gcc/ChangeLog

   1 2020-03-11  Eric Botcazou  <ebotcazou@adacore.com>
   2
   3         PR middle-end/93961
   4         * tree.c (variably_modified_type_p) <RECORD_TYPE>: Recurse into fields
   5         whose type is a qualified union.
   6
   7 2020-03-11  Jakub Jelinek  <jakub@redhat.com>
   8
   9         PR target/94121
  10         * config/aarch64/aarch64.c (aarch64_add_offset_1): Use absu_hwi
  11         instead of abs_hwi, change moffset type to unsigned HOST_WIDE_INT.
  12
  13         PR bootstrap/93962
  14         * value-prof.c (dump_histogram_value): Use abs_hwi instead of
  15         std::abs.
  16         (get_nth_most_common_value): Use abs_hwi instead of abs.
  17
  18         PR middle-end/94111
  19         * dfp.c (decimal_to_binary): Only use decimal128ToString if from->cl
  20         is rvc_normal, otherwise use real_to_decimal to print the number to
  21         string.
  22
  23         PR tree-optimization/94114
  24         * tree-loop-distribution.c (generate_memset_builtin): Call
  25         rewrite_to_non_trapping_overflow even on mem.
  26         (generate_memcpy_builtin): Call rewrite_to_non_trapping_overflow even
  27         on dest and src.
  28
  29 2020-03-10  Jeff Law  <law@redhat.com>
  30
  31         * config/bfin/bfin.md (movsi_insv): Add length attribute.
  32
  33 2020-03-10  Jiufu Guo  <guojiufu@linux.ibm.com>
  34
  35         PR target/93709
  36         * gcc/config/rs6000/rs6000.c (rs6000_emit_p9_fp_minmax): Check
  37         NAN and SIGNED_ZEROR for smax/smin.
  38
  39 2020-03-10  Will Schmidt  <will_schmidt@vnet.ibm.com>
  40
  41         PR target/90763
  42         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
  43         clause to handle P9V_BUILTIN_VEC_LXVL with const arguments.
  44
  45 2020-03-10  Roman Zhuykov  <zhroma@ispras.ru>
  46
  47         * loop-iv.c (find_simple_exit): Make it static.
  48         * cfgloop.h: Remove the corresponding prototype.
  49
  50 2020-03-10  Roman Zhuykov  <zhroma@ispras.ru>
  51
  52         * ddg.c (create_ddg): Fix intendation.
  53         (set_recurrence_length): Likewise.
  54         (create_ddg_all_sccs): Likewise.
  55
  56 2020-03-10  Jakub Jelinek  <jakub@redhat.com>
  57
  58         PR target/94088
  59         * config/i386/i386.md (*testqi_ext_3): Call ix86_match_ccmode with
  60         CCZmode instead of CCNOmode if operands[2] has DImode and pos + len
  61         is 32.
  62
  63 2020-03-09  Jason Merrill  <jason@redhat.com>
  64
  65         * gdbinit.in (pgs): Fix typo in documentation.
  66
  67 2020-03-09  Vladimir Makarov  <vmakarov@redhat.com>
  68
  69         Revert:
  70
  71         2020-02-28  Vladimir Makarov  <vmakarov@redhat.com>
  72
  73         PR rtl-optimization/93564
  74         * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
  75         do not honor reg alloc order.
  76
  77 2020-03-09  Andrew Pinski  <apinski@marvell.com>
  78
  79         PR inline-asm/94095
  80         * doc/extend.texi (x86 Operand Modifiers): Fix column
  81         for 'A' modifier.
  82
  83 2020-03-09  Martin Liska  <mliska@suse.cz>
  84
  85         PR target/93800
  86         * config/rs6000/rs6000.c (rs6000_option_override_internal):
  87         Remove set of str_align_loops and str_align_jumps as these
  88         should be set in previous 2 conditions in the function.
  89
  90 2020-03-09  Jakub Jelinek  <jakub@redhat.com>
  91
  92         PR rtl-optimization/94045
  93         * params.opt (-param=max-find-base-term-values=): New option.
