gcc/ChangeLog

   1 2020-02-12  Wilco Dijkstra  <wdijkstr@arm.com>
   2
   3         * config/aarch64/aarch64-simd.md
   4         (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
   5         * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
   6         generating separate ADDV and zero_extend patterns.
   7         * config/aarch64/iterators.md (VDQV_E): New iterator.
   8
   9 2020-02-12  Jeff Law  <law@redhat.com>
  10
  11         * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
  12         expanders, splits, etc.
  13         (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
  14         (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
  15         (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
  16         * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
  17         (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
  18         * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
  19         function prototype.
  20         (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
  21
  22 2020-02-12  Jakub Jelinek  <jakub@redhat.com>
  23
  24         PR target/93670
  25         * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
  26         (avx512vl_vextractf128<mode>): Use it instead of VI48F_256.  Remove
  27         TARGET_AVX512DQ from condition.
  28         (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
  29         instead of <mask_mode512bit_condition> in condition.  If
  30         TARGET_AVX512DQ is false, emit vextract*64x4 instead of
  31         vextract*32x8.
  32         (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
  33         from condition.
  34
  35 2020-02-12  Kewen Lin  <linkw@gcc.gnu.org>
  36
  37         PR target/91052
  38         * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
  39
  40 2020-02-12  Segher Boessenkool  <segher@kernel.crashing.org>
  41
  42         * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
  43         where strlen is more legible.
  44         (rs6000_builtin_vectorized_libmass): Ditto.
  45         (rs6000_print_options_internal): Ditto.
  46
  47 2020-02-11  Martin Sebor  <msebor@redhat.com>
  48
  49         PR tree-optimization/93683
  50         * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
  51
  52 2020-02-11  Michael Meissner  <meissner@linux.ibm.com>
  53
  54         * config/rs6000/predicates.md (cint34_operand): Rename the
  55         -mprefixed-addr option to be -mprefixed.
  56         * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
  57         the -mprefixed-addr option to be -mprefixed.
  58         (OTHER_FUTURE_MASKS): Likewise.
  59         (POWERPC_MASKS): Likewise.
  60         * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
  61         the -mprefixed-addr option to be -mprefixed.  Change error
  62         messages to refer to -mprefixed.
  63         (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
  64         -mprefixed.
  65         (rs6000_legitimate_offset_address_p): Likewise.
  66         (rs6000_mode_dependent_address): Likewise.
  67         (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
  68         "-mprefixed" for target attributes and pragmas.
  69         (address_to_insn_form): Rename the -mprefixed-addr option to be
  70         -mprefixed.
  71         (rs6000_adjust_insn_length): Likewise.
  72         * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
  73         -mprefixed-addr option to be -mprefixed.
  74         (ASM_OUTPUT_OPCODE): Likewise.
  75         * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
  76         -mprefixed-addr option to be -mprefixed.
  77         * config/rs6000/rs6000.opt (-mprefixed): Rename the
  78         -mprefixed-addr option to be prefixed.  Change the option from
  79         being undocumented to being documented.
  80         * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
  81         -mprefixed option.  Update the -mpcrel documentation to mention
  82         -mprefixed.
  83
  84 2020-02-11  Hans-Peter Nilsson  <hp@axis.com>
  85
  86         * ira-conflicts.c (print_hard_reg_set): Correct output for sets
  87         including FIRST_PSEUDO_REGISTER - 1.
  88         * ira-color.c (print_hard_reg_set): Ditto.
  89
  90 2020-02-11  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
  91
  92         * config/arm/arm-builtins.c (enum arm_type_qualifiers):
  93         (USTERNOP_QUALIFIERS): New define.
  94         (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
  95         (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
  96         (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
  97         (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
  98         * config/arm/arm_neon.h (vusdot_s32): New.
  99         (vusdot_lane_s32): New.
 100         (vusdotq_lane_s32): New.
 101         (vsudot_lane_s32): New.
 102         (vsudotq_lane_s32): New.
 103         * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
 104         * config/arm/iterators.md (DOTPROD_I8MM): New.
 105         (sup, opsuffix): Add <us/su>.
 106         * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
 107         * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
 108
 109 2020-02-11  Richard Biener  <rguenther@suse.de>
 110
 111         PR tree-optimization/93661
 112         PR tree-optimization/93662
 113         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
 114         tree_to_poly_int64.
 115         * tree-sra.c (get_access_for_expr): Likewise.
 116
 117 2020-02-10  Jakub Jelinek  <jakub@redhat.com>
 118
 119         PR target/93637
 120         * config/i386/sse.md (VI_256_AVX2): New mode iterator.
 121         (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
 122         Change condition from TARGET_AVX2 to TARGET_AVX.
 123
 124 2020-02-10  Iain Sandoe  <iain@sandoe.co.uk>
 125
 126         PR other/93641
 127         * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
 128         argument of strncmp.
 129
 130 2020-02-10  Hans-Peter Nilsson  <hp@axis.com>
 131
 132         Try to generate zero-based comparisons.
 133         * config/cris/cris.c (cris_reduce_compare): New function.
 134         * config/cris/cris-protos.h  (cris_reduce_compare): Add prototype.
 135         * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
 136         (cstore<mode>4"): Apply cris_reduce_compare in expanders.
 137
 138 2020-02-10  Richard Earnshaw  <rearnsha@arm.com>
 139
 140         PR target/91913
 141         * config/arm/arm.md (movsi_compare0): Allow SP as a source register
 142         in Thumb state and also as a destination in Arm state.  Add T16
 143         variants.
 144
 145 2020-02-10  Hans-Peter Nilsson  <hp@axis.com>
 146
 147         * md.texi (Define Subst): Match closing paren in example.
 148
 149 2020-02-10  Jakub Jelinek  <jakub@redhat.com>
 150
 151         PR target/58218
 152         PR other/93641
 153         * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
 154         arguments of strncmp.
 155
 156 2020-02-10  Feng Xue  <fxue@os.amperecomputing.com>
 157
 158         PR ipa/93203
 159         * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
 160         but different source value.
 161         (adjust_callers_for_value_intersection): New function.
 162         (gather_edges_for_value): Adjust order of callers to let a
 163         non-self-recursive caller be the first element.
 164         (self_recursive_pass_through_p): Add a new parameter "simple", and
 165         check generalized self-recursive pass-through jump function.
 166         (self_recursive_agg_pass_through_p): Likewise.
 167         (find_more_scalar_values_for_callers_subset): Compute value from
 168         pass-through jump function for self-recursive.
 169         (intersect_with_plats): Cleanup previous implementation code for value
 170         itersection with self-recursive call edge.
 171         (intersect_with_agg_replacements): Likewise.
 172         (intersect_aggregates_with_edge): Deduce value from pass-through jump
 173         function for self-recursive call edge.  Cleanup previous implementation
 174         code for value intersection with self-recursive call edge.
 175         (decide_whether_version_node): Remove dead callers and adjust order
 176         to let a non-self-recursive caller be the first element.
 177
 178 2020-02-09  Uroš Bizjak  <ubizjak@gmail.com>
 179
 180         * recog.c: Move pass_split_before_sched2 code in front of
 181         pass_split_before_regstack.
 182         (pass_data_split_before_sched2): Rename pass to split3 from split4.
 183         (pass_data_split_before_regstack): Rename pass to split4 from split3.
 184         (rest_of_handle_split_before_sched2): Remove.
 185         (pass_split_before_sched2::execute): Unconditionally call
 186         split_all_insns.
 187         (enable_split_before_sched2): New function.
 188         (pass_split_before_sched2::gate): Use enable_split_before_sched2.
 189         (pass_split_before_regstack::gate): Ditto.
 190         * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
 191         Update name check for renamed split4 pass.
 192         * config/sh/sh.c (register_sh_passes): Update pass insertion
 193         point for renamed split4 pass.
 194
 195 2020-02-09  Jakub Jelinek  <jakub@redhat.com>
 196
 197         * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
 198         DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
 199         copying them around between host and target.
 200
 201 2020-02-08  Andrew Pinski  <apinski@marvell.com>
 202
 203         PR target/91927
 204         * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
 205         STRICT_ALIGNMENT also.
 206
 207 2020-02-08  Jim Wilson  <jimw@sifive.com>
 208
 209         PR target/93532
 210         * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
 211
 212 2020-02-08  Uroš Bizjak  <ubizjak@gmail.com>
 213             Jakub Jelinek  <jakub@redhat.com>
 214
 215         PR target/65782
 216         * config/i386/i386.h (CALL_USED_REGISTERS): Make
 217         xmm16-xmm31 call-used even in 64-bit ms-abi.
 218
 219 2020-02-07  Dennis Zhang  <dennis.zhang@arm.com>
 220
 221         * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
 222         (simd_ummla, simd_usmmla): Likewise.
 223         * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
 224         * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
 225         (vusmmlaq_s32): New.
 226
 227 2020-02-07  Richard Biener  <rguenther@suse.de>
 228
 229         PR middle-end/93519
 230         * tree-inline.c (fold_marked_statements): Do a PRE walk,
 231         skipping unreachable regions.
 232         (optimize_inline_calls): Skip folding stmts when we didn't
 233         inline.
 234
 235 2020-02-07  H.J. Lu  <hongjiu.lu@intel.com>
 236
 237         PR target/85667
 238         * config/i386/i386.c (function_arg_ms_64): Add a type argument.
 239         Don't return aggregates with only SFmode and DFmode in SSE
 240         register.
 241         (ix86_function_arg): Pass arg.type to function_arg_ms_64.
 242
 243 2020-02-07  Jakub Jelinek  <jakub@redhat.com>
 244
 245         PR target/93122
 246         * config/rs6000/rs6000-logue.c
 247         (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
 248         if it fails, move rs into end_addr and retry.  Add
 249         REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
 250         the insn pattern doesn't describe well what exactly happens to
 251         dwarf2cfi.c.
 252
 253         PR target/93594
 254         * config/i386/predicates.md (avx_identity_operand): Remove.
 255         * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
 256         (avx_<castmode><avxsizesuffix>_<castmode>,
 257         avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
 258         a VEC_CONCAT of the operand and UNSPEC_CAST.
 259         (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
 260         a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
 261         UNSPEC_CAST.
 262
 263         PR target/93611
 264         * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
 265         recog_data.insn if distance_non_agu_define changed it.
 266
 267 2020-02-06  Michael Meissner  <meissner@linux.ibm.com>
 268
 269         PR target/93569
 270         * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
 271         we only had X-FORM (reg+reg) addressing for vectors.  Also before
 272         ISA 3.0, we only had X-FORM addressing for scalars in the
 273         traditional Altivec registers.
 274
 275 2020-02-06  <zhongyunde@huawei.com>
 276             Vladimir Makarov  <vmakarov@redhat.com>
 277
 278         PR rtl-optimization/93561
 279         * lra-assigns.c (spill_for): Check that tested hard regno is not out of
 280         hard register range.
 281
 282 2020-02-06  Richard Sandiford  <richard.sandiford@arm.com>
 283
 284         * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
 285         attribute.
 286
 287 2020-02-06  Segher Boessenkool  <segher@kernel.crashing.org>
 288
 289         * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
 290         where the low and the high 32 bits are equal to each other specially,
 291         with an rldimi instruction.
 292
 293 2020-02-06  Mihail Ionescu  <mihail.ionescu@arm.com>
 294
 295         * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
 296
 297 2020-02-06  Mihail Ionescu  <mihail.ionescu@arm.com>
 298
 299         * config/arm/arm-tables.opt: Regenerate.
 300
 301 2020-02-06  Richard Sandiford  <richard.sandiford@arm.com>
 302
 303         PR target/87763
 304         * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
 305         * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
 306         * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
 307
 308 2020-02-06  Richard Sandiford  <richard.sandiford@arm.com>
 309
 310         PR rtl-optimization/87763
 311         * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
 312
 313 2020-02-06  Delia Burduv  <delia.burduv@arm.com>
 314
 315         * config/aarch64/aarch64-simd-builtins.def
 316         (bfmlaq): New built-in function.
 317         (bfmlalb): New built-in function.
 318         (bfmlalt): New built-in function.
 319         (bfmlalb_lane): New built-in function.
 320         (bfmlalt_lane): New built-in function.
 321         * config/aarch64/aarch64-simd.md
 322         (aarch64_bfmmlaqv4sf): New pattern.
 323         (aarch64_bfmlal<bt>v4sf): New pattern.
 324         (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
 325         * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
 326         (vbfmlalbq_f32): New intrinsic.
 327         (vbfmlaltq_f32): New intrinsic.
 328         (vbfmlalbq_lane_f32): New intrinsic.
 329         (vbfmlaltq_lane_f32): New intrinsic.
 330         (vbfmlalbq_laneq_f32): New intrinsic.
 331         (vbfmlaltq_laneq_f32): New intrinsic.
 332         * config/aarch64/iterators.md (BF_MLA): New int iterator.
 333         (bt): New int attribute.
 334
 335 2020-02-06  Uroš Bizjak  <ubizjak@gmail.com>
 336
 337         * config/i386/i386.md (*pushtf): Emit "#" instead of
 338         calling gcc_unreachable in insn output.
 339         (*pushxf): Ditto.
 340         (*pushdf): Ditto.
 341         (*pushsf_rex64): Ditto for alternatives other than 1.
 342         (*pushsf): Ditto for alternatives other than 1.
 343
 344 2020-02-06  Martin Liska  <mliska@suse.cz>
 345
 346         PR gcov-profile/91971
 347         PR gcov-profile/93466
 348         * coverage.c (coverage_init): Revert mangling of
 349         path into filename.  It can lead to huge filename length.
 350         Creation of subfolders seem more natural.
 351
 352 2020-02-06  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
 353
 354         PR target/93300
 355         * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
 356         (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
 357         Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
 358
 359 2020-02-06  Jakub Jelinek  <jakub@redhat.com>
 360
 361         PR target/93594
 362         * config/i386/predicates.md (avx_identity_operand): New predicate.
 363         * config/i386/sse.md (*avx_vec_concat<mode>_1): New
 364         define_insn_and_split.
 365
 366         PR libgomp/93515
 367         * omp-low.c (use_pointer_for_field): For nested constructs, also
 368         look for map clauses on target construct.
 369         (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
 370         taskreg_nesting_level.
 371
 372         PR libgomp/93515
 373         * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
 374         shared clause, call omp_notice_variable on outer context if any.
 375
 376 2020-02-05  Jason Merrill  <jason@redhat.com>
 377
 378         PR c++/92003
 379         * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
 380         non-zero address even if weak and not yet defined.
 381
 382 2020-02-05  Martin Sebor  <msebor@redhat.com>
 383
 384         PR tree-optimization/92765
 385         * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
 386         * tree-ssa-strlen.c (compute_string_length): Remove.
 387         (determine_min_objsize): Remove.
 388         (get_len_or_size): Add an argument.  Call get_range_strlen_dynamic.
 389         Avoid using type size as the upper bound on string length.
 390         (handle_builtin_string_cmp): Add an argument.  Adjust.
 391         (strlen_check_and_optimize_call): Pass additional argument to
 392         handle_builtin_string_cmp.
 393
 394 2020-02-05  Uroš Bizjak  <ubizjak@gmail.com>
 395
 396         * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
 397         (*pushdi2_rex64 peephole2): Unconditionally split after
 398         epilogue_completed.
 399         (*ashl<mode>3_doubleword): Ditto.
 400         (*<shift_insn><mode>3_doubleword): Ditto.
 401
 402 2020-02-05  Michael Meissner  <meissner@linux.ibm.com>
 403
 404         PR target/93568
 405         * config/rs6000/rs6000.c (get_vector_offset): Fix
 406
 407 2020-02-05  Andrew Stubbs  <ams@codesourcery.com>
 408
 409         * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
 410
 411 2020-02-05  David Malcolm  <dmalcolm@redhat.com>
 412
 413         * doc/analyzer.texi
 414         (Special Functions for Debugging the Analyzer): Update description
 415         of __analyzer_dump_exploded_nodes.
 416
 417 2020-02-05  Jakub Jelinek  <jakub@redhat.com>
 418
 419         PR target/92190
 420         * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
 421         include sets and not clobbers in the vzeroupper pattern.
 422         * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
 423         the parallel has 17 (64-bit) or 9 (32-bit) elts.
 424         (*avx_vzeroupper_1): New define_insn_and_split.
 425
 426         PR target/92190
 427         * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
 428         don't run when !optimize.
 429         (pass_split_before_regstack::gate): For STACK_REGS targets, run even
 430         when !optimize.
 431
 432 2020-02-05  Richard Biener  <rguenther@suse.de>
 433
 434         PR middle-end/90648
 435         * genmatch.c (dt_node::gen_kids_1): Emit number of argument
 436         checks before matching calls.
 437
 438 2020-02-05  Jakub Jelinek  <jakub@redhat.com>
 439
 440         * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
 441         function comment typo.
 442
 443         PR middle-end/93555
 444         * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
 445         simd_clone_create failed when i == 0, adjust clone->nargs by
 446         clone->inbranch.
 447
 448 2020-02-05  Martin Liska  <mliska@suse.cz>
 449
 450         PR c++/92717
 451         * doc/invoke.texi: Document that one should
 452         not combine ASLR and -fpch.
 453
 454 2020-02-04  Richard Biener  <rguenther@suse.de>
 455
 456         PR tree-optimization/93538
 457         * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
 458
 459 2020-02-04  Richard Biener  <rguenther@suse.de>
 460
 461         PR tree-optimization/91123
 462         * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
 463         (vn_walk_cb_data::last_vuse): New member.
 464         (vn_walk_cb_data::saved_operands): Likewsie.
 465         (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
 466         (vn_walk_cb_data::push_partial_def): Use finish.
 467         (vn_reference_lookup_2): Update last_vuse and use finish if
 468         we've saved operands.
 469         (vn_reference_lookup_3): Use finish and update calls to
 470         push_partial_defs everywhere.  When translating through
 471         memcpy or aggregate copies save off operands and alias-set.
 472         (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
 473         operation for redundant store removal.