  94         * alias.c (find_base_term): Add cut-off for number of visited VALUEs
  95         in a single toplevel find_base_term call.
  96
  97 2020-03-06  Wilco Dijkstra  <wdijkstr@arm.com>
  98
  99         PR target/91598
 100         * config/aarch64/aarch64-builtins.c (TYPES_TERNOPU_LANE): Add define.
 101         * config/aarch64/aarch64-simd.md
 102         (aarch64_vec_<su>mult_lane<Qlane>): Add new insn for widening lane mul.
 103         (aarch64_vec_<su>mlal_lane<Qlane>): Likewise.
 104         * config/aarch64/aarch64-simd-builtins.def: Add intrinsics.
 105         * config/aarch64/arm_neon.h:
 106         (vmlal_lane_s16): Expand using intrinsics rather than inline asm.
 107         (vmlal_lane_u16): Likewise.
 108         (vmlal_lane_s32): Likewise.
 109         (vmlal_lane_u32): Likewise.
 110         (vmlal_laneq_s16): Likewise.
 111         (vmlal_laneq_u16): Likewise.
 112         (vmlal_laneq_s32): Likewise.
 113         (vmlal_laneq_u32): Likewise.
 114         (vmull_lane_s16): Likewise.
 115         (vmull_lane_u16): Likewise.
 116         (vmull_lane_s32): Likewise.
 117         (vmull_lane_u32): Likewise.
 118         (vmull_laneq_s16): Likewise.
 119         (vmull_laneq_u16): Likewise.
 120         (vmull_laneq_s32): Likewise.
 121         (vmull_laneq_u32): Likewise.
 122         * config/aarch64/iterators.md (Vcondtype): New iterator for lane mul.
 123         (Qlane): Likewise.
 124
 125 2020-03-06  Wilco Dijkstra  <wdijkstr@arm.com>
 126
 127         * aarch64/aarch64-simd.md (aarch64_mla_elt<mode>): Correct lane syntax.
 128         (aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
 129         (aarch64_mls_elt<mode>): Likewise.
 130         (aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
 131         (aarch64_fma4_elt<mode>): Likewise.
 132         (aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
 133         (aarch64_fma4_elt_to_64v2df): Likewise.
 134         (aarch64_fnma4_elt<mode>): Likewise.
 135         (aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
 136         (aarch64_fnma4_elt_to_64v2df): Likewise.
 137
 138 2020-03-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 139
 140         * config/aarch64/aarch64-sve2.md (@aarch64_sve_<sve_int_op><mode>:
 141         Specify movprfx attribute.
 142         (@aarch64_sve_<sve_int_op>_lane_<mode>): Likewise.
 143
 144 2020-03-06  David Edelsohn  <dje.gcc@gmail.com>
 145
 146         PR target/94065
 147         * config/rs6000/aix61.h (TARGET_NO_SUM_IN_TOC): Set to 1 for
 148         cmodel=large.
 149         (TARGET_NO_FP_IN_TOC): Same.
 150         * config/rs6000/aix71.h: Same.
 151         * config/rs6000/aix72.h: Same.
 152
 153 2020-03-06  Andrew Pinski  <apinski@marvell.com>
 154             Jeff Law  <law@redhat.com>
 155
 156         PR rtl-optimization/93996
 157         * haifa-sched.c (remove_notes): Be more careful when adding
 158         REG_SAVE_NOTE.
 159
 160 2020-03-06  Delia Burduv  <delia.burduv@arm.com>
 161
 162         * config/arm/arm_neon.h (vld2_bf16): New.
 163         (vld2q_bf16): New.
 164         (vld3_bf16): New.
 165         (vld3q_bf16): New.
 166         (vld4_bf16): New.
 167         (vld4q_bf16): New.
 168         (vld2_dup_bf16): New.
 169         (vld2q_dup_bf16): New.
 170         (vld3_dup_bf16): New.
 171         (vld3q_dup_bf16): New.
 172         (vld4_dup_bf16): New.
 173         (vld4q_dup_bf16): New.
 174         * config/arm/arm_neon_builtins.def
 175         (vld2): Changed to VAR13 and added v4bf, v8bf
 176         (vld2_dup): Changed to VAR8 and added v4bf, v8bf
 177         (vld3): Changed to VAR13 and added v4bf, v8bf
 178         (vld3_dup): Changed to VAR8 and added v4bf, v8bf
 179         (vld4): Changed to VAR13 and added v4bf, v8bf
 180         (vld4_dup): Changed to VAR8 and added v4bf, v8bf
 181         * config/arm/iterators.md (VDXBF2): New iterator.
 182         *config/arm/neon.md (neon_vld2): Use new iterators.
 183         (neon_vld2_dup<mode): Use new iterators.
 184         (neon_vld3<mode>): Likewise.
 185         (neon_vld3qa<mode>): Likewise.
 186         (neon_vld3qb<mode>): Likewise.
 187         (neon_vld3_dup<mode>): Likewise.
 188         (neon_vld4<mode>): Likewise.
 189         (neon_vld4qa<mode>): Likewise.
 190         (neon_vld4qb<mode>): Likewise.
 191         (neon_vld4_dup<mode>): Likewise.
 192         (neon_vld2_dupv8bf): New.
 193         (neon_vld3_dupv8bf): Likewise.
 194         (neon_vld4_dupv8bf): Likewise.
 195
 196 2020-03-06  Delia Burduv  <delia.burduv@arm.com>
 197
 198         * config/arm/arm_neon.h (bfloat16x4x2_t): New typedef.
 199         (bfloat16x8x2_t): New typedef.
 200         (bfloat16x4x3_t): New typedef.
 201         (bfloat16x8x3_t): New typedef.
 202         (bfloat16x4x4_t): New typedef.
 203         (bfloat16x8x4_t): New typedef.
 204         (vst2_bf16): New.
 205         (vst2q_bf16): New.
 206         (vst3_bf16): New.
 207         (vst3q_bf16): New.
 208         (vst4_bf16): New.
 209         (vst4q_bf16): New.
 210         * config/arm/arm-builtins.c (v2bf_UP): Define.
 211         (VAR13): New.
 212         (arm_init_simd_builtin_types): Init Bfloat16x2_t eltype.
 213         * config/arm/arm-modes.def (V2BF): New mode.
 214         * config/arm/arm-simd-builtin-types.def
 215         (Bfloat16x2_t): New entry.
 216         * config/arm/arm_neon_builtins.def
 217         (vst2): Changed to VAR13 and added v4bf, v8bf
 218         (vst3): Changed to VAR13 and added v4bf, v8bf
 219         (vst4): Changed to VAR13 and added v4bf, v8bf
 220         * config/arm/iterators.md (VDXBF): New iterator.
 221         (VQ2BF): New iterator.
 222         *config/arm/neon.md (neon_vst2<mode>): Used new iterators.
 223         (neon_vst2<mode>): Used new iterators.
 224         (neon_vst3<mode>): Used new iterators.
 225         (neon_vst3<mode>): Used new iterators.
 226         (neon_vst3qa<mode>): Used new iterators.
 227         (neon_vst3qb<mode>): Used new iterators.
 228         (neon_vst4<mode>): Used new iterators.
 229         (neon_vst4<mode>): Used new iterators.
 230         (neon_vst4qa<mode>): Used new iterators.
 231         (neon_vst4qb<mode>): Used new iterators.
 232
 233 2020-03-06  Delia Burduv  <delia.burduv@arm.com>
 234
 235         * config/aarch64/aarch64-simd-builtins.def
 236         (bfcvtn): New built-in function.
 237         (bfcvtn_q): New built-in function.
 238         (bfcvtn2): New built-in function.
 239         (bfcvt): New built-in function.