 474
 475 2020-02-04  Richard Biener  <rguenther@suse.de>
 476
 477         PR tree-optimization/92819
 478         * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
 479         generating more stmts than before.
 480
 481 2020-02-04  Martin Liska  <mliska@suse.cz>
 482
 483         * config/arm/arm.c (arm_gen_far_branch): Move the function
 484         outside of selftests.
 485
 486 2020-02-03  Michael Meissner  <meissner@linux.ibm.com>
 487
 488         * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
 489         function to adjust PC-relative vector addresses.
 490         (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
 491         handle vectors with PC-relative addresses.
 492
 493 2020-02-03  Michael Meissner  <meissner@linux.ibm.com>
 494
 495         * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
 496         reference.
 497         (hard_reg_and_mode_to_addr_mask): Delete.
 498         (rs6000_adjust_vec_address): If the original vector address
 499         was REG+REG or REG+OFFSET and the element is not zero, do the add
 500         of the elements in the original address before adding the offset
 501         for the vector element.  Use address_to_insn_form to validate the
 502         address using the register being loaded, rather than guessing
 503         whether the address is a DS-FORM or DQ-FORM address.
 504
 505 2020-02-03  Michael Meissner  <meissner@linux.ibm.com>
 506
 507         * config/rs6000/rs6000.c (get_vector_offset): New helper function
 508         to calculate the offset in memory from the start of a vector of a
 509         particular element.  Add code to keep the element number in
 510         bounds if the element number is variable.
 511         (rs6000_adjust_vec_address): Move calculation of offset of the
 512         vector element to get_vector_offset.
 513         (rs6000_split_vec_extract_var): Do not do the initial AND of
 514         element here, move the code to get_vector_offset.
 515
 516 2020-02-03  Michael Meissner  <meissner@linux.ibm.com>
 517
 518         * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
 519         gcc_asserts.
 520
 521 2020-02-03  Segher Boessenkool  <segher@kernel.crashing.org>
 522
 523         * config/rs6000/constraints.md: Improve documentation.
 524
 525 2020-02-03  Richard Earnshaw  <rearnsha@arm.com>
 526
 527         PR target/93548
 528         * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
 529         ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
 530
 531 2020-02-03  Andrew Stubbs  <ams@codesourcery.com>
 532
 533         * config.gcc: Remove "carrizo" support.
 534         * config/gcn/gcn-opts.h (processor_type): Likewise.
 535         * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
 536         * config/gcn/gcn.opt (gpu_type): Likewise.
 537         * config/gcn/t-omp-device: Likewise.
 538
 539 2020-02-03  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
 540
 541         PR target/91816
 542         * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
 543         * config/arm/arm.c (arm_gen_far_branch): New function
 544         arm_gen_far_branch.
 545         * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
 546
 547 2020-02-03  Julian Brown  <julian@codesourcery.com>
 548             Tobias Burnus  <tobias@codesourcery.com>
 549
 550         * doc/invoke.texi: Update mention of OpenACC version to 2.6.
 551
 552 2020-02-03  Jakub Jelinek  <jakub@redhat.com>
 553
 554         PR target/93533
 555         * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
 556         valid RTL to sum up the lowest and second lowest bytes of the popcnt
 557         result.
 558
 559 2020-02-02  Vladimir Makarov  <vmakarov@redhat.com>
 560
 561         PR rtl-optimization/91333
 562         * ira-color.c (struct allocno_color_data): Add member
 563         hard_reg_prefs.
 564         (init_allocno_threads): Set the member up.
 565         (bucket_allocno_compare_func): Add compare hard reg
 566         prefs.
 567
 568 2020-01-31  Sandra Loosemore  <sandra@codesourcery.com>
 569
 570         nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
 571
 572         * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
 573         * config.in: Regenerated.
 574         * configure: Regenerated.
 575         * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
 576         for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
 577         (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
 578
 579 2020-02-01  Andrew Burgess  <andrew.burgess@embecosm.com>
 580
 581         * configure: Regenerate.
 582
 583 2020-01-31  Vladimir Makarov  <vmakarov@redhat.com>
 584
 585         PR rtl-optimization/91333
 586         * ira-color.c (bucket_allocno_compare_func): Move conflict hard
 587         reg preferences comparison up.
 588
 589 2020-01-31  Richard Sandiford  <richard.sandiford@arm.com>
 590
 591         * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
 592         * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
 593         aarch64-sve-builtins-base.h.
 594         * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
 595         aarch64-sve-builtins-base.cc.
 596         * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
 597         (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
 598         (svcvtnt): Declare.
 599         * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
 600         (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
 601         (svcvtnt): New functions.
 602         * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
 603         (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
 604         (svcvtnt): New functions.
 605         (svcvt): Add a form that converts f32 to bf16.
 606         * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
 607         (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
 608         Declare.
 609         * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
 610         Treat B as bfloat16_t.
 611         (ternary_bfloat_lane_base): New class.
 612         (ternary_bfloat_def): Likewise.
 613         (ternary_bfloat): New shape.
 614         (ternary_bfloat_lane_def): New class.
 615         (ternary_bfloat_lane): New shape.
 616         (ternary_bfloat_lanex2_def): New class.
 617         (ternary_bfloat_lanex2): New shape.
 618         (ternary_bfloat_opt_n_def): New class.
 619         (ternary_bfloat_opt_n): New shape.
 620         * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
 621         * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
 622         (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
 623         (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
 624         (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
 625         (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
 626         (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
 627         * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
 628         the pattern off the narrow mode instead of the wider one.
 629         * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
 630         (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
 631         (sve_fp_op): Handle them.
 632         (SVE_BFLOAT_TERNARY_LONG): New int itertor.
 633         (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
 634
 635 2020-01-31  Richard Sandiford  <richard.sandiford@arm.com>
 636
 637         * config/aarch64/arm_sve.h: Include arm_bf16.h.
 638         * config/aarch64/aarch64-modes.def (BF): Move definition before
 639         VECTOR_MODES.  Remove separate VECTOR_MODES for V4BF and V8BF.
 640         (SVE_MODES): Handle BF modes.
 641         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
 642         BF modes.
 643         (aarch64_full_sve_mode): Likewise.
 644         * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
 645         and VNx32BF.
 646         (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
 647         (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
 648         (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
 649         (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
 650         new SVE BF modes.
 651         * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
 652         type_class_index.
 653         * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
 654         (TYPES_all_data): Add bf16.
 655         (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
 656         (register_tuple_type): Increase buffer size.
 657         * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
 658         (bf16): New type suffix.
 659         * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
 660         (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
 661         (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
 662         Change type from all_data to all_arith.
 663         * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
 664         (svminp): Likewise.
 665
 666 2020-01-31  Dennis Zhang  <dennis.zhang@arm.com>
 667             Matthew Malcomson  <matthew.malcomson@arm.com>
 668             Richard Sandiford  <richard.sandiford@arm.com>
 669
 670         * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
 671         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
 672         __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
 673         __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate.  Don't define
 674         __ARM_FEATURE_MATMUL_FP64.
 675         * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
 676         (sve): Add AARCH64_FL_F32MM to the list of extensions that should
 677         be disabled at the same time.
 678         (f32mm): New extension.
 679         * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
 680         (AARCH64_FL_F64MM): Bump to the next bit up.
 681         (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
 682         (TARGET_SVE_F64MM): New macros.
 683         * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
 684         (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
 685         (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
 686         (UNSPEC_ZIP2Q): New unspeccs.
 687         (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
 688         (optab, sur, perm_insn): Handle the new unspecs.
 689         (sve_fp_op): Handle UNSPEC_FMMLA.  Resort.
 690         * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
 691         TARGET_SVE_F64MM instead of separate tests.
 692         (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
 693         (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
 694         (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
 695         (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
 696         (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
 697         * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
 698         (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
 699         (TYPES_s_signed): New macro.
 700         (TYPES_s_integer): Use it.
 701         (TYPES_d_float): New macro.
 702         (TYPES_d_data): Use it.
 703         * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
 704         (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
 705         (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
 706         * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
 707         (svmmla): New shape.
 708         (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
 709         template parameters.
 710         (ternary_resize2_lane_base): Likewise.
 711         (ternary_resize2_base): New class.
 712         (ternary_qq_lane_base): Likewise.
 713         (ternary_intq_uintq_lane_def): Likewise.
 714         (ternary_intq_uintq_lane): New shape.
 715         (ternary_intq_uintq_opt_n_def): New class
 716         (ternary_intq_uintq_opt_n): New shape.
 717         (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
 718         (ternary_uintq_intq_def): New class.
 719         (ternary_uintq_intq): New shape.
 720         (ternary_uintq_intq_lane_def): New class.
 721         (ternary_uintq_intq_lane): New shape.
 722         (ternary_uintq_intq_opt_n_def): New class.
 723         (ternary_uintq_intq_opt_n): New shape.
 724         * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
 725         (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
 726         (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
 727         * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
 728         Generalize to...
 729         (svdotprod_lane_impl): ...this new class.
 730         (svmmla_impl, svusdot_impl): New classes.
 731         (svdot_lane): Update to use svdotprod_lane_impl.
 732         (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
 733         (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
 734         functions.
 735         * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
 736         function, with no types defined.
 737         (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
 738         AARCH64_FL_I8MM functions.
 739         (svmmla): New AARCH64_FL_F32MM function.
 740         (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
 741         (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
 742         AARCH64_FL_F64MM function.
 743         (REQUIRED_EXTENSIONS):
 744
 745 2020-01-31  Andrew Stubbs  <ams@codesourcery.com>
 746
 747         * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
 748         alternative only.
 749
 750 2020-01-31  Uroš Bizjak  <ubizjak@gmail.com>
 751
 752         * config/i386/i386.md (*movoi_internal_avx): Do not check for
 753         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.  Remove MODE_V8SF handling.
 754         (*movti_internal): Do not check for
 755         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
 756         (*movtf_internal): Move check for TARGET_SSE2 and size optimization
 757         just after check for TARGET_AVX.
 758         (*movdf_internal): Ditto.
 759         * config/i386/mmx.md (*mov<mode>_internal): Do not check for
 760         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
 761         * config/i386/sse.md (mov<mode>_internal): Only check
 762         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode.  Move check
 763         for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
 764         (<sse>_andnot<mode>3<mask_name>): Move check for
 765         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
 766         (<code><mode>3<mask_name>): Ditto.
 767         (*andnot<mode>3): Ditto.
 768         (*andnottf3): Ditto.
 769         (*<code><mode>3): Ditto.
 770         (*<code>tf3): Ditto.
 771         (*andnot<VI:mode>3): Remove
 772         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
 773         (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
 774         (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
 775         (sse4_1_blendv<ssemodesuffix>): Ditto.
 776         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
 777         Explain that tune applies to 128bit instructions only.
 778
 779 2020-01-31  Kwok Cheung Yeung  <kcy@codesourcery.com>
 780
 781         * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
 782         to definition of hsa_kernel_description.  Parse assembly to find SGPR
 783         and VGPR count of kernel and store in hsa_kernel_description.
 784
 785 2020-01-31  Tamar Christina  <tamar.christina@arm.com>
 786
 787         PR rtl-optimization/91838
 788         * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
 789         to truncate if allowed or reject combination.
 790
 791 2020-01-31  Andrew Stubbs  <ams@codesourcery.com>
 792
 793         * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
 794         (find_inv_vars_cb): Likewise.
 795
 796 2020-01-31  David Malcolm  <dmalcolm@redhat.com>
 797
 798         * calls.c (special_function_p): Split out the check for DECL_NAME
 799         being non-NULL and fndecl being extern at file scope into a
 800         new maybe_special_function_p and call it.  Drop check for fndecl
 801         being non-NULL that was after a usage of DECL_NAME (fndecl).
 802         * tree.h (maybe_special_function_p): New inline function.
 803
 804 2020-01-30  Andrew Stubbs  <ams@codesourcery.com>
 805
 806         * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
 807         (mask_gather_load<mode>): ... here, and zero-initialize the
 808         destination.
 809         (maskload<mode>di): Zero-initialize the destination.
 810         * config/gcn/gcn.c:
 811
 812 2020-01-30  David Malcolm  <dmalcolm@redhat.com>
 813
 814         PR analyzer/93356
 815         * doc/analyzer.texi (Limitations): Note that constraints on
 816         floating-point values are currently ignored.
 817
 818 2020-01-30  Jakub Jelinek  <jakub@redhat.com>
 819
 820         PR lto/93384
 821         * symtab.c (symtab_node::noninterposable_alias): If localalias
 822         already exists, but is not usable, append numbers after it until
 823         a unique name is found.  Formatting fix.
 824
 825         PR middle-end/93505
 826         * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
 827         rotate counts.
 828
 829 2020-01-30  Andrew Stubbs  <ams@codesourcery.com>
 830
 831         * config/gcn/gcn.c (print_operand): Handle LTGT.
 832         * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
 833
 834 2020-01-30  Richard Biener  <rguenther@suse.de>
 835
 836         * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
 837         and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
 838
 839 2020-01-30  John David Anglin  <danglin@gcc.gnu.org>
 840
 841         * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
 842         without a DECL in .data.rel.ro.local.
 843
 844 2020-01-30  Jakub Jelinek  <jakub@redhat.com>
 845
 846         PR target/93494
 847         * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
 848         returned.
 849
 850         PR target/91824
 851         * config/i386/sse.md
 852         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
 853         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this.  Use
 854         any_extend code iterator instead of always zero_extend.
 855         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
 856         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
 857         Use any_extend code iterator instead of always zero_extend.
 858         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
 859         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
 860         Use any_extend code iterator instead of always zero_extend.
 861         (*sse2_pmovmskb_ext): New define_insn.
 862         (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
 863
 864         PR target/91824
 865         * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
 866         (*popcountsi2_zext_falsedep): New define_insn.
 867
 868 2020-01-30  Dragan Mladjenovic  <dmladjenovic@wavecomp.com>
 869
 870         * config.in: Regenerated.
 871         * configure: Regenerated.
 872
 873 2020-01-29  Tobias Burnus  <tobias@codesourcery.com>
 874
 875         PR bootstrap/93409
 876         * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
 877         LLVM's assembler changed the default in version 9.
 878
 879 2020-01-24  Jeff Law  <law@redhat.com>
 880
 881         PR tree-optimization/89689
 882         * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
 883
 884 2020-01-29  Richard Sandiford  <richard.sandiford@arm.com>
 885
 886         Revert:
 887
 888         2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
 889
 890         PR rtl-optimization/87763
 891         * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
 892         simplification to handle subregs as well as bare regs.
 893         * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
 894
 895 2020-01-29  Joel Hutton  <Joel.Hutton@arm.com>
 896
 897         PR target/93221
 898         * ira.c (ira): Revert use of simplified LRA algorithm.
 899
 900 2020-01-29  Martin Jambor  <mjambor@suse.cz>
 901
 902         PR tree-optimization/92706
 903         * tree-sra.c (struct access): Fields first_link, last_link,
 904         next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
 905         next_rhs_queued and grp_rhs_queued respectively, new fields
 906         first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
 907         (struct assign_link): Field next renamed to next_rhs, new field
 908         next_lhs.  Updated comment.
 909         (work_queue_head): Renamed to rhs_work_queue_head.
 910         (lhs_work_queue_head): New variable.
 911         (add_link_to_lhs): New function.
 912         (relink_to_new_repr): Also relink LHS lists.
 913         (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
 914         (add_access_to_lhs_work_queue): New function.
 915         (pop_access_from_work_queue): Renamed to
 916         pop_access_from_rhs_work_queue.
 917         (pop_access_from_lhs_work_queue): New function.
 918         (build_accesses_from_assign): Also add links to LHS lists and to LHS
 919         work_queue.
 920         (child_would_conflict_in_lacc): Renamed to
 921         child_would_conflict_in_acc.  Adjusted parameter names.
 922         (create_artificial_child_access): New parameter set_grp_read, use it.
 923         (subtree_mark_written_and_enqueue): Renamed to
 924         subtree_mark_written_and_rhs_enqueue.
 925         (propagate_subaccesses_across_link): Renamed to
 926         propagate_subaccesses_from_rhs.
 927         (propagate_subaccesses_from_lhs): New function.
 928         (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
 929         RHSs.
 930
 931 2020-01-29  Martin Jambor  <mjambor@suse.cz>
 932
 933         PR tree-optimization/92706
 934         * tree-sra.c (struct access): Adjust comment of
 935         grp_total_scalarization.
 936         (find_access_in_subtree): Look for single children spanning an entire
 937         access.
 938         (scalarizable_type_p): Allow register accesses, adjust callers.
 939         (completely_scalarize): Remove function.
 940         (scalarize_elem): Likewise.
 941         (create_total_scalarization_access): Likewise.
 942         (sort_and_splice_var_accesses): Do not track total scalarization
 943         flags.
 944         (analyze_access_subtree): New parameter totally, adjust to new meaning
 945         of grp_total_scalarization.
 946         (analyze_access_trees): Pass new parameter to analyze_access_subtree.
 947         (can_totally_scalarize_forest_p): New function.
 948         (create_total_scalarization_access): Likewise.
 949         (create_total_access_and_reshape): Likewise.
 950         (total_should_skip_creating_access): Likewise.
 951         (totally_scalarize_subtree): Likewise.
 952         (analyze_all_variable_accesses): Perform total scalarization after
 953         subaccess propagation using the new functions above.
 954         (initialize_constant_pool_replacements): Output initializers by
 955         traversing the access tree.
 956
 957 2020-01-29  Martin Jambor  <mjambor@suse.cz>
 958
 959         * tree-sra.c (verify_sra_access_forest): New function.
 960         (verify_all_sra_access_forests): Likewise.
 961         (create_artificial_child_access): Set parent.
 962         (analyze_all_variable_accesses): Call the verifier.
 963
 964 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
 965
 966         * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
 967         if called on indirect edge.
 968         (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
 969         speculative call if needed.
 970
 971 2020-01-29  Richard Biener  <rguenther@suse.de>
 972
 973         PR tree-optimization/93428
 974         * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
 975         permutation when the load node is created.
 976         (vect_analyze_slp_instance): Re-use it here.
 977
 978 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
 979
 980         * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
 981
 982 2020-01-28  Vladimir Makarov  <vmakarov@redhat.com>
 983
 984         PR rtl-optimization/93272
 985         * ira-lives.c (process_out_of_region_eh_regs): New function.
 986         (process_bb_node_lives): Call it.
 987
 988 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
 989
 990         * coverage.c (read_counts_file): Make error message lowercase.
 991
 992 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
 993
 994         * profile-count.c (profile_quality_display_names): Fix ordering.
 995
 996 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
 997
 998         PR lto/93318
 999         * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1000         hash only when edge is first within the sequence.
1001         (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1002         (symbol_table::create_edge): Do not set target_prob.
1003         (cgraph_edge::remove_caller): Watch for speculative calls when updating
1004         the call site hash.
1005         (cgraph_edge::make_speculative): Drop target_prob parameter.
1006         (cgraph_edge::speculative_call_info): Remove.
1007         (cgraph_edge::first_speculative_call_target): New member function.
1008         (update_call_stmt_hash_for_removing_direct_edge): New function.
1009         (cgraph_edge::resolve_speculation): Rewrite to new API.
1010         (cgraph_edge::speculative_call_for_target): New member function.
1011         (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1012         multiple speculation targets.
1013         (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1014         of profile.
1015         (verify_speculative_call): Verify that targets form an interval.
1016         * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1017         (cgraph_edge::first_speculative_call_target): New member function.
1018         (cgraph_edge::next_speculative_call_target): New member function.
1019         (cgraph_edge::speculative_call_target_ref): New member function.
1020         (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1021         (cgraph_edge): Remove target_prob.
1022         * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1023         Fix handling of speculative calls.
1024         * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1025         * ipa-fnsummary.c (analyze_function_body): Likewise.
1026         * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1027         * ipa-profile.c (dump_histogram): Fix formating.
1028         (ipa_profile_generate_summary): Watch for overflows.
1029         (ipa_profile): Do not require probablity to be 1/2; update to new API.
1030         * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1031         (update_indirect_edges_after_inlining): Update to new API.
1032         * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1033         profiles.
1034         * profile-count.h: (profile_probability::adjusted): New.
1035         * tree-inline.c (copy_bb): Update to new speculative call API; fix
1036         updating of profile.
1037         * value-prof.c (gimple_ic_transform): Rename to ...
1038         (dump_ic_profile): ... this one; update dumping.
1039         (stream_in_histogram_value): Fix formating.
1040         (gimple_value_profile_transformations): Update.
1041
1042 2020-01-28  H.J. Lu  <hongjiu.lu@intel.com>
1043
1044         PR target/91461
1045         * config/i386/i386.md (*movoi_internal_avx): Remove
1046         TARGET_SSE_TYPELESS_STORES check.
1047         (*movti_internal): Prefer TARGET_AVX over
1048         TARGET_SSE_TYPELESS_STORES.
1049         (*movtf_internal): Likewise.
1050         * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1051         TARGET_SSE_TYPELESS_STORES.  Remove "<MODE_SIZE> == 16" check
1052         from TARGET_SSE_TYPELESS_STORES.
1053
1054 2020-01-28  David Malcolm  <dmalcolm@redhat.com>
1055
1056         * diagnostic-core.h (warning_at): Rename overload to...
1057         (warning_meta): ...this.
1058         (emit_diagnostic_valist): Delete decl of overload taking
1059         diagnostic_metadata.
1060         * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1061         (warning_at): Rename overload taking diagnostic_metadata to...
1062         (warning_meta): ...this.
1063
1064 2020-01-28  Richard Biener  <rguenther@suse.de>
1065
1066         PR tree-optimization/93439
1067         * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1068         * tree-cfg.c (move_sese_region_to_fn): ... here.
1069         (verify_types_in_gimple_reference): Verify used cliques are
1070         tracked.
1071
1072 2020-01-28  H.J. Lu  <hongjiu.lu@intel.com>
1073
1074         PR target/91399
1075         * config/i386/i386-options.c (set_ix86_tune_features): Add an
1076         argument of a pointer to struct gcc_options and pass it to
1077         parse_mtune_ctrl_str.
1078         (ix86_function_specific_restore): Pass opts to
1079         set_ix86_tune_features.
1080         (ix86_option_override_internal): Likewise.
1081         (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1082         gcc_options and use it for x_ix86_tune_ctrl_string.
1083
1084 2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
1085
1086         PR rtl-optimization/87763
1087         * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1088         simplification to handle subregs as well as bare regs.
1089         * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1090
1091 2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
1092
1093         * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1094         for reduction chains that (now) include a call.
1095
1096 2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
1097
1098         PR tree-optimization/92822
1099         * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1100         out the don't-care elements of a vector whose significant elements
1101         are duplicates, make the don't-care elements duplicates too.
1102
1103 2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
1104
1105         PR tree-optimization/93434
1106         * tree-predcom.c (split_data_refs_to_components): Record which
1107         components have had aliasing loads removed.  Prevent store-store
1108         commoning for all such components.
1109
1110 2020-01-28  Jakub Jelinek  <jakub@redhat.com>
1111
1112         PR target/93418
1113         * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1114         -1 or is_vshift is true, use new_vector with number of elts npatterns
1115         rather than new_unary_operation.
1116
1117         PR tree-optimization/93454
1118         * gimple-fold.c (fold_array_ctor_reference): Perform
1119         elt_size.to_uhwi () just once, instead of calling it in every
1120         iteration.  Punt if that value is above size of the temporary
1121         buffer.  Decrease third native_encode_expr argument when
1122         bufoff + elt_sz is above size of buf.
1123
1124 2020-01-27  Joseph Myers  <joseph@codesourcery.com>
1125
1126         * config/mips/mips.c (mips_declare_object_name)
1127         [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1128
1129 2020-01-27  Martin Liska  <mliska@suse.cz>
1130
1131         PR gcov-profile/93403
1132         * tree-profile.c (gimple_init_gcov_profiler): Generate
1133         both __gcov_indirect_call_profiler_v4 and
1134         __gcov_indirect_call_profiler_v4_atomic.
1135
1136 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
1137
1138         PR target/92822
1139         * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1140         expander.
1141         (@aarch64_split_simd_mov<mode>): Use it.
1142         (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1143         Leave the vec_extract patterns to handle 2-element vectors.
1144         (aarch64_simd_mov_from_<mode>high): Likewise.
1145         (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1146         (vec_extractv2dfv1df): Likewise.
1147
1148 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
1149
1150         * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1151         jump conditions for *compare_condjump<GPI:mode>.
1152
1153 2020-01-27  David Malcolm  <dmalcolm@redhat.com>
1154
1155         PR analyzer/93276
1156         * digraph.cc (test_edge::test_edge): Specify template for base
1157         class initializer.
1158
1159 2020-01-27  Claudiu Zissulescu  <claziss@synopsys.com>
1160
1161         * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1162
1163 2020-01-27  Claudiu Zissulescu  <claziss@synopsys.com>
1164
1165         * config/arc/arc-protos.h (gen_mlo): Remove.
1166         (gen_mhi): Likewise.
1167         * config/arc/arc.c (AUX_MULHI): Define.
1168         (arc_must_save_reister): Special handling for r58/59.
1169         (arc_compute_frame_size): Consider mlo/mhi registers.
1170         (arc_save_callee_saves): Emit fp/sp move only when emit_move
1171         paramter is true.
1172         (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1173         mlo/mhi name selection.
1174         (arc_restore_callee_saves): Don't early restore blink when ISR.
1175         (arc_expand_prologue): Add mlo/mhi saving.
1176         (arc_expand_epilogue): Add mlo/mhi restoring.
1177         (gen_mlo): Remove.
1178         (gen_mhi): Remove.
1179         * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1180         numbering when MUL64 option is used.
1181         (DWARF2_FRAME_REG_OUT): Define.
1182         * config/arc/arc.md (arc600_stall): New pattern.
1183         (VUNSPEC_ARC_ARC600_STALL): Define.
1184         (mulsi64): Use correct mlo/mhi registers.
1185         (mulsi_600): Clean it up.
1186         * config/arc/predicates.md (mlo_operand): Remove any dependency on
1187         TARGET_BIG_ENDIAN.
1188         (mhi_operand): Likewise.
1189
1190 2020-01-27  Claudiu Zissulescu  <claziss@synopsys.com>
1191             Petro Karashchenko  <petro.karashchenko@ring.com>
1192
1193         * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1194         attributes if needed.
1195         (prepare_move_operands): Generate special unspec instruction for
1196         direct access.
1197         (arc_isuncached_mem_p): Propagate uncached attribute to each
1198         structure member.
1199         * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1200         (VUNSPEC_ARC_STDI): Likewise.
1201         (ALLI): New mode iterator.
1202         (mALLI): New mode attribute.
1203         (lddi): New instruction pattern.
1204         (stdi): Likewise.
1205         (stdidi_split): Split instruction for architectures which are not
1206         supporting ll64 option.
1207         (lddidi_split): Likewise.
1208
1209 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
1210
1211         PR rtl-optimization/92989
1212         * lra-lives.c (process_bb_lives): Update the live-in set before
1213         processing additional clobbers.
1214
1215 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
1216
1217         PR rtl-optimization/93170
1218         * cselib.c (cselib_invalidate_regno_val): New function, split out
1219         from...
1220         (cselib_invalidate_regno): ...here.
1221         (cselib_invalidated_by_call_p): New function.
1222         (cselib_process_insn): Iterate over all the hard-register entries in
1223         REG_VALUES and invalidate any that cross call-clobbered registers.
1224
1225 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
1226
1227         * dojump.c (split_comparison): Use HONOR_NANS rather than
1228         HONOR_SNANS when splitting LTGT.
1229
1230 2020-01-27  Martin Liska  <mliska@suse.cz>
1231
1232         PR driver/91220
1233         * opts.c (print_filtered_help): Exclude language-specific
1234         options from --help=common unless enabled in all FEs.
1235
1236 2020-01-27  Martin Liska  <mliska@suse.cz>
1237
1238         * opts.c (print_help): Exclude params from
1239         all except --help=param.
1240
1241 2020-01-27  Martin Liska  <mliska@suse.cz>
1242
1243         PR target/93274
1244         * config/i386/i386-features.c (make_resolver_func):
1245         Align the code with ppc64 target implementation.
1246         Do not generate a unique name for resolver function.
1247
1248 2020-01-27  Richard Biener  <rguenther@suse.de>
1249
1250         PR tree-optimization/93397
1251         * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1252         converted reduction chain SLP graph adjustment.
1253
1254 2020-01-26  Marek Polacek  <polacek@redhat.com>
1255
1256         PR sanitizer/93436
1257         * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1258         null DECL_NAME.
1259
1260 2020-01-26  Jason Merrill  <jason@redhat.com>
1261
1262         PR c++/92601
1263         * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1264         of complete types.
1265
1266 2020-01-26  Darius Galis  <darius.galis@cyberthorstudios.com>
1267
1268         * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1269         (rx_setmem): Likewise.
1270
1271 2020-01-26  Jakub Jelinek  <jakub@redhat.com>
1272
1273         PR target/93412
1274         * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1275         Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1276         drop <di> from constraint of last operand.
1277
1278         PR target/93430
1279         * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1280         TARGET_AVX2 and V4DFmode not in the split condition, but in the
1281         pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1282
1283 2020-01-25  Feng Xue  <fxue@os.amperecomputing.com>
1284
1285         PR ipa/93166
1286         * ipa-cp.c (get_info_about_necessary_edges): Remove value
1287         check assertion.
1288
1289 2020-01-24  Jeff Law  <law@redhat.com>
1290
1291         PR tree-optimization/92788
1292         * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1293         not EDGE_ABNORMAL.
1294
1295 2020-01-24  Jakub Jelinek  <jakub@redhat.com>
1296
1297         PR target/93395
1298         * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1299         *avx_vperm_broadcast_<mode>,
1300         <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1301         *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1302         Move before avx2_perm<mode>/avx512f_perm<mode>.
1303
1304         PR target/93376
1305         * simplify-rtx.c (simplify_const_unary_operation,
1306         simplify_const_binary_operation): Punt for mode precision above
1307         MAX_BITSIZE_MODE_ANY_INT.
1308
1309 2020-01-24  Andrew Pinski  <apinski@marvell.com>
1310
1311         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1312         alu.shift_reg to 0.
1313
1314 2020-01-24  Jeff Law  <law@redhat.com>
1315
1316         PR target/13721
1317         * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1318         for REGs.  Call output_operand_lossage to get more reasonable
1319         diagnostics.
1320
1321 2020-01-24  Andrew Stubbs  <ams@codesourcery.com>
1322
1323         * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1324         gcn_fp_compare_operator.
1325         (vec_cmpu<mode>di): Use gcn_compare_operator.
1326         (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1327         (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1328         (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1329         (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1330         (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1331         (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1332         (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1333         gcn_fp_compare_operator.
1334         (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1335         gcn_fp_compare_operator.
1336         (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1337         gcn_fp_compare_operator.
1338         (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1339         gcn_fp_compare_operator.
1340
1341 2020-01-24  Maciej W. Rozycki  <macro@wdc.com>
1342
1343         * doc/install.texi (Cross-Compiler-Specific Options): Document
1344         `--with-toolexeclibdir' option.
1345
1346 2020-01-24  Hans-Peter Nilsson  <hp@axis.com>
1347
1348         * target.def (flags_regnum): Also mention effect on delay slot filling.
1349         * doc/tm.texi: Regenerate.
1350
1351 2020-01-23  Jeff Law  <law@redhat.com>
1352
1353         PR translation/90162
1354         * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1355
1356 2020-01-23  Mikael Tillenius  <mti-1@tillenius.com>
1357
1358         PR target/92269
1359         * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1360         profiling label
1361
1362 2020-01-23  Jakub Jelinek  <jakub@redhat.com>
1363
1364         PR rtl-optimization/93402
1365         * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1366         USE insns.
1367
1368 2020-01-23  Dragan Mladjenovic  <dmladjenovic@wavecomp.com>
1369
1370         * config.in: Regenerated.
1371         * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1372         for TARGET_LIBC_GNUSTACK.
1373         * configure: Regenerated.
1374         * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1375         found to be 2.31 or greater.
1376
1377 2020-01-23  Dragan Mladjenovic  <dmladjenovic@wavecomp.com>
1378
1379         * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1380         TARGET_SOFT_FLOAT.
1381         * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1382         (mips_asm_file_end): New function. Delegate to
1383         file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1384         * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1385
1386 2020-01-23  Jakub Jelinek  <jakub@redhat.com>
1387
1388         PR target/93376
1389         * config/i386/i386-modes.def (POImode): New mode.
1390         (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1391         * config/i386/i386.md (DPWI): New mode attribute.
1392         (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1393         (QWI): Rename to...
1394         (QPWI): ... this.  Use POI instead of OI for TImode.
1395         (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1396         *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1397         instead of <QWI>.
1398
1399 2020-01-23  Richard Sandiford  <richard.sandiford@arm.com>
1400
1401         PR target/93341
1402         * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1403         unspec.
1404         (speculation_tracker_rev): New pattern.
1405         * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1406         Use speculation_tracker_rev to track the inverse condition.
1407
1408 2020-01-23  Richard Biener  <rguenther@suse.de>
1409
1410         PR tree-optimization/93381
1411         * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1412         alias-set of the def as argument and record the first one.
1413         (vn_walk_cb_data::first_set): New member.
1414         (vn_reference_lookup_3): Pass the alias-set of the current def
1415         to push_partial_def.  Fix alias-set used in the aggregate copy
1416         case.
1417         (vn_reference_lookup): Consistently set *last_vuse_ptr.
1418         * real.c (clear_significand_below): Fix out-of-bound access.
1419
1420 2020-01-23  Jakub Jelinek  <jakub@redhat.com>
1421
1422         PR target/93346
1423         * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1424         New define_insn patterns.
1425
1426 2020-01-23  Richard Sandiford  <richard.sandiford@arm.com>
1427
1428         * doc/sourcebuild.texi (check-function-bodies): Add an
1429         optional target/xfail selector.
1430
1431 2020-01-23  Richard Sandiford  <richard.sandiford@arm.com>
1432
1433         PR rtl-optimization/93124
1434         * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1435         bare USE and CLOBBER insns.
1436
1437 2020-01-22  Andrew Pinski  <apinski@marvell.com>
1438
1439         * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1440
1441 2020-01-22  David Malcolm  <dmalcolm@redhat.com>
1442
1443         PR analyzer/93307
1444         * gdbinit.in (break-on-saved-diagnostic): Update for move of
1445         diagnostic_manager into "ana" namespace.
1446         * selftest-run-tests.c (selftest::run_tests): Update for move of
1447         selftest::run_analyzer_selftests to
1448         ana::selftest::run_analyzer_selftests.
1449
1450 2020-01-22  Richard Sandiford  <richard.sandiford@arm.com>
1451
1452         * cfgexpand.c (union_stack_vars): Update the size.
1453
1454 2020-01-22  Richard Biener  <rguenther@suse.de>
1455
1456         PR tree-optimization/93381
1457         * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1458         throughout, handle all conversions the same.
1459
1460 2020-01-22  Jakub Jelinek  <jakub@redhat.com>
1461
1462         PR target/93335
1463         * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1464         gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1465         predicate, not whenever it is CONST_INT.  Otherwise, force_reg it.
1466         Call force_reg on high_in2 unconditionally.
1467
1468 2020-01-22  Martin Liska  <mliska@suse.cz>
1469
1470         PR tree-optimization/92924
1471         * profile.c (compute_value_histograms): Divide
1472         all counter values.
1473
1474 2020-01-22  Jakub Jelinek  <jakub@redhat.com>
1475
1476         PR target/91298
1477         * output.h (assemble_name_resolve): Declare.
1478         * varasm.c (assemble_name_resolve): New function.
1479         (assemble_name): Use it.