 240         * config/aarch64/aarch64-simd.md
 241         (aarch64_bfcvtn<q><mode>): New pattern.
 242         (aarch64_bfcvtn2v8bf): New pattern.
 243         (aarch64_bfcvtbf): New pattern.
 244         * config/aarch64/arm_bf16.h (float32_t): New typedef.
 245         (vcvth_bf16_f32): New intrinsic.
 246         * config/aarch64/arm_bf16.h (vcvt_bf16_f32): New intrinsic.
 247         (vcvtq_low_bf16_f32): New intrinsic.
 248         (vcvtq_high_bf16_f32): New intrinsic.
 249         * config/aarch64/iterators.md (V4SF_TO_BF): New mode iterator.
 250         (UNSPEC_BFCVTN): New UNSPEC.
 251         (UNSPEC_BFCVTN2): New UNSPEC.
 252         (UNSPEC_BFCVT): New UNSPEC.
 253         * config/arm/types.md (bf_cvt): New type.
 254
 255 2020-03-06  Andreas Krebbel  <krebbel@linux.ibm.com>
 256
 257         * config/s390/s390.md ("tabort"): Get rid of two consecutive
 258         blanks in format string.
 259
 260 2020-03-05  H.J. Lu  <hongjiu.lu@intel.com>
 261
 262         PR target/89229
 263         PR target/89346
 264         * config/i386/i386-protos.h (ix86_output_ssemov): New prototype.
 265         * config/i386/i386.c (ix86_get_ssemov): New function.
 266         (ix86_output_ssemov): Likewise.
 267         * config/i386/sse.md (VMOVE:mov<mode>_internal): Call
 268         ix86_output_ssemov for TYPE_SSEMOV.  Remove TARGET_AVX512VL
 269         check.
 270         (*movxi_internal_avx512f): Call ix86_output_ssemov for TYPE_SSEMOV.
 271         (*movoi_internal_avx): Call ix86_output_ssemov for TYPE_SSEMOV.
 272         Remove ext_sse_reg_operand and TARGET_AVX512VL check.
 273         (*movti_internal): Likewise.
 274         (*movtf_internal): Call ix86_output_ssemov for TYPE_SSEMOV.
 275
 276 2020-03-05  Jeff Law  <law@redhat.com>
 277
 278         PR tree-optimization/91890
 279         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Remove LOC argument.
 280         Use gimple_or_expr_nonartificial_location.
 281         (check_bounds_overlap): Drop LOC argument to maybe_diag_access_bounds.
 282         Use gimple_or_expr_nonartificial_location.
 283         * gimple.c (gimple_or_expr_nonartificial_location): New function.
 284         * gimple.h (gimple_or_expr_nonartificial_location): Declare it.
 285         * tree-ssa-strlen.c (maybe_warn_overflow): Use
 286         gimple_or_expr_nonartificial_location.
 287         (maybe_diag_stxncpy_trunc, handle_builtin_stxncpy_strncat): Likewise.
 288         (maybe_warn_pointless_strcmp): Likewise.
 289
 290 2020-03-05  Jakub Jelinek  <jakub@redhat.com>
 291
 292         PR target/94046
 293         * config/i386/avx2intrin.h (_mm_mask_i32gather_ps): Fix first cast of
 294         SRC and MASK arguments to __m128 from __m128d.
 295         (_mm256_mask_i32gather_ps): Fix first cast of MASK argument to __m256
 296         from __m256d.
 297         (_mm_mask_i64gather_ps): Fix first cast of MASK argument to __m128
 298         from __m128d.
 299         * config/i386/xopintrin.h (_mm_permute2_pd): Fix first cast of C
 300         argument to __m128i from __m128d.
 301         (_mm256_permute2_pd): Fix first cast of C argument to __m256i from
 302         __m256d.
 303         (_mm_permute2_ps): Fix first cast of C argument to __m128i from __m128.
 304         (_mm256_permute2_ps): Fix first cast of C argument to __m256i from
 305         __m256.