1480         * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
1481
1482 2020-01-22  Joseph Myers  <joseph@codesourcery.com>
1483
1484         * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
1485         update_web_docs_git instead of update_web_docs_svn.
1486
1487 2020-01-21  Andrew Pinski  <apinski@marvell.com>
1488
1489         PR target/9311
1490         * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
1491         as PTR mode. Have operand 1 as being modeless, it can be P mode.
1492         (*tlsgd_small_<mode>): Likewise.
1493         * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
1494         <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
1495         register.  Convert that register back to dest using convert_mode.
1496
1497 2020-01-21  Jim Wilson  <jimw@sifive.com>
1498
1499         * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
1500         instead of XINT.
1501
1502 2020-01-21  H.J. Lu  <hongjiu.lu@intel.com>
1503             Uros Bizjak    <ubizjak@gmail.com>
1504
1505         PR target/93319
1506         * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
1507         with ptr_mode.
1508         (legitimize_tls_address): Do GNU2 TLS address computation in
1509         ptr_mode and zero-extend result to Pmode.
1510         *  config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
1511         :P with :PTR and Pmode with ptr_mode.
1512         (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
1513         (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
1514         (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
1515
1516 2020-01-21  Jakub Jelinek  <jakub@redhat.com>
1517
1518         PR target/93333
1519         * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
1520         the last two operands are CONST_INT_P before using them as such.
1521
1522 2020-01-21  Richard Sandiford  <richard.sandiford@arm.com>
1523
1524         * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
1525         to get the integer element types.
1526
1527 2020-01-21  Richard Sandiford  <richard.sandiford@arm.com>
1528
1529         * config/aarch64/aarch64-sve-builtins.h
1530         (function_expander::convert_to_pmode): Declare.
1531         * config/aarch64/aarch64-sve-builtins.cc
1532         (function_expander::convert_to_pmode): New function.
1533         (function_expander::get_contiguous_base): Use it.
1534         (function_expander::prepare_gather_address_operands): Likewise.
1535         * config/aarch64/aarch64-sve-builtins-sve2.cc
1536         (svwhilerw_svwhilewr_impl::expand): Likewise.
1537
1538 2020-01-21  Szabolcs Nagy  <szabolcs.nagy@arm.com>
1539
1540         PR target/92424
1541         * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
1542         cfun->machine->label_is_assembled.
1543         (aarch64_print_patchable_function_entry): New.
1544         (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
1545         * config/aarch64/aarch64.h (struct machine_function): New field,
1546         label_is_assembled.
1547
1548 2020-01-21  David Malcolm  <dmalcolm@redhat.com>
1549
1550         PR ipa/93315
1551         * ipa-profile.c (ipa_profile): Delete call_sums and set it to
1552         NULL on exit.
1553
1554 2020-01-18  Jan Hubicka  <hubicka@ucw.cz>
1555
1556         PR lto/93318
1557         * cgraph.c (cgraph_edge::resolve_speculation,
1558         cgraph_edge::redirect_call_stmt_to_callee): Fix update of
1559         call_stmt_site_hash.
1560
1561 2020-01-21  Martin Liska  <mliska@suse.cz>
1562
1563         * config/rs6000/rs6000.c (common_mode_defined): Remove
1564         unused variable.
1565
1566 2020-01-21  Richard Biener  <rguenther@suse.de>
1567
1568         PR tree-optimization/92328
1569         * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
1570         type when value-numbering same-sized store by inserting a
1571         VIEW_CONVERT_EXPR.
1572         (eliminate_dom_walker::eliminate_stmt): When eliminating
1573         a redundant store handle bit-reinterpretation of the same value.
1574
1575 2020-01-21  Andrew Pinski  <apinski@marvel.com>
1576
1577         PR tree-opt/93321
1578         * tree-into-ssa.c (prepare_block_for_update_1): Split out
1579         from ...
1580         (prepare_block_for_update): This.  Use a worklist instead of
1581         recursing.
1582
1583 2020-01-21  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
1584
1585         * gcc/config/arm/arm.c (clear_operation_p):
1586         Initialise last_regno, skip first iteration
1587         based on the first_set value and use ints instead
1588         of the unnecessary HOST_WIDE_INTs.
1589
1590 2020-01-21  Jakub Jelinek  <jakub@redhat.com>
1591
1592         PR target/93073
1593         * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
1594         compare_mode other than SFmode or DFmode.
1595
1596 2020-01-21  Kito Cheng  <kito.cheng@sifive.com>
1597
1598         PR target/93304
1599         * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
1600         * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
1601         * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
1602
1603 2020-01-20  Wilco Dijkstra  <wdijkstr@arm.com>
1604
1605         * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
1606
1607 2020-01-20  Andrew Pinski  <apinski@marvell.com>
1608
1609         PR middle-end/93242
1610         * targhooks.c (default_print_patchable_function_entry): Use
1611         output_asm_insn to emit the nop instruction.
1612
1613 2020-01-20  Fangrui Song  <maskray@google.com>
1614
1615         PR middle-end/93194
1616         * targhooks.c (default_print_patchable_function_entry): Align to
1617         POINTER_SIZE.
1618
1619 2020-01-20  H.J. Lu  <hongjiu.lu@intel.com>
1620
1621         PR target/93319
1622         * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
1623         gen_tls_dynamic_gnu2_64.  Compute GNU2 TLS address in ptr_mode.
1624         * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
1625         (@tls_dynamic_gnu2_64_<mode>): This.  Replace DI with P.
1626         (*tls_dynamic_gnu2_lea_64): Renamed to ...
1627         (*tls_dynamic_gnu2_lea_64_<mode>): This.  Replace DI with P.
1628         Remove the {q} suffix from lea.
1629         (*tls_dynamic_gnu2_call_64): Renamed to ...
1630         (*tls_dynamic_gnu2_call_64_<mode>): This.  Replace DI with P.
1631         (*tls_dynamic_gnu2_combine_64): Renamed to ...
1632         (*tls_dynamic_gnu2_combine_64_<mode>): This.  Replace DI with P.
1633         Pass Pmode to gen_tls_dynamic_gnu2_64.
1634
1635 2020-01-20  Wilco Dijkstra  <wdijkstr@arm.com>
1636
1637         * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
1638
1639 2020-01-20  Richard Sandiford  <richard.sandiford@arm.com>
1640
1641         * config/aarch64/aarch64-sve-builtins-base.cc
1642         (svld1ro_impl::memory_vector_mode): Remove parameter name.
1643
1644 2020-01-20  Richard Biener  <rguenther@suse.de>
1645
1646         PR debug/92763
1647         * dwarf2out.c (prune_unused_types): Unconditionally mark
1648         called function DIEs.
1649
1650 2020-01-20  Martin Liska  <mliska@suse.cz>
1651
1652         PR tree-optimization/93199
1653         * tree-eh.c (struct leh_state): Add
1654         new field outer_non_cleanup.
1655         (cleanup_is_dead_in): Pass leh_state instead
1656         of eh_region.  Add a checking that state->outer_non_cleanup
1657         points to outer non-clean up region.
1658         (lower_try_finally): Record outer_non_cleanup
1659         for this_state.
1660         (lower_catch): Likewise.
1661         (lower_eh_filter): Likewise.
1662         (lower_eh_must_not_throw): Likewise.
1663         (lower_cleanup): Likewise.
1664
1665 2020-01-20  Richard Biener  <rguenther@suse.de>
1666
1667         PR tree-optimization/93094
1668         * tree-vectorizer.h (vect_loop_versioning): Adjust.
1669         (vect_transform_loop): Likewise.
1670         * tree-vectorizer.c (try_vectorize_loop_1): Pass down
1671         loop_vectorized_call to vect_transform_loop.
1672         * tree-vect-loop.c (vect_transform_loop): Pass down
1673         loop_vectorized_call to vect_loop_versioning.
1674         * tree-vect-loop-manip.c (vect_loop_versioning): Use
1675         the earlier discovered loop_vectorized_call.
1676
1677 2020-01-19  Eric S. Raymond <esr@thyrsus.com>
1678
1679         * doc/contribute.texi: Update for SVN -> Git transition.
1680         * doc/install.texi: Likewise.
1681
1682 2020-01-18  Jan Hubicka  <hubicka@ucw.cz>
1683
1684         PR lto/93318
1685         * cgraph.c (cgraph_edge::make_speculative): Increase number of
1686         speculative targets.
1687         (verify_speculative_call): New function
1688         (cgraph_node::verify_node): Use it.
1689         * ipa-profile.c (ipa_profile): Fix formating; do not set number of
1690         speculations.
1691
1692 2020-01-18  Jan Hubicka  <hubicka@ucw.cz>
1693
1694         PR lto/93318
1695         * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
1696         (cgraph_edge::make_direct): Remove all indirect targets.
1697         (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
1698         (cgraph_node::verify_node): Verify that only one call_stmt or
1699         lto_stmt_uid is set.
1700         * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
1701         lto_stmt_uid.
1702         * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
1703         (lto_output_ref): Simplify streaming of stmt.
1704         * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
1705
1706 2020-01-18  Tamar Christina  <tamar.christina@arm.com>
1707
1708         * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
1709         Mark parameter unused.
1710
1711 2020-01-18  Hans-Peter Nilsson  <hp@axis.com>
1712
1713         * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
1714
1715 2019-01-18  Gerald Pfeifer  <gerald@pfeifer.com>
1716
1717         * varpool.c (ctor_useable_for_folding_p): Fix grammar.
1718
1719 2020-01-18  Iain Sandoe  <iain@sandoe.co.uk>
1720
1721         * Makefile.in: Add coroutine-passes.o.
1722         * builtin-types.def (BT_CONST_SIZE): New.
1723         (BT_FN_BOOL_PTR): New.
1724         (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
1725         * builtins.def (DEF_COROUTINE_BUILTIN): New.
1726         * coroutine-builtins.def: New file.
1727         * coroutine-passes.cc: New file.
1728         * function.h (struct GTY function): Add a bit to indicate that the
1729         function is a coroutine component.
1730         * internal-fn.c (expand_CO_FRAME): New.
1731         (expand_CO_YIELD): New.
1732         (expand_CO_SUSPN): New.
1733         (expand_CO_ACTOR): New.
1734         * internal-fn.def (CO_ACTOR): New.
1735         (CO_YIELD): New.
1736         (CO_SUSPN): New.
1737         (CO_FRAME): New.
1738         * passes.def: Add pass_coroutine_lower_builtins,
1739         pass_coroutine_early_expand_ifns.
1740         * tree-pass.h (make_pass_coroutine_lower_builtins): New.
1741         (make_pass_coroutine_early_expand_ifns): New.
1742         * doc/invoke.texi: Document the fcoroutines command line
1743         switch.
1744
1745 2020-01-18  Jakub Jelinek  <jakub@redhat.com>
1746
1747         * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
1748
1749         PR target/93312
1750         * config/arm/arm.c (clear_operation_p): Don't use REGNO until
1751         after checking the argument is a REG.  Don't use REGNO (reg)
1752         again to set last_regno, reuse regno variable instead.
1753
1754 2020-01-17  David Malcolm  <dmalcolm@redhat.com>
1755
1756         * doc/analyzer.texi (Limitations): Add note about NaN.
1757
1758 2020-01-17  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
1759             Sudakshina Das  <sudi.das@arm.com>
1760
1761         * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
1762         and valid immediate.
1763         (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
1764         (lshrdi3): Generate thumb2_lsrl for valid immediates.
1765         * config/arm/constraints.md (Pg): New.
1766         * config/arm/predicates.md (long_shift_imm): New.
1767         (arm_reg_or_long_shift_imm): Likewise.
1768         * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
1769         (thumb2_lsll): Likewise.
1770         (thumb2_lsrl): New.
1771
1772 2020-01-17  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
1773             Sudakshina Das  <sudi.das@arm.com>
1774
1775         * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
1776         (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
1777         * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
1778         register pairs for doubleword quantities for ARMv8.1M-Mainline.
1779         * config/arm/thumb2.md (thumb2_asrl): New.
1780         (thumb2_lsll): Likewise.
1781
1782 2020-01-17  Jakub Jelinek  <jakub@redhat.com>
1783
1784         * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
1785         unused variable.
1786
1787 2020-01-17  Alexander Monakov  <amonakov@ispras.ru>
1788
1789         * gdbinit.in (help-gcc-hooks): New command.
1790         (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
1791         pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
1792         documentation.
1793
1794 2020-01-17  Matthew Malcomson  <matthew.malcomson@arm.com>
1795
1796         * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
1797         correct target macro.
1798
1799 2020-01-17  Matthew Malcomson  <matthew.malcomson@arm.com>
1800
1801         * config/aarch64/aarch64-protos.h
1802         (aarch64_sve_ld1ro_operand_p): New.
1803         * config/aarch64/aarch64-sve-builtins-base.cc
1804         (class load_replicate): New.
1805         (class svld1ro_impl): New.
1806         (class svld1rq_impl): Change to inherit from load_replicate.
1807         (svld1ro): New sve intrinsic function base.
1808         * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
1809         New DEF_SVE_FUNCTION.
1810         * config/aarch64/aarch64-sve-builtins-base.h
1811         (svld1ro): New decl.
1812         * config/aarch64/aarch64-sve-builtins.cc
1813         (function_expander::add_mem_operand): Modify assert to allow
1814         OImode.
1815         * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
1816         pattern.
1817         * config/aarch64/aarch64.c
1818         (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
1819         (aarch64_sve_ld1rq_ld1ro_operand_p): This.
1820         (aarch64_sve_ld1ro_operand_p): New.
1821         * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
1822         * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
1823         * config/aarch64/predicates.md
1824         (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
1825
1826 2020-01-17  Matthew Malcomson  <matthew.malcomson@arm.com>
1827
1828         * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
1829         Introduce this ACLE specified predefined macro.
1830         * config/aarch64/aarch64-option-extensions.def (f64mm): New.
1831         (fp): Disabling this disables f64mm.
1832         (simd): Disabling this disables f64mm.
1833         (fp16): Disabling this disables f64mm.
1834         (sve): Disabling this disables f64mm.
1835         * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
1836         (AARCH64_ISA_F64MM): New.
1837         (TARGET_F64MM): New.
1838         * doc/invoke.texi (f64mm): Document new option.
1839
1840 2020-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
1841
1842         * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
1843         (neoversen1_tunings): Likewise.
1844
1845 2020-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
1846
1847         PR target/92692
1848         * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
1849         Add assert to ensure prolog has been emitted.
1850         (aarch64_split_atomic_op): Likewise.
1851         * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
1852         Use epilogue_completed rather than reload_completed.
1853         (aarch64_atomic_exchange<mode>): Likewise.
1854         (aarch64_atomic_<atomic_optab><mode>): Likewise.
1855         (atomic_nand<mode>): Likewise.
1856         (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
1857         (atomic_fetch_nand<mode>): Likewise.
1858         (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
1859         (atomic_nand_fetch<mode>): Likewise.
1860
1861 2020-01-17  Richard Sandiford  <richard.sandiford@arm.com>
1862
1863         PR target/93133
1864         * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
1865         for FP modes.
1866         (REVERSE_CONDITION): Delete.
1867         * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
1868         (CCFP_CCFPE): Likewise.
1869         (e): New mode attribute.
1870         * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
1871         (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
1872         (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
1873         (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
1874         (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
1875         (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
1876         * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
1877         name of generator from gen_ccmpdi to gen_ccmpccdi.
1878         (aarch64_gen_ccmp_next): Use code_for_ccmp.  If we want to reverse
1879         the previous comparison but aren't able to, use the new ccmp_rev
1880         patterns instead.
1881
1882 2020-01-17  Richard Sandiford  <richard.sandiford@arm.com>
1883
1884         * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
1885         than testing directly for INTEGER_CST.
1886         (gimplify_target_expr, gimplify_omp_depend): Likewise.
1887
1888 2020-01-17  Jakub Jelinek  <jakub@redhat.com>
1889
1890         PR tree-optimization/93292
1891         * tree-vect-stmts.c (vectorizable_comparison): Punt also if
1892         get_vectype_for_scalar_type returns NULL.
1893
1894 2020-01-16  Jan Hubicka  <hubicka@ucw.cz>
1895
1896         * params.opt (-param=max-predicted-iterations): Increase range from 0.
1897         * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
1898
1899 2020-01-16  Jan Hubicka  <hubicka@ucw.cz>
1900
1901         * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
1902         dump.
1903         * params.opt: (max-predicted-iterations): Set bounds.
1904         * predict.c (real_almost_one, real_br_prob_base,
1905         real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
1906         (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
1907         probabilities; do not truncate to reg_br_prob_bases.
1908         (estimate_loops_at_level): Pass max_cyclic_prob.
1909         (estimate_loops): Compute max_cyclic_prob.
1910         (estimate_bb_frequencies): Do not initialize real_*; update calculation
1911         of back edge prob.
1912         * profile-count.c (profile_probability::to_sreal): New.
1913         * profile-count.h (class sreal): Move up in file.
1914         (profile_probability::to_sreal): Declare.
1915
1916 2020-01-16  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
1917
1918         * config/arm/arm.c
1919         (arm_invalid_conversion): New function for target hook.
1920         (arm_invalid_unary_op): New function for target hook.
1921         (arm_invalid_binary_op): New function for target hook.
1922
1923 2020-01-16  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
1924
1925         * config.gcc: Add arm_bf16.h.
1926         * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
1927         (arm_simd_builtin_std_type): Add BFmode.
1928         (arm_init_simd_builtin_types): Define element types for vector types.
1929         (arm_init_bf16_types): New function.
1930         (arm_init_builtins): Add arm_init_bf16_types function call.
1931         * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
1932         * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
1933         * config/arm/arm.c (aapcs_vfp_sub_candidate):  Add BFmode.
1934         (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
1935         (arm_vector_mode_supported_p): Add V4BF, V8BF.
1936         (arm_mangle_type):  Add __bf16.
1937         * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
1938         VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
1939         arm_bf16_ptr_type_node.
1940         * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
1941         define_split between ARM registers.
1942         * config/arm/arm_bf16.h: New file.