 306
 307 2020-03-05  Delia Burduv  <delia.burduv@arm.com>
 308
 309         * config/arm/arm_neon.h (vbfmmlaq_f32): New.
 310         (vbfmlalbq_f32): New.
 311         (vbfmlaltq_f32): New.
 312         (vbfmlalbq_lane_f32): New.
 313         (vbfmlaltq_lane_f32): New.
 314         (vbfmlalbq_laneq_f32): New.
 315         (vbfmlaltq_laneq_f32): New.
 316         * config/arm/arm_neon_builtins.def (vmmla): New.
 317         (vfmab): New.
 318         (vfmat): New.
 319         (vfmab_lane): New.
 320         (vfmat_lane): New.
 321         (vfmab_laneq): New.
 322         (vfmat_laneq): New.
 323         * config/arm/iterators.md (BF_MA): New int iterator.
 324         (bt): New int attribute.
 325         (VQXBF): Copy of VQX with V8BF.
 326         * config/arm/neon.md (neon_vmmlav8bf): New insn.
 327         (neon_vfma<bt>v8bf): New insn.
 328         (neon_vfma<bt>_lanev8bf): New insn.
 329         (neon_vfma<bt>_laneqv8bf): New expand.
 330         (neon_vget_high<mode>): Changed iterator to VQXBF.
 331         * config/arm/unspecs.md (UNSPEC_BFMMLA): New UNSPEC.
 332         (UNSPEC_BFMAB): New UNSPEC.
 333         (UNSPEC_BFMAT): New UNSPEC.
 334
 335 2020-03-05  Jakub Jelinek  <jakub@redhat.com>
 336
 337         PR middle-end/93399
 338         * tree-pretty-print.h (pretty_print_string): Declare.
 339         * tree-pretty-print.c (pretty_print_string): Remove forward
 340         declaration, no longer static.  Change nbytes parameter type
 341         from unsigned to size_t.
 342         * print-rtl.c (print_value) <case CONST_STRING>: Use
 343         pretty_print_string and for shrink way too long strings.
 344
 345 2020-03-05  Richard Biener  <rguenther@suse.de>
 346             Jakub Jelinek  <jakub@redhat.com>
 347
 348         PR tree-optimization/93582
 349         * tree-ssa-sccvn.c (vn_reference_lookup_3): Treat POINTER_PLUS_EXPR
 350         last operand as signed when looking for memset offset.  Formatting
 351         fix.
 352
 353 2020-03-04  Andrew Pinski  <apinski@marvell.com>
 354
 355         PR bootstrap/93962
 356         * value-prof.c (dump_histogram_value): Use std::abs.
 357
 358 2020-03-04  Martin Sebor  <msebor@redhat.com>
 359
 360         PR tree-optimization/93986
 361         * tree-ssa-strlen.c (maybe_warn_overflow): Convert all wide_int
 362         operands to the same precision widest_int to avoid ICEs.
 363
 364 2020-03-04  Bill Schmidt  <wschmidt@linux.ibm.com>
 365
 366         PR target/87560
 367         * rs6000-cpus.def (OTHER_ALTIVEC_MASKS): New #define.
 368         * rs6000.c (rs6000_disable_incompatible_switches): Add table entry
 369         for OPTION_MASK_ALTIVEC.
 370
 371 2020-03-04  Andreas Krebbel  <krebbel@linux.ibm.com>
 372
 373         * config.gcc: Include the glibc-stdint.h header for zTPF.
 374
 375 2020-03-04  Andreas Krebbel  <krebbel@linux.ibm.com>
 376
 377         * config/s390/s390.c (s390_secondary_memory_needed): Disallow
 378         direct FPR-GPR copies.
 379         (s390_register_info_gprtofpr): Disallow GPR content to be saved in
 380         FPRs.
 381
 382 2020-03-04  Andreas Krebbel  <krebbel@linux.ibm.com>
 383
 384         * config/s390/s390.c (s390_emit_prologue): Specify the 2 new
 385         operands to the prologue_tpf expander.
 386         (s390_emit_epilogue): Likewise.
 387         (s390_option_override_internal): Do error checking and setup for
 388         the new options.
 389         * config/s390/tpf.h (TPF_TRACE_PROLOGUE_CHECK)
 390         (TPF_TRACE_EPILOGUE_CHECK, TPF_TRACE_PROLOGUE_TARGET)
 391         (TPF_TRACE_EPILOGUE_TARGET, TPF_TRACE_PROLOGUE_SKIP_TARGET)
 392         (TPF_TRACE_EPILOGUE_SKIP_TARGET): New macro definitions.
 393         * config/s390/tpf.md ("prologue_tpf", "epilogue_tpf"): Add two new
 394         operands for the check flag and the branch target.
 395         * config/s390/tpf.opt ("mtpf-trace-hook-prologue-check")
 396         ("mtpf-trace-hook-prologue-target")
 397         ("mtpf-trace-hook-epilogue-check")
 398         ("mtpf-trace-hook-epilogue-target", "mtpf-trace-skip"): New
 399         options.
 400         * doc/invoke.texi: Document -mtpf-trace-skip option. The other
 401         options are for debugging purposes and will not be documented
 402         here.
 403
 404 2020-03-04  Jakub Jelinek  <jakub@redhat.com>
 405
 406         PR debug/93888
 407         * tree-inline.c (copy_decl_to_var): Copy DECL_BY_REFERENCE flag.
 408
 409         * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Add offseti
 410         argument.  Change pd argument so that it can be modified.  Turn
 411         constant non-CONSTRUCTOR store into non-constant if it is too large.
 412         Adjust offset and size of CONSTRUCTOR or non-constant store to avoid
 413         overflows.
 414         (vn_walk_cb_data::vn_walk_cb_data, vn_reference_lookup_3): Adjust
 415         callers.
 416
 417 2020-02-04  Richard Biener  <rguenther@suse.de>
 418
 419         PR tree-optimization/93964
 420         * graphite-isl-ast-to-gimple.c
 421         (gcc_expression_from_isl_ast_expr_id): Add intermediate
 422         conversion for pointer to integer converts.
 423         * graphite-scop-detection.c (assign_parameter_index_in_region):
 424         Relax assert.
 425
 426 2020-03-04  Martin Liska  <mliska@suse.cz>
 427
 428         PR c/93886
 429         PR c/93887
 430         * doc/invoke.texi: Clarify --help=language and --help=common
 431         interaction.
 432
 433 2020-03-04  Jakub Jelinek  <jakub@redhat.com>
 434
 435         PR tree-optimization/94001
 436         * tree-tailcall.c (process_assignment): Before comparing op1 to
 437         *ass_var, verify *ass_var is non-NULL.
 438
 439 2020-03-04  Kito Cheng  <kito.cheng@sifive.com>
 440
 441         PR target/93995
 442         * config/riscv/riscv.c (riscv_emit_float_compare): Using NE to compare
 443         the result of IOR.
 444
 445 2020-03-03  Dennis Zhang  <dennis.zhang@arm.com>
 446
 447         * config/arm/arm_bf16.h (vcvtah_f32_bf16, vcvth_bf16_f32): New.
 448         * config/arm/arm_neon.h (vcvt_f32_bf16, vcvtq_low_f32_bf16): New.
 449         (vcvtq_high_f32_bf16, vcvt_bf16_f32): New.
 450         (vcvtq_low_bf16_f32, vcvtq_high_bf16_f32): New.
 451         * config/arm/arm_neon_builtins.def (vbfcvt, vbfcvt_high): New entries.
 452         (vbfcvtv4sf, vbfcvtv4sf_high): Likewise.
 453         * config/arm/iterators.md (VBFCVT, VBFCVTM): New mode iterators.
 454         (V_bf_low, V_bf_cvt_m): New mode attributes.
 455         * config/arm/neon.md (neon_vbfcvtv4sf<VBFCVT:mode>): New.