1943         * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
1944         * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
1945         (VQXMOV): Add V8BF.
1946         * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
1947         * config/arm/vfp.md: Add BFmode to movhf patterns.
1948
1949 2020-01-16  Mihail Ionescu  <mihail.ionescu@arm.com>
1950             Andre Vieira  <andre.simoesdiasvieira@arm.com>
1951
1952         * config/arm/arm-cpus.in (mve, mve_float): New features.
1953         (dsp, mve, mve.fp): New options.
1954         * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
1955         * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
1956         * doc/invoke.texi: Document the armv8.1-m mve and dps options.
1957
1958 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
1959             Thomas Preud'homme  <thomas.preudhomme@arm.com>
1960
1961         * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
1962         Armv8-M Mainline.
1963         * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
1964         error for using -mcmse when targeting Armv8.1-M Mainline.
1965
1966 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
1967             Thomas Preud'homme  <thomas.preudhomme@arm.com>
1968
1969         * config/arm/arm.md (nonsecure_call_internal): Do not force memory
1970         address in r4 when targeting Armv8.1-M Mainline.
1971         (nonsecure_call_value_internal): Likewise.
1972         * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
1973         a register match_operand again.  Emit BLXNS when targeting
1974         Armv8.1-M Mainline.
1975         (nonsecure_call_value_reg_thumb2): Likewise.
1976
1977 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
1978             Thomas Preud'homme  <thomas.preudhomme@arm.com>
1979
1980         * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
1981         (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
1982         variable as true when floating-point ABI is not hard.  Replace
1983         check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
1984         Generate VLSTM and VLLDM instruction respectively before and
1985         after a function call to cmse_nonsecure_call function.
1986         * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
1987         (VUNSPEC_VLLDM): Likewise.
1988         * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
1989         (lazy_load_multiple_insn): Likewise.
1990
1991 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
1992             Thomas Preud'homme  <thomas.preudhomme@arm.com>
1993
1994         * config/arm/arm.c (vfp_emit_fstmd): Declare early.
1995         (arm_emit_vfp_multi_reg_pop): Likewise.
1996         (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
1997         registers to clear in max_fp_regno.  Emit VPUSH and VPOP to save and
1998         restore callee-saved VFP registers.
1999
2000 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
2001             Thomas Preud'homme  <thomas.preudhomme@arm.com>
2002
2003         * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2004         (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2005         (cmse_nonsecure_call_inline_register_clear): This.  Save and clear
2006         callee-saved GPRs as well as clear ip register before doing a nonsecure
2007         call then restore callee-saved GPRs after it when targeting
2008         Armv8.1-M Mainline.
2009         (arm_reorg): Adapt to function rename.
2010
2011 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
2012             Thomas Preud'homme  <thomas.preudhomme@arm.com>
2013
2014         * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2015         * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2016         clear_vfp_multiple pattern based on a new vfp parameter.
2017         (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2018         targeting Armv8.1-M Mainline.
2019         (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2020         unconditionally when targeting Armv8.1-M Mainline architecture.  Check
2021         whether VFP registers are available before looking call_used_regs for a
2022         VFP register.
2023         * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2024         of prototype of clear_operation_p.
2025         (clear_vfp_multiple_operation): New predicate.
2026         * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2027         * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2028
2029 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
2030             Thomas Preud'homme  <thomas.preudhomme@arm.com>
2031
2032         * config/arm/arm-protos.h (clear_operation_p): Declare.
2033         * config/arm/arm.c (clear_operation_p): New function.
2034         (cmse_clear_registers): Generate clear_multiple instruction pattern if
2035         targeting Armv8.1-M Mainline or successor.
2036         (output_return_instruction): Only output APSR register clearing if
2037         Armv8.1-M Mainline instructions not available.
2038         (thumb_exit): Likewise.
2039         * config/arm/predicates.md (clear_multiple_operation): New predicate.
2040         * config/arm/thumb2.md (clear_apsr): New define_insn.
2041         (clear_multiple): Likewise.
2042         * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2043
2044 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
2045             Thomas Preud'homme  <thomas.preudhomme@arm.com>
2046
2047         * config/arm/arm.c (fp_sysreg_names): Declare and define.
2048         (use_return_insn): Also return false for Armv8.1-M Mainline.
2049         (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2050         Mainline instructions are available.
2051         (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2052         when targeting Armv8.1-M Mainline Security Extensions.
2053         (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2054         Mainline entry function.
2055         (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2056         targeting Armv8.1-M Mainline or successor.
2057         (arm_expand_epilogue): Fix indentation of caller-saved register
2058         clearing.  Restore FPCXTNS if this is an Armv8.1-M Mainline
2059         entry function.
2060         * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2061         (FP_SYSREGS): Likewise.
2062         (enum vfp_sysregs_encoding): Define enum.
2063         (fp_sysreg_names): Declare.
2064         * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2065         * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2066         (pop_fpsysreg_insn): Likewise.
2067
2068 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
2069             Thomas Preud'homme  <thomas.preudhomme@arm.com>
2070
2071         * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2072         (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2073         ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2074         ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2075         ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2076         (ARMv8_1m_main): New feature group.
2077         (armv8.1-m.main): New architecture.
2078         * config/arm/arm-tables.opt: Regenerate.
2079         * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2080         (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2081         (arm_options_perform_arch_sanity_checks): Error out when targeting
2082         Armv8.1-M Mainline Security Extensions.
2083         * config/arm/arm.h (arm_arch8_1m_main): Declare.
2084
2085 2020-01-16  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
2086
2087         * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2088         aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2089         * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2090         aarch64_bfdot_laneq): New.
2091         * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2092         vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2093         vbfdotq_laneq_f32): New.
2094         * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2095         VBFMLA_W, VBF): New.
2096         (isquadop): Add V4BF, V8BF.
2097
2098 2020-01-16  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
2099
2100         * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2101         New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2102         TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2103         (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2104         (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2105         * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2106         usdot_laneq, sudot_lane,sudot_laneq): New.
2107         * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2108         (aarch64_<sur>dot_lane): New.
2109         * config/aarch64/arm_neon.h (vusdot_s32): New.
2110         (vusdotq_s32): New.
2111         (vusdot_lane_s32): New.
2112         (vsudot_lane_s32): New.
2113         * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2114         (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2115
2116 2020-01-16  Martin Liska  <mliska@suse.cz>
2117
2118         * value-prof.c (dump_histogram_value): Fix
2119         obvious spacing issue.
2120
2121 2020-01-16  Andrew Pinski  <apinski@marvell.com>
2122
2123         * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2124         !storage_order_barrier_p.
2125
2126 2020-01-16  Andrew Pinski  <apinski@marvell.com>
2127
2128         * sched-int.h (_dep): Add unused bit-field field for the padding.
2129         * sched-deps.c (init_dep_1): Init unused field.
2130
2131 2020-01-16  Andrew Pinski  <apinski@marvell.com>
2132
2133         * optabs.h (create_expand_operand): Initialize target field also.
2134
2135 2020-01-16  Andre Vieira  <andre.simoesdiasvieira@arm.com>
2136
2137         PR tree-optimization/92429
2138         * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2139         * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2140         control folding.
2141         * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2142         tree.
2143
2144 2020-01-16  Richard Sandiford  <richard.sandiford@arm.com>
2145
2146         * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2147         aarch64_sve_int_mode to each mode.
2148
2149 2020-01-15  David Malcolm  <dmalcolm@redhat.com>
2150
2151         * doc/analyzer.texi (Overview): Add note about
2152         -fdump-ipa-analyzer.
2153
2154 2020-01-15  Wilco Dijkstra  <wdijkstr@arm.com>
2155
2156         PR tree-optimization/93231
2157         * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2158         input_type is unsigned.  Use tree_to_shwi for shift constant.
2159         Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2160         (simplify_count_trailing_zeroes): Add test to handle known non-zero
2161         inputs more efficiently.
2162
2163 2020-01-15  Uroš Bizjak  <ubizjak@gmail.com>
2164
2165         * config/i386/i386.md (*movsf_internal): Do not require
2166         SSE2 ISA for alternatives 14 and 15.
2167
2168 2020-01-15  Richard Biener  <rguenther@suse.de>
2169
2170         PR middle-end/93273
2171         * tree-eh.c (sink_clobbers): If we already visited the destination
2172         block do not defer insertion.
2173         (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2174         the purpose of defered insertion.
2175
2176 2020-01-15  Jakub Jelinek  <jakub@redhat.com>
2177
2178         * BASE-VER: Bump to 10.0.1.
2179
2180 2020-01-15  Richard Sandiford  <richard.sandiford@arm.com>
2181
2182         PR tree-optimization/93247
2183         * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2184         type of the stmt that we're going to vectorize.
2185
2186 2020-01-15  Richard Sandiford  <richard.sandiford@arm.com>
2187
2188         * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2189         VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2190         type from the lhs.
2191
2192 2020-01-15  Martin Liska  <mliska@suse.cz>
2193
2194         * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2195         2 calls of streamer_read_hwi in a function call.
2196
2197 2020-01-15  Richard Biener  <rguenther@suse.de>
2198
2199         * alias.c (record_alias_subset): Avoid redundant work when
2200         subset is already recorded.
2201
2202 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
2203
2204         * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2205         the analyzer options provide CWE identifiers.
2206
2207 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
2208
2209         * tree-diagnostic-path.cc (path_summary::event_range::print):
2210         When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2211         using get_pure_location.
2212
2213 2020-01-15  Jakub Jelinek  <jakub@redhat.com>
2214
2215         PR tree-optimization/93262
2216         * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2217         perform head trimming only if the last argument is constant,
2218         either all ones, or larger or equal to head trim, in the latter
2219         case decrease the last argument by head_trim.
2220
2221         PR tree-optimization/93249
2222         * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2223         (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2224         function body scope, reindent.  For BUILTIN_IN_STRNCPY*, don't
2225         perform head trim unless we can prove there are no '\0' chars
2226         from the source among the first head_trim chars.
2227
2228 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
2229
2230         * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2231
2232 2020-01-15  Jakub Jelinek  <jakub@redhat.com>
2233
2234         PR target/93009
2235         * config/i386/sse.md
2236         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2237         *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2238         *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2239         *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2240         just a single alternative instead of two, make operands 1 and 2
2241         commutative.
2242
2243 2020-01-14  Jan Hubicka  <hubicka@ucw.cz>
2244
2245         PR lto/91576
2246         * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2247         TYPE_MODE.
2248
2249 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
2250
2251         * Makefile.in (lang_opt_files): Add analyzer.opt.
2252         (ANALYZER_OBJS): New.
2253         (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2254         tristate.o and ANALYZER_OBJS.
2255         (TEXI_GCCINT_FILES): Add analyzer.texi.
2256         * common.opt (-fanalyzer): New driver option.
2257         * config.in: Regenerate.
2258         * configure: Regenerate.
2259         * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2260         (gccdepdir): Also create depdir for "analyzer" subdir.
2261         * digraph.cc: New file.
2262         * digraph.h: New file.
2263         * doc/analyzer.texi: New file.
2264         * doc/gccint.texi ("Static Analyzer") New menu item.
2265         (analyzer.texi): Include it.
2266         * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2267         ("Warning Options"): Add static analysis warnings to the list.
2268         (-Wno-analyzer-double-fclose): New option.
2269         (-Wno-analyzer-double-free): New option.
2270         (-Wno-analyzer-exposure-through-output-file): New option.
2271         (-Wno-analyzer-file-leak): New option.
2272         (-Wno-analyzer-free-of-non-heap): New option.
2273         (-Wno-analyzer-malloc-leak): New option.
2274         (-Wno-analyzer-possible-null-argument): New option.
2275         (-Wno-analyzer-possible-null-dereference): New option.
2276         (-Wno-analyzer-null-argument): New option.
2277         (-Wno-analyzer-null-dereference): New option.
2278         (-Wno-analyzer-stale-setjmp-buffer): New option.
2279         (-Wno-analyzer-tainted-array-index): New option.
2280         (-Wno-analyzer-use-after-free): New option.
2281         (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2282         (-Wno-analyzer-use-of-uninitialized-value): New option.
2283         (-Wanalyzer-too-complex): New option.
2284         (-fanalyzer-call-summaries): New warning.
2285         (-fanalyzer-checker=): New warning.
2286         (-fanalyzer-fine-grained): New warning.
2287         (-fno-analyzer-state-merge): New warning.
2288         (-fno-analyzer-state-purge): New warning.
2289         (-fanalyzer-transitivity): New warning.
2290         (-fanalyzer-verbose-edges): New warning.
2291         (-fanalyzer-verbose-state-changes): New warning.
2292         (-fanalyzer-verbosity=): New warning.
2293         (-fdump-analyzer): New warning.
2294         (-fdump-analyzer-callgraph): New warning.
2295         (-fdump-analyzer-exploded-graph): New warning.
2296         (-fdump-analyzer-exploded-nodes): New warning.
2297         (-fdump-analyzer-exploded-nodes-2): New warning.
2298         (-fdump-analyzer-exploded-nodes-3): New warning.
2299         (-fdump-analyzer-supergraph): New warning.
2300         * doc/sourcebuild.texi (dg-require-dot): New.
2301         (dg-check-dot): New.
2302         * gdbinit.in (break-on-saved-diagnostic): New command.
2303         * graphviz.cc: New file.
2304         * graphviz.h: New file.
2305         * ordered-hash-map-tests.cc: New file.
2306         * ordered-hash-map.h: New file.
2307         * passes.def (pass_analyzer): Add before
2308         pass_ipa_whole_program_visibility.
2309         * selftest-run-tests.c (selftest::run_tests): Call
2310         selftest::ordered_hash_map_tests_cc_tests.
2311         * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2312         decl.
2313         * shortest-paths.h: New file.
2314         * timevar.def (TV_ANALYZER): New timevar.
2315         (TV_ANALYZER_SUPERGRAPH): Likewise.
2316         (TV_ANALYZER_STATE_PURGE): Likewise.
2317         (TV_ANALYZER_PLAN): Likewise.
2318         (TV_ANALYZER_SCC): Likewise.
2319         (TV_ANALYZER_WORKLIST): Likewise.
2320         (TV_ANALYZER_DUMP): Likewise.
2321         (TV_ANALYZER_DIAGNOSTICS): Likewise.
2322         (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2323         * tree-pass.h (make_pass_analyzer): New decl.
2324         * tristate.cc: New file.
2325         * tristate.h: New file.
2326
2327 2020-01-14  Uroš Bizjak  <ubizjak@gmail.com>
2328
2329         PR target/93254
2330         * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2331         alternatives 9 and 10.
2332
2333 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
2334
2335         * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2336         * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2337         * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2338         * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2339         (selftest::hash_map_tests_c_tests): Call it.
2340         * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2341         New static constant, using the value of = H::empty_zero_p.
2342         (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2343         from default_hash_traits <Value>.
2344         * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2345         from Traits.
2346         * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2347         * hash-table.h (hash_table::alloc_entries): Guard the loop of
2348         calls to mark_empty with !Descriptor::empty_zero_p.
2349         (hash_table::empty_slow): Conditionalize the memset call with a
2350         check that Descriptor::empty_zero_p; otherwise, loop through the
2351         entries calling mark_empty on them.
2352         * hash-traits.h (int_hash::empty_zero_p): New static constant.
2353         (pointer_hash::empty_zero_p): Likewise.
2354         (pair_hash::empty_zero_p): Likewise.
2355         * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2356         Likewise.
2357         * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2358         (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2359         * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2360         * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2361         (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2362         * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2363         * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2364         * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2365         * tree-vectorizer.h
2366         (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2367         Likewise.
2368
2369 2020-01-14  Kewen Lin  <linkw@gcc.gnu.org>
2370
2371         * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2372         fix typo on return value.
2373
2374 2020-01-14  Xiong Hu Luo  <luoxhu@linux.ibm.com>
2375
2376         PR ipa/69678
2377         * cgraph.c (symbol_table::create_edge): Init speculative_id and
2378         target_prob.
2379         (cgraph_edge::make_speculative): Add param for setting speculative_id
2380         and target_prob.
2381         (cgraph_edge::speculative_call_info): Update comments and find reference
2382         by speculative_id for multiple indirect targets.
2383         (cgraph_edge::resolve_speculation): Decrease the speculations
2384         for indirect edge, drop it's speculative if not direct target
2385         left. Update comments.
2386         (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2387         (cgraph_node::dump): Print num_speculative_call_targets.
2388         (cgraph_node::verify_node): Don't report error if speculative
2389         edge not include statement.
2390         (cgraph_edge::num_speculative_call_targets_p): New function.
2391         * cgraph.h (int common_target_id): Remove.
2392         (int common_target_probability): Remove.
2393         (num_speculative_call_targets): New variable.
2394         (make_speculative): Add param for setting speculative_id.
2395         (cgraph_edge::num_speculative_call_targets_p): New declare.
2396         (target_prob): New variable.
2397         (speculative_id): New variable.
2398         * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2399           call summaries for multiple speculative call targets.
2400         * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2401         * ipa-profile.c (struct speculative_call_target): New struct.
2402         (class speculative_call_summary): New class.
2403         (class speculative_call_summaries): New class.
2404         (call_sums): New variable.
2405         (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2406         (ipa_profile_write_edge_summary): New function.
2407         (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2408         (ipa_profile_dump_all_summaries): New function.
2409         (ipa_profile_read_edge_summary): New function.
2410         (ipa_profile_read_summary_section): New function.
2411         (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2412         (ipa_profile): Generate num_speculative_call_targets from
2413         profile summaries.
2414         * ipa-ref.h (speculative_id): New variable.
2415         * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2416         * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2417         common_target_probability.   Stream out speculative_id and
2418         num_speculative_call_targets.
2419         (input_edge): Likewise.
2420         * predict.c (dump_prediction): Remove edges count assert to be
2421         precise.
2422         * symtab.c (symtab_node::create_reference): Init speculative_id.
2423         (symtab_node::clone_references): Clone speculative_id.
2424         (symtab_node::clone_referring): Clone speculative_id.