 456         (neon_vbfcvtv4sf_highv8bf, neon_vbfcvtsf): New.
 457         (neon_vbfcvt<VBFCVT:mode>, neon_vbfcvt_highv8bf): New.
 458         (neon_vbfcvtbf_cvtmode<mode>, neon_vbfcvtbf): New
 459         * config/arm/unspecs.md (UNSPEC_BFCVT, UNSPEC_BFCVT_HIG): New.
 460
 461 2020-03-03  Jakub Jelinek  <jakub@redhat.com>
 462
 463         PR tree-optimization/93582
 464         * tree-ssa-sccvn.h (vn_reference_lookup): Add mask argument.
 465         * tree-ssa-sccvn.c (struct vn_walk_cb_data): Add mask and masked_result
 466         members, initialize them in the constructor and if mask is non-NULL,
 467         artificially push_partial_def {} for the portions of the mask that
 468         contain zeros.
 469         (vn_walk_cb_data::finish): If mask is non-NULL, set masked_result to
 470         val and return (void *)-1.  Formatting fix.
 471         (vn_reference_lookup_pieces): Adjust vn_walk_cb_data initialization.
 472         Formatting fix.
 473         (vn_reference_lookup): Add mask argument.  If non-NULL, don't call
 474         fully_constant_vn_reference_p nor vn_reference_lookup_1 and return
 475         data.mask_result.
 476         (visit_nary_op): Handle BIT_AND_EXPR of a memory load and INTEGER_CST
 477         mask.
 478         (visit_stmt): Formatting fix.
 479
 480 2020-03-03  Richard Biener  <rguenther@suse.de>
 481
 482         PR tree-optimization/93946
 483         * alias.h (refs_same_for_tbaa_p): Declare.
 484         * alias.c (refs_same_for_tbaa_p): New function.
 485         * tree-ssa-alias.c (ao_ref_alias_set): For a NULL ref return
 486         zero.
 487         * tree-ssa-scopedtables.h
 488         (avail_exprs_stack::lookup_avail_expr): Add output argument
 489         giving access to the hashtable entry.
 490         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
 491         Likewise.
 492         * tree-ssa-dom.c: Include alias.h.
 493         (dom_opt_dom_walker::optimize_stmt): Validate TBAA state before
 494         removing redundant store.
 495         * tree-ssa-sccvn.h (vn_reference_s::base_set): New member.
 496         (ao_ref_init_from_vn_reference): Adjust prototype.
 497         (vn_reference_lookup_pieces): Likewise.
 498         (vn_reference_insert_pieces): Likewise.
 499         * tree-ssa-sccvn.c: Track base alias set in addition to alias
 500         set everywhere.
 501         (eliminate_dom_walker::eliminate_stmt): Also check base alias
 502         set when removing redundant stores.
 503         (visit_reference_op_store): Likewise.
 504         * dse.c (record_store): Adjust valdity check for redundant
 505         store removal.
 506
 507 2020-03-03  Jakub Jelinek  <jakub@redhat.com>
 508
 509         PR target/26877
 510         * config/s390/s390.h (OPTION_DEFAULT_SPECS): Reorder.
 511
 512         PR rtl-optimization/94002
 513         * explow.c (plus_constant): Punt if cst has VOIDmode and
 514         get_pool_mode is different from mode.
 515
 516 2020-03-03  Claudiu Zissulescu  <claziss@synopsys.com>
 517
 518         * config/arc/arc.c (leigitimate_small_data_address_p): Check if an
 519         address has an offset which fits the scalling constraint for a
 520         load/store operation.
 521         (legitimate_scaled_address_p): Update use
 522         leigitimate_small_data_address_p.
 523         (arc_print_operand): Likewise.
 524         (arc_legitimate_address_p): Likewise.
 525         (legitimate_small_data_address_p): Likewise.
 526
 527 2020-03-03  Claudiu Zissulescu  <claziss@synopsys.com>
 528
 529         * config/arc/arc.md (fmasf4_fpu): Use accl_operand predicate.