2425         (symtab_node::clone_reference): Clone speculative_id.
2426         (symtab_node::clear_stmts_in_references): Clear speculative_id.
2427         * tree-inline.c (copy_bb): Duplicate all the speculative edges
2428         if indirect call contains multiple speculative targets.
2429         * value-prof.h  (check_ic_target): Remove.
2430         * value-prof.c  (gimple_value_profile_transformations):
2431         Use void function gimple_ic_transform.
2432         * value-prof.c  (gimple_ic_transform): Handle topn case.
2433         Fix comment typos.  Change it to a void function.
2434
2435 2020-01-13  Andrew Pinski  <apinski@marvell.com>
2436
2437         * config/aarch64/aarch64-cores.def (octeontx2): New define.
2438         (octeontx2t98): New define.
2439         (octeontx2t96): New define.
2440         (octeontx2t93): New define.
2441         (octeontx2f95): New define.
2442         (octeontx2f95n): New define.
2443         (octeontx2f95mm): New define.
2444         * config/aarch64/aarch64-tune.md: Regenerate.
2445         * doc/invoke.texi (-mcpu=): Document the new cpu types.
2446
2447 2020-01-13  Jason Merrill  <jason@redhat.com>
2448
2449         PR c++/33799 - destroy return value if local cleanup throws.
2450         * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2451
2452 2020-01-13  Martin Liska  <mliska@suse.cz>
2453
2454         * ipa-cp.c (get_max_overall_size): Use newly
2455         renamed param param_ipa_cp_unit_growth.
2456         * params.opt: Remove legacy param name.
2457
2458 2020-01-13  Martin Sebor  <msebor@redhat.com>
2459
2460         PR tree-optimization/93213
2461         * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2462         stores to be eliminated.
2463
2464 2020-01-13  Martin Liska  <mliska@suse.cz>
2465
2466         * opts.c (print_help): Do not print CL_PARAM
2467         and CL_WARNING for CL_OPTIMIZATION.
2468
2469 2020-01-13  Jonathan Wakely  <jwakely@redhat.com>
2470
2471         PR driver/92757
2472         * doc/invoke.texi (Warning Options): Add caveat about some warnings
2473         depending on optimization settings.
2474
2475 2020-01-13  Jakub Jelinek  <jakub@redhat.com>
2476
2477         PR tree-optimization/90838
2478         * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2479         SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
2480         argument rather than to initialize temporary for targets that
2481         don't use the mode argument at all.  Initialize ctzval to avoid
2482         warning at -O0.
2483
2484 2020-01-10  Thomas Schwinge  <thomas@codesourcery.com>
2485
2486         * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
2487         * tree-core.h: Document it.
2488         * gimplify.c (gimplify_omp_workshare): Set it.
2489         * omp-low.c (lower_omp_target): Use it.
2490         * tree-pretty-print.c (dump_omp_clause): Print it.
2491
2492         * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
2493         Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
2494
2495 2020-01-10  David Malcolm  <dmalcolm@redhat.com>
2496
2497         * Makefile.in (OBJS): Add tree-diagnostic-path.o.
2498         * common.opt (fdiagnostics-path-format=): New option.
2499         (diagnostic_path_format): New enum.
2500         (fdiagnostics-show-path-depths): New option.
2501         * coretypes.h (diagnostic_event_id_t): New forward decl.
2502         * diagnostic-color.c (color_dict): Add "path".
2503         * diagnostic-event-id.h: New file.
2504         * diagnostic-format-json.cc (json_from_expanded_location): Make
2505         non-static.
2506         (json_end_diagnostic): Call context->make_json_for_path if it
2507         exists and the diagnostic has a path.
2508         (diagnostic_output_format_init): Clear context->print_path.
2509         * diagnostic-path.h: New file.
2510         * diagnostic-show-locus.c (colorizer::set_range): Special-case
2511         when printing a run of events in a diagnostic_path so that they
2512         all get the same color.
2513         (layout::m_diagnostic_path_p): New field.
2514         (layout::layout): Initialize it.
2515         (layout::print_any_labels): Don't colorize the label text for an
2516         event in a diagnostic_path.
2517         (gcc_rich_location::add_location_if_nearby): Add
2518         "restrict_to_current_line_spans" and "label" params.  Pass the
2519         former to layout.maybe_add_location_range; pass the latter
2520         when calling add_range.
2521         * diagnostic.c: Include "diagnostic-path.h".
2522         (diagnostic_initialize): Initialize context->path_format and
2523         context->show_path_depths.
2524         (diagnostic_show_any_path): New function.
2525         (diagnostic_path::interprocedural_p): New function.
2526         (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
2527         (simple_diagnostic_path::num_events): New function.
2528         (simple_diagnostic_path::get_event): New function.
2529         (simple_diagnostic_path::add_event): New function.
2530         (simple_diagnostic_event::simple_diagnostic_event): New ctor.
2531         (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
2532         (debug): New overload taking a diagnostic_path *.
2533         * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
2534         * diagnostic.h (enum diagnostic_path_format): New enum.
2535         (json::value): New forward decl.
2536         (diagnostic_context::path_format): New field.
2537         (diagnostic_context::show_path_depths): New field.
2538         (diagnostic_context::print_path): New callback field.
2539         (diagnostic_context::make_json_for_path): New callback field.
2540         (diagnostic_show_any_path): New decl.
2541         (json_from_expanded_location): New decl.
2542         * doc/invoke.texi (-fdiagnostics-path-format=): New option.
2543         (-fdiagnostics-show-path-depths): New option.
2544         (-fdiagnostics-color): Add "path" to description of default
2545         GCC_COLORS; describe it.
2546         (-fdiagnostics-format=json): Document how diagnostic paths are
2547         represented in the JSON output format.
2548         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2549         Add optional params "restrict_to_current_line_spans" and "label".
2550         * opts.c (common_handle_option): Handle
2551         OPT_fdiagnostics_path_format_ and
2552         OPT_fdiagnostics_show_path_depths.
2553         * pretty-print.c: Include "diagnostic-event-id.h".
2554         (pp_format): Implement "%@" format code for printing
2555         diagnostic_event_id_t *.
2556         (selftest::test_pp_format): Add tests for "%@".
2557         * selftest-run-tests.c (selftest::run_tests): Call
2558         selftest::tree_diagnostic_path_cc_tests.
2559         * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
2560         * toplev.c (general_init): Initialize global_dc->path_format and
2561         global_dc->show_path_depths.
2562         * tree-diagnostic-path.cc: New file.
2563         * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
2564         non-static.  Drop "diagnostic" param in favor of storing the
2565         original value of "where" and re-using it.
2566         (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
2567         maybe_unwind_expanded_macro_loc.
2568         (tree_diagnostics_defaults): Initialize context->print_path and
2569         context->make_json_for_path.
2570         * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
2571         decl.
2572         (default_tree_make_json_for_path): New decl.
2573         (maybe_unwind_expanded_macro_loc): New decl.
2574
2575 2020-01-10  Jakub Jelinek  <jakub@redhat.com>
2576
2577         PR tree-optimization/93210
2578         * fold-const.h (native_encode_initializer,
2579         can_native_interpret_type_p): Declare.
2580         * fold-const.c (native_encode_string): Fix up handling with off != -1,
2581         simplify.
2582         (native_encode_initializer): New function, moved from dwarf2out.c.
2583         Adjust to native_encode_expr compatible arguments, including dry-run
2584         and partial extraction modes.  Don't handle STRING_CST.
2585         (can_native_interpret_type_p): No longer static.
2586         * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
2587         offset / BITS_PER_UNIT fits into int and don't call it if
2588         can_native_interpret_type_p fails.  If suboff is NULL and for
2589         CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
2590         native_encode_initializer.
2591         (fold_const_aggregate_ref_1): Formatting fix.
2592         * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
2593         (tree_add_const_value_attribute): Adjust caller.
2594
2595         PR tree-optimization/90838
2596         * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2597         SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
2598         CTZ_DEFINED_VALUE_AT_ZERO.
2599
2600 2020-01-10  Vladimir Makarov  <vmakarov@redhat.com>
2601
2602         PR inline-asm/93027
2603         * lra-constraints.c (match_reload): Permit input operands have the
2604         same mode as output while other input operands have a different
2605         mode.
2606
2607 2020-01-10  Wilco Dijkstra  <wdijkstr@arm.com>
2608
2609         PR tree-optimization/90838
2610         * tree-ssa-forwprop.c (check_ctz_array): Add new function.
2611         (check_ctz_string): Likewise.
2612         (optimize_count_trailing_zeroes): Likewise.
2613         (simplify_count_trailing_zeroes): Likewise.
2614         (pass_forwprop::execute): Try ctz simplification.
2615         * match.pd: Add matching for ctz idioms.
2616
2617 2020-01-10  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
2618
2619         * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
2620         for target hook.
2621         (aarch64_invalid_unary_op): New function for target hook.
2622         (aarch64_invalid_binary_op): New function for target hook.
2623
2624 2020-01-10  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
2625
2626         * config.gcc: Add arm_bf16.h.
2627         * config/aarch64/aarch64-builtins.c
2628         (aarch64_simd_builtin_std_type): Add BFmode.
2629         (aarch64_init_simd_builtin_types): Define element types for vector
2630         types.
2631         (aarch64_init_bf16_types): New function.
2632         (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
2633         * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
2634         modes.
2635         * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
2636         * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
2637         patterns.
2638         * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
2639         (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
2640         * config/aarch64/aarch64.c
2641         (aarch64_classify_vector_mode): Add support for BF types.
2642         (aarch64_gimplify_va_arg_expr): Add support for BF types.
2643         (aarch64_vq_mode): Add support for BF types.
2644         (aarch64_simd_container_mode): Add support for BF types.
2645         (aarch64_mangle_type): Add support for BF scalar type.
2646         * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
2647         * config/aarch64/arm_bf16.h: New file.
2648         * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2649         * config/aarch64/iterators.md: Add BF types to mode attributes.
2650         (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
2651
2652 2020-01-10  Jason Merrill  <jason@redhat.com>
2653
2654         PR c++/93173 - incorrect tree sharing.
2655         * gimplify.c (copy_if_shared): No longer static.
2656         * gimplify.h: Declare it.
2657
2658 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
2659
2660         * doc/invoke.texi (-msve-vector-bits=): Document that
2661         -msve-vector-bits=128 now generates VL-specific code for
2662         little-endian targets.
2663         * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
2664         build_vector_type_for_mode to construct the data vector types.
2665         * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
2666         VL-specific code for -msve-vector-bits=128 on little-endian targets.
2667         (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
2668         for 128-bit vectors.
2669
2670 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
2671
2672         * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
2673         invocation.
2674
2675 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
2676
2677         * config/aarch64/aarch64-builtins.c
2678         (aarch64_builtin_vectorized_function): Check for specific vector modes,
2679         rather than checking the number of elements and the element mode.
2680
2681 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
2682
2683         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
2684         get_related_vectype_for_scalar_type rather than build_vector_type
2685         to create the index type for a conditional reduction.
2686
2687 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
2688
2689         * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
2690         for any type of gather or scatter, including strided accesses.
2691
2692 2020-01-10  Andre Vieira  <andre.simoesdiasvieira@arm.com>
2693
2694         * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
2695          comment.
2696
2697 2020-01-10  Andre Vieira  <andre.simoesdiasvieira@arm.com>
2698
2699         * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
2700         get_dr_vinfo_offset
2701         * tree-vect-loop.c (update_epilogue_loop_vinfo):  Remove orig_drs_init
2702         parameter and its use to reset DR_OFFSET's.
2703         (vect_transform_loop): Remove orig_drs_init argument.
2704         * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
2705         member of dr_vec_info rather than the offset of the associated
2706         data_reference's innermost_loop_behavior.
2707         (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
2708         (vect_do_peeling): Remove orig_drs_init parameter and its construction.
2709         * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
2710         get_dr_vinfo_offset.
2711         (vectorizable_store): Likewise.
2712         (vectorizable_load): Likewise.
2713
2714 2020-01-10  Richard Biener  <rguenther@suse.de>
2715
2716         * gimple-ssa-store-merging
2717         (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
2718
2719 2020-01-10  Martin Liska  <mliska@suse.cz>
2720
2721         PR ipa/93217
2722         * ipa-inline-analysis.c (offline_size): Make proper parenthesis
2723         encapsulation that was there before r280040.
2724
2725 2020-01-10  Richard Biener  <rguenther@suse.de>
2726
2727         PR middle-end/93199
2728         * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
2729         sequences to avoid walking them again for secondary opportunities.
2730         (pass_lower_eh_dispatch::execute): Instead actually insert
2731         them here.
2732
2733 2020-01-10  Richard Biener  <rguenther@suse.de>
2734
2735         PR middle-end/93199
2736         * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
2737         (cleanup_all_empty_eh): Walk landing pads in reverse order to
2738         avoid quadraticness.
2739
2740 2020-01-10  Martin Jambor  <mjambor@suse.cz>
2741
2742         * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
2743         * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
2744         to get param_ipa_sra_max_replacements.
2745         (param_splitting_across_edge): Pass the caller to
2746         pull_accesses_from_callee.
2747
2748 2020-01-10  Martin Jambor  <mjambor@suse.cz>
2749
2750         * params.opt (param_ipcp_unit_growth): Mark as Optimization.
2751         * ipa-cp.c (max_new_size): Removed.
2752         (orig_overall_size): New variable.
2753         (get_max_overall_size): New function.
2754         (estimate_local_effects): Use it.  Adjust dump.
2755         (decide_about_value): Likewise.
2756         (ipcp_propagate_stage): Do not calculate max_new_size, just store
2757         orig_overall_size.  Adjust dump.
2758         (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
2759
2760 2020-01-10  Martin Jambor  <mjambor@suse.cz>
2761
2762         * params.opt (param_ipa_max_agg_items): Mark as Optimization
2763         * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
2764         instead of param_ipa_max_agg_items.
2765         (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
2766         optimization info for the callee.
2767
2768 2020-01-09  Kwok Cheung Yeung  <kcy@codesourcery.com>
2769
2770         * lto-streamer-in.c (input_function): Remove streamed-in inline debug
2771         markers if debug_inline_points is false.
2772
2773 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
2774
2775         * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
2776         extra_objs.
2777         * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
2778         aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
2779         aarch64-sve-builtins-sve2.h.
2780         (aarch64-sve-builtins-sve2.o): New rule.
2781         * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
2782         (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
2783         (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
2784         (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
2785         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
2786         TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
2787         TARGET_SVE2_SM4.
2788         * config/aarch64/aarch64-sve.md: Update comments with SVE2
2789         instructions that are handled here.
2790         (@cond_asrd<mode>): Generalize to...
2791         (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
2792         (*cond_asrd<mode>_2): Generalize to...
2793         (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
2794         (*cond_asrd<mode>_z): Generalize to...
2795         (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
2796         * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
2797         (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
2798         (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
2799         * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
2800         pattern.
2801         (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
2802         (@aarch64_scatter_stnt<mode>): Likewise.
2803         (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
2804         (@aarch64_mul_lane_<mode>): Likewise.
2805         (@aarch64_sve_suqadd<mode>_const): Likewise.
2806         (*<sur>h<addsub><mode>): Generalize to...
2807         (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
2808         new pattern.
2809         (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
2810         (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
2811         (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
2812         (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
2813         (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
2814         (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
2815         (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
2816         (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
2817         (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
2818         (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
2819         (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
2820         (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
2821         (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
2822         (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
2823         (@aarch64_sve_add_mul_lane_<mode>): Likewise.
2824         (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
2825         (@aarch64_sve2_xar<mode>): Likewise.
2826         (@aarch64_sve2_bcax<mode>): Likewise.
2827         (*aarch64_sve2_eor3<mode>): Rename to...
2828         (@aarch64_sve2_eor3<mode>): ...this.
2829         (@aarch64_sve2_bsl<mode>): New expander.
2830         (@aarch64_sve2_nbsl<mode>): Likewise.
2831         (@aarch64_sve2_bsl1n<mode>): Likewise.
2832         (@aarch64_sve2_bsl2n<mode>): Likewise.
2833         (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
2834         (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
2835         (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
2836         (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
2837         (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
2838         (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
2839         (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
2840         (<su>mull<bt><Vwide>): Generalize to...
2841         (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
2842         pattern.
2843         (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
2844         (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
2845         (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
2846         (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2847         (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
2848         (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2849         (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
2850         (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2851         (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
2852         (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2853         (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
2854         (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
2855         (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
2856         (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
2857         (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
2858         (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
2859         (<SHRNB:r>shrnb<mode>): Generalize to...
2860         (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
2861         new pattern.
2862         (<SHRNT:r>shrnt<mode>): Generalize to...
2863         (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
2864         new pattern.
2865         (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
2866         (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
2867         (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
2868         (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
2869         (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
2870         (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
2871         (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
2872         (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
2873         (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
2874         (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
2875         (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
2876         (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
2877         (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
2878         (@aarch64_sve2_cvtnt<mode>): Likewise.
2879         (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
2880         (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
2881         (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
2882         (@aarch64_sve2_cvtxnt<mode>): Likewise.
2883         (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
2884         (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
2885         (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
2886         (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
2887         (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
2888         (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
2889         (@aarch64_sve2_pmul<mode>): Likewise.
2890         (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
2891         (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
2892         (@aarch64_sve2_tbl2<mode>): Likewise.
2893         (@aarch64_sve2_tbx<mode>): Likewise.
2894         (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
2895         (@aarch64_sve2_histcnt<mode>): Likewise.
2896         (@aarch64_sve2_histseg<mode>): Likewise.
2897         (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
2898         (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
2899         (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
2900         (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
2901         (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
2902         (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
2903         (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
2904         (<su>mulh<r>s<mode>3): Update after above pattern name changes.
2905         * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
2906         (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
2907         (SVE2_PMULL_PAIR_I): New mode iterators.
2908         (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
2909         (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
2910         (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
2911         (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
2912         (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
2913         (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
2914         (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
2915         (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
2916         (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
2917         (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
2918         (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
2919         (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
2920         (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
2921         (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
2922         (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
2923         (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
2924         (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
2925         (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
2926         (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
2927         (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
2928         (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
2929         (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
2930         (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
2931         (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
2932         (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
2933         (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
2934         (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
2935         (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
2936         (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
2937         (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
2938         (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
2939         further down file.
2940         (VNARROW, Ventype): New mode attributes.
2941         (Vewtype): Handle VNx2DI.  Fix typo in comment.
2942         (VDOUBLE): New mode attribute.
2943         (sve_lane_con): Handle VNx8HI.
2944         (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
2945         (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
2946         (sve_int_op, sve_int_op_rev): Handle the above codes.
2947         (sve_pred_int_rhs2_operand): Likewise.
2948         (MULLBT, SHRNB, SHRNT): Delete.
2949         (SVE_INT_SHIFT_IMM): New int iterator.
2950         (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
2951         and UNSPEC_WHILEHS for TARGET_SVE2.
2952         (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
2953         (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
2954         (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
2955         (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
2956         (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
2957         (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
2958         (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
2959         (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
2960         (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
2961         (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
2962         (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
2963         (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
2964         (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
2965         (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
2966         (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
2967         (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
2968         (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
2969         (optab): Handle the new unspecs.
2970         (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
2971         and UNSPEC_RSHRNT.
2972         (lr): Handle the new unspecs.
2973         (bt): Delete.
2974         (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
2975         (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
2976         (sve_int_qsub_op): New int attributes.
2977         (sve_fp_op, rot): Handle the new unspecs.
2978         * config/aarch64/aarch64-sve-builtins.h
2979         (function_resolver::require_matching_pointer_type): Declare.
2980         (function_resolver::resolve_unary): Add an optional boolean argument.
2981         (function_resolver::finish_opt_n_resolution): Add an optional
2982         type_suffix_index argument.
2983         (gimple_folder::redirect_call): Declare.
2984         (gimple_expander::prepare_gather_address_operands): Add an optional
2985         bool parameter.
2986         * config/aarch64/aarch64-sve-builtins.cc: Include
2987         aarch64-sve-builtins-sve2.h.
2988         (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
2989         (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
2990         (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
2991         (TYPES_hsd_integer): Use TYPES_hsd_signed.
2992         (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
2993         (TYPES_s_unsigned): Likewise.
2994         (TYPES_s_integer): Use TYPES_s_unsigned.
2995         (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
2996         (TYPES_sd_integer): Use them.
2997         (TYPES_d_unsigned): New macro.
2998         (TYPES_d_integer): Use it.
2999         (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3000         (TYPES_cvt_narrow): Likewise.
3001         (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3002         (preds_mx): New variable.
3003         (function_builder::add_overloaded_function): Allow the new feature
3004         set to be more restrictive than the original one.
3005         (function_resolver::infer_pointer_type): Remove qualifiers from
3006         the pointer type before printing it.
3007         (function_resolver::require_matching_pointer_type): New function.
3008         (function_resolver::resolve_sv_displacement): Handle functions
3009         that don't support 32-bit vector indices or svint32_t vector offsets.
3010         (function_resolver::finish_opt_n_resolution): Take the inferred type
3011         as a separate argument.
3012         (function_resolver::resolve_unary): Optionally treat all forms in
3013         the same way as normal merging functions.
3014         (gimple_folder::redirect_call): New function.
3015         (function_expander::prepare_gather_address_operands): Add an argument
3016         that says whether scaled forms are available.  If they aren't,
3017         handle scaling of vector indices and don't add the extension and
3018         scaling operands.
3019         (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3020         fall back to using cond_* instead.
3021         * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3022         Split out the member variables into...
3023         (rtx_code_function_base): ...this new base class.
3024         (rtx_code_function_rotated): Inherit rtx_code_function_base.
3025         (unspec_based_function): Split out the member variables into...
3026         (unspec_based_function_base): ...this new base class.
3027         (unspec_based_function_rotated): Inherit unspec_based_function_base.
3028         (unspec_based_function_exact_insn): New class.
3029         (unspec_based_add_function, unspec_based_add_lane_function)
3030         (unspec_based_lane_function, unspec_based_pred_function)
3031         (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3032         (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3033         (unspec_based_sub_function, unspec_based_sub_lane_function): New
3034         typedefs.
3035         (unspec_based_fused_function): New class.
3036         (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3037         (unspec_based_fused_lane_function): New class.
3038         (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3039         typedefs.
3040         (CODE_FOR_MODE1): New macro.
3041         (fixed_insn_function): New class.
3042         (while_comparison): Likewise.
3043         * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3044         (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3045         (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3046         (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3047         (load_gather_sv_restricted, shift_left_imm_long): Declare.
3048         (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3049         (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3050         (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3051         (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3052         (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3053         (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3054         (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3055         (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3056         * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3057         Also add an initial argument for unary_convert_narrowt, regardless
3058         of the predication type.
3059         (build_32_64): Allow loads and stores to specify MODE_none.
3060         (build_sv_index64, build_sv_uint_offset): New functions.
3061         (long_type_suffix): New function.
3062         (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3063         (binary_imm_long_base, load_gather_sv_base): Likewise.
3064         (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3065         (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3066         (unary_narrowb_base, unary_narrowt_base): Likewise.
3067         (binary_long_lane_def, binary_long_lane): New shape.
3068         (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3069         (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3070         (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3071         (binary_to_uint_def, binary_to_uint): Likewise.
3072         (binary_wide_def, binary_wide): Likewise.
3073         (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3074         (compare_def, compare): Likewise.
3075         (compare_ptr_def, compare_ptr): Likewise.
3076         (load_ext_gather_index_restricted_def,
3077         load_ext_gather_index_restricted): Likewise.
3078         (load_ext_gather_offset_restricted_def,
3079         load_ext_gather_offset_restricted): Likewise.
3080         (load_gather_sv_def): Inherit from load_gather_sv_base.
3081         (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3082         (shift_left_imm_def, shift_left_imm): Likewise.
3083         (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3084         (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3085         (store_scatter_index_restricted_def,
3086         store_scatter_index_restricted): Likewise.
3087         (store_scatter_offset_restricted_def,
3088         store_scatter_offset_restricted): Likewise.
3089         (tbl_tuple_def, tbl_tuple): Likewise.
3090         (ternary_long_lane_def, ternary_long_lane): Likewise.
3091         (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3092         (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3093         (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3094         (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3095         (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3096         (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3097         (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3098         (ternary_uint_def, ternary_uint): Likewise.
3099         (unary_convert): Fix typo in comment.
3100         (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3101         (unary_long_def, unary_long): Likewise.
3102         (unary_narrowb_def, unary_narrowb): Likewise.
3103         (unary_narrowt_def, unary_narrowt): Likewise.
3104         (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3105         (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3106         (unary_to_int_def, unary_to_int): Likewise.
3107         * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3108         (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3109         (svasrd_impl): Delete.
3110         (svcadd_impl::expand): Handle integer operations too.
3111         (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3112         new functions to derive the unspec numbers.
3113         (svmla_svmls_lane_impl): Replace with...
3114         (svmla_lane_impl, svmls_lane_impl): ...these new classes.  Handle
3115         integer operations too.
3116         (svwhile_impl): Rename to...
3117         (svwhilelx_impl): ...this and inherit from while_comparison.
3118         (svasrd): Use unspec_based_function.
3119         (svmla_lane): Use svmla_lane_impl.
3120         (svmls_lane): Use svmls_lane_impl.
3121         (svrecpe, svrsqrte): Handle unsigned integer operations too.
3122         (svwhilele, svwhilelt): Use svwhilelx_impl.
3123         * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3124         * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3125         * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3126         * config/aarch64/aarch64-sve-builtins.def: Include
3127         aarch64-sve-builtins-sve2.def.
3128
3129 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
3130
3131         * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3132         (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3133         * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3134         (aarch64_sve_sqadd_sqsub_immediate_p): Likewise.  Handle scalar
3135         immediates as well as vector ones.
3136         * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3137         (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3138         (aarch64_sve_qsub_immediate): Update calls accordingly.
3139
3140 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
3141
3142         * config/aarch64/aarch64-sve2.md: Add banner comments.
3143         (<su>mulh<r>s<mode>3): Move further up file.
3144         (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3145         (*aarch64_sve2_sra<mode>): Move further down file.
3146         * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3147
3148 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
3149
3150         * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3151         and UNSPEC_WHILEWR.
3152         (while_optab_cmp): Handle them.
3153         * config/aarch64/aarch64-sve.md
3154         (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3155         and add a "@" marker.
3156         * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3157         instead of gen_aarch64_sve2_while_ptest.
3158         (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3159
3160 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
3161
3162         * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3163         (UNSPEC_WHILELE): ...this.
3164         (UNSPEC_WHILE_LO): Rename to...
3165         (UNSPEC_WHILELO): ...this.
3166         (UNSPEC_WHILE_LS): Rename to...
3167         (UNSPEC_WHILELS): ...this.
3168         (UNSPEC_WHILE_LT): Rename to...
3169         (UNSPEC_WHILELT): ...this.
3170         * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3171         (cmp_op, while_optab_cmp): Likewise.
3172         * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3173         * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3174         (svwhilelt): Likewise.
3175
3176 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
3177
3178         * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3179         (unary_to_uint): Define.
3180         * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3181         (unary_count): Rename to...
3182         (unary_to_uint_def, unary_to_uint): ...this.
3183         * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3184
3185 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
3186
3187         * config/aarch64/aarch64-sve-builtins-functions.h
3188         (code_for_mode_function): New class.
3189         (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3190         * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3191         (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3192         (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3193         (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3194
3195 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
3196
3197         * config/aarch64/iterators.md (addsub): New code attribute.
3198         * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3199         Re-express as...
3200         (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3201         in the asm string and attributes.  Fix indentation.
3202         * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3203         Re-express as...
3204         (@aarch64_sve_<optab><mode>): ...this.
3205         * config/aarch64/aarch64-sve-builtins.h
3206         (function_expander::expand_signed_unpred_op): Delete.
3207         * config/aarch64/aarch64-sve-builtins.cc
3208         (function_expander::expand_signed_unpred_op): Likewise.
3209         (function_expander::map_to_rtx_codes): If the optab isn't defined,
3210         try using code_for_aarch64_sve instead.
3211         * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3212         (svqsub_impl): Likewise.
3213         (svqadd, svqsub): Use rtx_code_function instead.
3214
3215 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
3216
3217         * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3218         (HADDSUB, sur, addsub): Remove them.
3219
3220 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
3221
3222         * tree-nrv.c (pass_return_slot::execute): Handle all internal
3223         functions the same way, rather than singling out those that
3224         aren't mapped directly to optabs.
3225
3226 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
3227
3228         * target.def (compatible_vector_types_p): New target hook.
3229         * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3230         * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3231         * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3232         * doc/tm.texi: Regenerate.
3233         * gimple-expr.c: Include target.h.
3234         (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3235         * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3236         function.
3237         (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3238         * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3239         Use the original predicate if it already has a suitable type.
3240
3241 2020-01-09  Martin Jambor  <mjambor@suse.cz>
3242
3243         * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3244         resolve_speculation and redirect_call_stmt_to_callee static.  Change
3245         return type of set_call_stmt to cgraph_edge *.
3246         * auto-profile.c (afdo_indirect_call): Adjust call to
3247         redirect_call_stmt_to_callee.
3248         * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3249         make the this pointer explicit, adjust self-recursive calls and the
3250         call top make_direct.  Return the resulting edge.
3251         (cgraph_edge::remove): Make this pointer explicit.
3252         (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3253         (cgraph_edge::make_direct): Likewise, adjust call to
3254         resolve_speculation.
3255         (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3256         call to set_call_stmt.
3257         (cgraph_update_edges_for_call_stmt_node): Update call to
3258         set_call_stmt and remove.
3259         * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3260         Renamed edge to master_edge.  Adjusted calls to set_call_stmt.
3261         (cgraph_node::create_edge_including_clones): Moved "first" definition
3262         of edge to the block where it was used.  Adjusted calls to
3263         set_call_stmt.
3264         (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3265         cgraph_edge::remove.
3266         * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3267         make_direct and redirect_call_stmt_to_callee.
3268         * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3269         resolve_speculation and make_direct.
3270         * ipa-inline-transform.c (inline_transform): Adjust call to
3271         redirect_call_stmt_to_callee.
3272         (check_speculations_1):: Adjust call to resolve_speculation.
3273         * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3274         resolve-speculation.
3275         (inline_small_functions): Adjust call to resolve_speculation.
3276         (ipa_inline): Likewise.
3277         * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3278         make_direct.
3279         * ipa-visibility.c (function_and_variable_visibility): Make iteration
3280         safe with regards to edge removal, adjust calls to
3281         redirect_call_stmt_to_callee.
3282         * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3283         and redirect_call_stmt_to_callee.
3284         * multiple_target.c (create_dispatcher_calls): Adjust call to
3285         redirect_call_stmt_to_callee
3286         (redirect_to_specific_clone): Likewise.
3287         * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3288         Adjust calls to cgraph_edge::remove.
3289         * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3290         (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3291         (expand_call_inline): Adjust call to cgraph_edge::remove.
3292
3293 2020-01-09  Martin Liska  <mliska@suse.cz>
3294
3295         * params.opt: Set Optimization for
3296         param_max_speculative_devirt_maydefs.
3297
3298 2020-01-09  Martin Sebor  <msebor@redhat.com>
3299
3300         PR middle-end/93200
3301         PR fortran/92956
3302         * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3303
3304 2020-01-09  Martin Liska  <mliska@suse.cz>
3305
3306         * auto-profile.c (auto_profile): Use opt_for_fn
3307         for a parameter.
3308         * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3309         (propagate_vals_across_arith_jfunc): Likewise.
3310         (hint_time_bonus): Likewise.
3311         (incorporate_penalties): Likewise.
3312         (good_cloning_opportunity_p): Likewise.
3313         (perform_estimation_of_a_value): Likewise.
3314         (estimate_local_effects): Likewise.
3315         (ipcp_propagate_stage): Likewise.
3316         * ipa-fnsummary.c (decompose_param_expr): Likewise.
3317         (set_switch_stmt_execution_predicate): Likewise.
3318         (analyze_function_body): Likewise.
3319         * ipa-inline-analysis.c (offline_size): Likewise.
3320         * ipa-inline.c (early_inliner): Likewise.
3321         * ipa-prop.c (ipa_analyze_node): Likewise.
3322         (ipcp_transform_function): Likewise.
3323         * ipa-sra.c (process_scan_results): Likewise.
3324         (ipa_sra_summarize_function): Likewise.
3325         * params.opt: Rename ipcp-unit-growth to
3326         ipa-cp-unit-growth.  Add Optimization for various
3327         IPA-related parameters.
3328
3329 2020-01-09  Richard Biener  <rguenther@suse.de>
3330
3331         PR middle-end/93054
3332         * gimplify.c (gimplify_expr): Deal with NOP definitions.
3333
3334 2020-01-09  Richard Biener  <rguenther@suse.de>
3335
3336         PR tree-optimization/93040
3337         * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3338
3339 2020-01-09  Georg-Johann Lay  <avr@gjlay.de>
3340
3341         * common/config/avr/avr-common.c (avr_option_optimization_table)
3342         [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3343
3344 2020-01-09  Martin Liska  <mliska@suse.cz>
3345
3346         * cgraphclones.c (symbol_table::materialize_all_clones):
3347         Use cgraph_node::dump_name.
3348
3349 2020-01-09  Jakub Jelinek  <jakub@redhat.com>
3350
3351         PR inline-asm/93202
3352         * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3353         output_operand_lossage instead of gcc_unreachable.
3354         * doc/md.texi (riscv f constraint): Fix typo.
3355
3356         PR target/93141
3357         * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3358         SWI.  Use <general_hilo_operand> instead of <general_operand>.  Use
3359         CONST_SCALAR_INT_P instead of CONST_INT_P.
3360         (*subv<mode>4_1): Rename to ...
3361         (subv<mode>4_1): ... this.
3362         (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3363         define_insn_and_split patterns.
3364         (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3365         patterns.
3366
3367 2020-01-08  David Malcolm  <dmalcolm@redhat.com>
3368
3369         * vec.c (class selftest::count_dtor): New class.
3370         (selftest::test_auto_delete_vec): New test.
3371         (selftest::vec_c_tests): Call it.
3372         * vec.h (class auto_delete_vec): New class template.
3373         (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3374
3375 2020-01-08  David Malcolm  <dmalcolm@redhat.com>
3376
3377         * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3378
3379 2020-01-08  Jim Wilson  <jimw@sifive.com>
3380
3381         * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3382         use of TLS_MODEL_LOCAL_EXEC when not pic.
3383
3384 2020-01-08  David Malcolm  <dmalcolm@redhat.com>
3385
3386         * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3387         memory leak.
3388
3389 2020-01-08  Jakub Jelinek  <jakub@redhat.com>
3390
3391         PR target/93187
3392         * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3393         *stack_protect_set_3 peephole2): Also check that the second
3394         insns source is general_operand.
3395
3396         PR target/93174
3397         * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3398         predicate for output operand instead of register_operand.
3399         (addcarry<mode>, addcarry<mode>_1): Likewise.  Add alternative with
3400         memory destination and non-memory operands[2].
3401
3402 2020-01-08  Martin Liska  <mliska@suse.cz>
3403
3404         * cgraph.c (cgraph_node::dump): Use ::dump_name or
3405         ::dump_asm_name instead of (::name or ::asm_name).
3406         * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3407         * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3408         (analyze_functions): Likewise.
3409         (expand_all_functions): Likewise.
3410         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3411         (propagate_bits_across_jump_function): Likewise.
3412         (dump_profile_updates): Likewise.
3413         (ipcp_store_bits_results): Likewise.
3414         (ipcp_store_vr_results): Likewise.
3415         * ipa-devirt.c (dump_targets): Likewise.
3416         * ipa-fnsummary.c (analyze_function_body): Likewise.
3417         * ipa-hsa.c (check_warn_node_versionable): Likewise.
3418         (process_hsa_functions): Likewise.
3419         * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3420         (set_alias_uids): Likewise.
3421         * ipa-inline-transform.c (save_inline_function_body): Likewise.
3422         * ipa-inline.c (recursive_inlining): Likewise.
3423         (inline_to_all_callers_1): Likewise.
3424         (ipa_inline): Likewise.
3425         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3426         (ipa_propagate_frequency): Likewise.
3427         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3428         (remove_described_reference): Likewise.
3429         * ipa-pure-const.c (worse_state): Likewise.
3430         (check_retval_uses): Likewise.
3431         (analyze_function): Likewise.
3432         (propagate_pure_const): Likewise.
3433         (propagate_nothrow): Likewise.
3434         (dump_malloc_lattice): Likewise.
3435         (propagate_malloc): Likewise.
3436         (pass_local_pure_const::execute): Likewise.
3437         * ipa-visibility.c (optimize_weakref): Likewise.
3438         (function_and_variable_visibility): Likewise.
3439         * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3440         (ipa_discover_variable_flags): Likewise.
3441         * lto-streamer-out.c (output_function): Likewise.
3442         (output_constructor): Likewise.
3443         * tree-inline.c (copy_bb): Likewise.
3444         * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3445         * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3446
3447 2020-01-08  Richard Biener  <rguenther@suse.de>
3448
3449         PR middle-end/93199
3450         * tree-eh.c (sink_clobbers): Update virtual operands for
3451         the first and last stmt only.  Add a dry-run capability.
3452         (pass_lower_eh_dispatch::execute): Perform clobber sinking
3453         after CFG manipulations and in RPO order to catch all
3454         secondary opportunities reliably.
3455
3456 2020-01-08  Georg-Johann Lay  <avr@gjlay.de>
3457
3458         PR target/93182
3459         * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3460
3461 2019-01-08  Richard Biener  <rguenther@suse.de>
3462
3463         PR middle-end/93199
3464         * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3465         * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3466         virtual operand, also updating SSA use.
3467         * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3468         Update stmt after resetting virtual operand.
3469         (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3470         * gimple-iterator.c (gsi_remove): When not removing the stmt
3471         permanently do not delink immediate uses or mark the stmt modified.
3472
3473 2020-01-08  Martin Liska  <mliska@suse.cz>
3474
3475         * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
3476         (ipa_call_context::estimate_size_and_time): Likewise.
3477         (inline_analyze_function): Likewise.
3478
3479 2020-01-08  Martin Liska  <mliska@suse.cz>
3480
3481         * cgraph.c (cgraph_node::dump): Use systematically
3482         dump_asm_name.
3483
3484 2020-01-08  Georg-Johann Lay  <avr@gjlay.de>
3485
3486         Add -nodevicespecs option for avr.
3487
3488         PR target/93182
3489         * config/avr/avr.opt (-nodevicespecs): New driver option.
3490         * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
3491         "-specs=device-specs/..." if that option is not set.
3492         * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3493
3494 2020-01-08  Georg-Johann Lay  <avr@gjlay.de>
3495
3496         Implement 64-bit double functions for avr.
3497
3498         PR target/92055
3499         * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
3500         --with-double-comparison.
3501         * doc/install.texi: Document them.
3502         * config/avr/avr-c.c (avr_cpu_cpp_builtins)
3503         <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
3504         <WITH_DOUBLE_COMPARISON>: New built-in defines.
3505         * doc/invoke.texi (AVR Built-in Macros): Document them.
3506         * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
3507         * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
3508         * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
3509
3510 2020-01-08  Richard Earnshaw  <rearnsha@arm.com>
3511
3512         PR target/93188
3513         * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
3514         armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
3515         when only building rm-profile multilibs.
3516
3517 2020-01-08  Feng Xue  <fxue@os.amperecomputing.com>
3518
3519         PR ipa/93084
3520         * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
3521         lattice for a value to check.
3522         (propagate_vals_across_arith_jfunc): Add an assertion to ensure
3523         finite propagation in self-recursive scc.
3524
3525 2020-01-08  Luo Xiong Hu  <luoxhu@linux.ibm.com>
3526
3527         * ipa-inline.c (caller_growth_limits): Restore the AND.
3528
3529 2020-01-07  Andrew Stubbs  <ams@codesourcery.com>
3530
3531         * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
3532         (VEC_ALLREG_ALT): New iterator.
3533         (VEC_ALLREG_INT_MODE): New iterator.
3534         (VCMP_MODE): New iterator.
3535         (VCMP_MODE_INT): New iterator.
3536         (vec_cmpu<mode>di): Use VCMP_MODE_INT.
3537         (vec_cmp<u>v64qidi): New define_expand.
3538         (vec_cmp<mode>di_exec): Use VCMP_MODE.
3539         (vec_cmpu<mode>di_exec): New define_expand.
3540         (vec_cmp<u>v64qidi_exec): New define_expand.
3541         (vec_cmp<mode>di_dup): Use VCMP_MODE.
3542         (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
3543         (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
3544         (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
3545         (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
3546         (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
3547         (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
3548         (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
3549         (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
3550         (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
3551         this.
3552         * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
3553         * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
3554
3555 2020-01-07  Andrew Stubbs  <ams@codesourcery.com>
3556
3557         * config/gcn/constraints.md (DA): Update description and match.
3558         (DB): Likewise.
3559         (Db): New constraint.
3560         * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
3561         parameter.
3562         * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
3563         Implement 'Db' mixed immediate type.
3564         * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
3565         (addcv64si3_dup<exec_vcc>): Delete.
3566         (subcv64si3<exec_vcc>): Rework constraints.
3567         (addv64di3): Rework constraints.
3568         (addv64di3_exec): Rework constraints.
3569         (subv64di3): Rework constraints.
3570         (addv64di3_dup): Delete.
3571         (addv64di3_dup_exec): Delete.
3572         (addv64di3_zext): Rework constraints.
3573         (addv64di3_zext_exec): Rework constraints.
3574         (addv64di3_zext_dup): Rework constraints.
3575         (addv64di3_zext_dup_exec): Rework constraints.
3576         (addv64di3_zext_dup2): Rework constraints.
3577         (addv64di3_zext_dup2_exec): Rework constraints.
3578         (addv64di3_sext_dup2): Rework constraints.
3579         (addv64di3_sext_dup2_exec): Rework constraints.
3580
3581 2020-01-07  Andre Vieira  <andre.simoesdiasvieira@arm.com>
3582
3583         * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
3584         existing target checks.
3585
3586 2020-01-07  Richard Biener  <rguenther@suse.de>
3587
3588         * doc/install.texi: Bump minimal supported MPC version.
3589
3590 2020-01-07  Richard Sandiford  <richard.sandiford@arm.com>
3591
3592         * langhooks-def.h (lhd_simulate_enum_decl): Declare.
3593         (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
3594         * langhooks.c: Include stor-layout.h.
3595         (lhd_simulate_enum_decl): New function.
3596         * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
3597         handle_arm_sve_h for the LTO frontend.
3598         (register_vector_type): Cope with null returns from pushdecl.
3599
3600 2020-01-07  Richard Sandiford  <richard.sandiford@arm.com>
3601
3602         * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
3603         (aarch64_sve::nvectors_if_data_type): Replace with...
3604         (aarch64_sve::builtin_type_p): ...this.
3605         * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
3606         (find_vector_type): Delete.
3607         (add_sve_type_attribute): New function.
3608         (lookup_sve_type_attribute): Likewise.
3609         (register_builtin_types): Add an "SVE type" attribute to each type.
3610         (register_tuple_type): Likewise.
3611         (svbool_type_p, nvectors_if_data_type): Delete.
3612         (mangle_builtin_type): Use lookup_sve_type_attribute.
3613         (builtin_type_p): Likewise.  Add an overload that returns the
3614         number of constituent vector and predicate registers.
3615         * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
3616         (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
3617         instead of aarch64_sve_argument_p.
3618         (aarch64_takes_arguments_in_sve_regs_p): Likewise.
3619         (aarch64_pass_by_reference): Likewise.
3620         (aarch64_function_value_1): Likewise.
3621         (aarch64_return_in_memory): Likewise.
3622         (aarch64_layout_arg): Likewise.
3623
3624 2020-01-07  Jakub Jelinek  <jakub@redhat.com>
3625
3626         PR tree-optimization/93156
3627         * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
3628         least significant bit is always clear.
3629
3630         PR tree-optimization/93118
3631         * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?.  Add new
3632         simplifier with two intermediate conversions.
3633
3634 2020-01-07  Martin Liska  <mliska@suse.cz>
3635
3636         * params.opt: Add Optimization for various parameters.
3637
3638 2020-01-07  Martin Liska  <mliska@suse.cz>
3639
3640         PR ipa/83411
3641         * doc/extend.texi: Explain cloning for target_clone
3642         attribute.
3643
3644 2020-01-07  Martin Liska  <mliska@suse.cz>
3645
3646         PR tree-optimization/92860
3647         * common.opt: Make in Optimization option
3648         as it is affected by -O0, which is an Optimization
3649         option.
3650         * tree-inline.c (tree_inlinable_function_p):
3651         Use opt_for_fn for warn_inline.
3652         (expand_call_inline): Likewise.
3653
3654 2020-01-07  Martin Liska  <mliska@suse.cz>
3655
3656         PR tree-optimization/92860
3657         * common.opt: Make flag_ree as optimization
3658         attribute.
3659
3660 2020-01-07  Martin Liska  <mliska@suse.cz>
3661
3662         PR optimization/92860
3663         * params.opt: Mark param_min_crossjump_insns with Optimization
3664         keyword.
3665
3666 2020-01-07  Luo Xiong Hu  <luoxhu@linux.ibm.com>
3667
3668         * ipa-inline-analysis.c (estimate_growth): Fix typo.
3669         * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
3670
3671 2020-01-06  Michael Meissner  <meissner@linux.ibm.com>
3672
3673         * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
3674         helper function to return the valid addressing formats for a given
3675         hard register and mode.
3676         (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
3677
3678         * config/rs6000/constraints.md (Q constraint): Update
3679         documentation.
3680         * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
3681         documentation.
3682
3683         * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
3684         Use 'Q' for doing vector extract from memory.
3685         (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
3686         memory.
3687         (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
3688         doing vector extract from memory.
3689         (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
3690         extract from memory.
3691
3692         * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
3693         for the offset being 34-bits when -mcpu=future is used.
3694
3695 2020-01-06  John David Anglin  <danglin@gcc.gnu.org>
3696
3697         * config/pa/pa.md: Revert change to use ordered_comparison_operator
3698         instead of cmpib_comparison_operator in cmpib patterns.
3699         * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
3700         of cmpib_comparison_operator.  Revise comment.
3701
3702 2020-01-06  Richard Sandiford  <richard.sandiford@arm.com>
3703
3704         * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
3705         in an IFN_DIV_POW2 node to be equal.
3706
3707 2020-01-06  Richard Sandiford  <richard.sandiford@arm.com>
3708
3709         * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
3710         (vect_check_scalar_mask): ...this.
3711         (vectorizable_store, vectorizable_load): Update call accordingly.
3712         (vectorizable_call): Use vect_check_scalar_mask to check the mask
3713         argument in calls to conditional internal functions.
3714
3715 2020-01-06  Andrew Stubbs  <ams@codesourcery.com>
3716
3717         * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
3718         '0' matching inputs.
3719         (subv64di3_exec): Likewise.
3720
3721 2020-01-06  Bryan Stenson  <bryan@siliconvortex.com>
3722
3723         * config/mips/mips.c (vr4130_align_insns): Fix typo.
3724         * doc/md.texi (movstr): Likewise.
3725
3726 2020-01-06  Andrew Stubbs  <ams@codesourcery.com>
3727
3728         * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
3729         clobber.
3730
3731 2020-01-06  Richard Sandiford  <richard.sandiford@arm.com>
3732
3733         * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
3734         Depend on...
3735         (s-aarch64-tune-md): ...this new stamp file.  Pipe the new contents
3736         to a temporary file and use move-if-change to update the real
3737         file where necessary.
3738
3739 2020-01-06  Richard Sandiford  <richard.sandiford@arm.com>
3740
3741         * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
3742         rather than Upa for CPY /M.
3743
3744 2020-01-06  Andrew Stubbs  <ams@codesourcery.com>
3745
3746         * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
3747         immediate.
3748
3749 2020-01-06  Martin Liska  <mliska@suse.cz>
3750
3751     PR tree-optimization/92860
3752     * params.opt: Mark param_max_combine_insns with Optimization
3753     keyword.
3754
3755 2020-01-05  Jakub Jelinek  <jakub@redhat.com>
3756
3757         PR target/93141
3758         * config/i386/i386.md (SWIDWI): New mode iterator.
3759         (DWI, dwi): Add TImode variants.
3760         (addv<mode>4): Use SWIDWI iterator instead of SWI.  Use
3761         <general_hilo_operand> instead of <general_operand>.  Use
3762         CONST_SCALAR_INT_P instead of CONST_INT_P.
3763         (*addv<mode>4_1): Rename to ...
3764         (addv<mode>4_1): ... this.
3765         (QWI): New mode attribute.
3766         (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3767         define_insn_and_split patterns.
3768         (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3769         patterns.
3770         (uaddv<mode>4): Use SWIDWI iterator instead of SWI.  Use
3771         <general_hilo_operand> instead of <general_operand>.
3772         (*addcarry<mode>_1): New define_insn.
3773         (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
3774
3775 2020-01-03  Konstantin Kharlamov  <Hi-Angel@yandex.ru>
3776
3777         * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
3778         Use "call" instead of "set".
3779
3780 2020-01-03  Martin Jambor  <mjambor@suse.cz>
3781
3782         PR ipa/92917
3783         * ipa-cp.c (print_all_lattices): Skip functions without info.
3784
3785 2020-01-03  Jakub Jelinek  <jakub@redhat.com>
3786
3787         PR target/93089
3788         * config/i386/i386-options.c (ix86_simd_clone_adjust): If
3789         TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
3790         simd clones.  If TARGET_PREFER_AVX256, use prefer-vector-width=512
3791         for 'e' simd clones.
3792
3793         PR target/93089
3794         * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
3795         entry.
3796         (mprefer-vector-width=): Add Save.
3797         * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
3798         -mprefer-vector-width= if non-zero.  Fix up -mfpmath= comment.
3799         (ix86_debug_options, ix86_function_specific_print): Adjust
3800         ix86_target_string callers.
3801         (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
3802         (ix86_valid_target_attribute_tree): Likewise.
3803         * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
3804         * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
3805         ix86_target_string caller.
3806
3807         PR target/93110
3808         * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
3809         emitting ASHIFTRT, XOR and MINUS by hand.  Use gen_int_mode with QImode
3810         instead of gen_int_shift_amount + convert_modes.
3811
3812         PR rtl-optimization/93088
3813         * loop-iv.c (find_single_def_src): Punt after looking through
3814         128 reg copies for regs with single definitions.  Move definitions
3815         to first uses.
3816
3817 2020-01-02  Dennis Zhang  <dennis.zhang@arm.com>
3818
3819         * config/arm/arm-c.c (arm_cpu_builtins): Define
3820         __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
3821         __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
3822         __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
3823         * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
3824         * config/arm/arm-tables.opt: Regenerated.
3825         * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
3826         arm_arch_i8mm and arm_arch_bf16 when enabled.
3827         * config/arm/arm.h (TARGET_I8MM): New macro.
3828         (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
3829         * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
3830         * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
3831         * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
3832         (v8_6_a_simd_variants): New.
3833         (v8_*_a_simd_variants): Add i8mm and bf16.
3834         * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
3835
3836 2020-01-02  Jakub Jelinek  <jakub@redhat.com>
3837
3838         PR ipa/93087
3839         * predict.c (compute_function_frequency): Don't call
3840         warn_function_cold on functions that already have cold attribute.
3841
3842 2020-01-01  John David Anglin  <danglin@gcc.gnu.org>
3843
3844         PR target/67834
3845         * config/pa/pa.c (pa_elf_select_rtx_section): New.  Put references to
3846         COMDAT group function labels in .data.rel.ro.local section.
3847         * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
3848
3849         PR target/93111
3850         * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
3851         comparison_operator in B and S integer comparisons.  Likewise, use
3852         ordered_comparison_operator instead of cmpib_comparison_operator in
3853         cmpib patterns.
3854         * config/pa/predicates.md (cmpib_comparison_operator): Remove.
3855
3856 2020-01-01  Jakub Jelinek  <jakub@redhat.com>
3857
3858         Update copyright years.
3859
3860         * gcc.c (process_command): Update copyright notice dates.
3861         * gcov-dump.c (print_version): Ditto.
3862         * gcov.c (print_version): Ditto.
3863         * gcov-tool.c (print_version): Ditto.
3864         * gengtype.c (create_file): Ditto.
3865         * doc/cpp.texi: Bump @copying's copyright year.
3866         * doc/cppinternals.texi: Ditto.
3867         * doc/gcc.texi: Ditto.
3868         * doc/gccint.texi: Ditto.
3869         * doc/gcov.texi: Ditto.
3870         * doc/install.texi: Ditto.
3871         * doc/invoke.texi: Ditto.
3872
3873 2020-01-01  Jan Hubicka  <hubicka@ucw.cz>
3874
3875         * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
3876         summary.
3877
3878 2020-01-01  Jakub Jelinek  <jakub@redhat.com>
3879
3880         PR tree-optimization/93098
3881         * match.pd (popcount): For shift amounts, use integer_onep
3882         or wi::to_widest () == cst instead of tree_to_uhwi () == cst
3883         tests.  Make sure that precision is power of two larger than or equal
3884         to 16.  Ensure shift is never negative.  Use HOST_WIDE_INT_UC macro
3885         instead of ULL suffixed constants.  Formatting fixes.
3886 \f
3887 Copyright (C) 2020 Free Software Foundation, Inc.
3888
3889 Copying and distribution of this file, with or without modification,
3890 are permitted in any medium without royalty provided the copyright
3891 notice and this notice are preserved.