 530         (fnmasf4_fpu): Likewise.
 531
 532 2020-03-03  Claudiu Zissulescu  <claziss@synopsys.com>
 533
 534         * config/arc/arc.md (adddi3): Early expand the 64bit operation into
 535         32bit ops.
 536         (subdi3): Likewise.
 537         (adddi3_i): Remove pattern.
 538         (subdi3_i): Likewise.
 539
 540 2020-03-03  Claudiu Zissulescu  <claziss@synopsys.com>
 541
 542         * config/arc/arc.md (eh_return): Add length info.
 543
 544 2020-03-02  David Malcolm  <dmalcolm@redhat.com>
 545
 546         * doc/invoke.texi (-fanalyzer-show-duplicate-count): New.
 547
 548 2020-03-02  David Malcolm  <dmalcolm@redhat.com>
 549
 550         * doc/invoke.texi (Static Analyzer Options): Add
 551         -Wanalyzer-stale-setjmp-buffer to the list of options enabled
 552         by -fanalyzer.
 553
 554 2020-03-02  Uroš Bizjak  <ubizjak@gmail.com>
 555
 556         PR target/93997
 557         * config/i386/i386.md (movstrict<mode>): Allow only
 558         registers with VALID_INT_MODE_P modes.
 559
 560 2020-03-02  Andrew Stubbs  <ams@codesourcery.com>
 561
 562         * config/gcn/gcn-valu.md (dpp_move<mode>): New.
 563         (reduc_insn): Use 'U' and 'B' operand codes.
 564         (reduc_<reduc_op>_scal_<mode>): Allow all types.
 565         (reduc_<reduc_op>_scal_v64di): Delete.
 566         (*<reduc_op>_dpp_shr_<mode>): Allow all 1reg types.
 567         (*plus_carry_dpp_shr_v64si): Change to ...
 568         (*plus_carry_dpp_shr_<mode>): ... this and allow all 1reg int types.
 569         (mov_from_lane63_v64di): Change to ...
 570         (mov_from_lane63_<mode>): ... this, and allow all 64-bit modes.
 571         * config/gcn/gcn.c (gcn_expand_dpp_shr_insn): Increase buffer size.
 572         Support UNSPEC_MOV_DPP_SHR output formats.
 573         (gcn_expand_reduc_scalar): Add "use_moves" reductions.
 574         Add "use_extends" reductions.
 575         (print_operand_address): Add 'I' and 'U' codes.
 576         * config/gcn/gcn.md (unspec): Add UNSPEC_MOV_DPP_SHR.
 577
 578 2020-03-02  Martin Liska  <mliska@suse.cz>
 579
 580         * lto-wrapper.c: Fix typo in comment about
 581         C++ standard version.
 582
 583 2020-03-01  Martin Sebor  <msebor@redhat.com>
 584
 585         PR c++/92721
 586         * calls.c (init_attr_rdwr_indices): Correctly handle attribute.
 587
 588 2020-03-01  Martin Sebor  <msebor@redhat.com>
 589
 590         PR middle-end/93829
 591         * tree-ssa-strlen.c (count_nonzero_bytes): Set the size to that
 592           of a pointer in the outermost ADDR_EXPRs.
 593
 594 2020-02-28  Jeff Law  <law@redhat.com>
 595
 596         * config/v850/v850.h (STATIC_CHAIN_REGNUM): Change to r19.
 597         * config/v850/v850.c (v850_asm_trampoline_template): Update
 598         accordingly.
 599
 600 2020-02-28  Michael Meissner  <meissner@linux.ibm.com>
 601
 602         PR target/93937
 603         * config/rs6000/vsx.md (vsx_extract_<mode>_<VS_scalar>mode_var):
 604         Delete insn.
 605
 606 2020-02-28  Martin Liska  <mliska@suse.cz>
 607
 608         PR other/93965
 609         * configure.ac: Improve detection of ld_date by requiring
 610         either two dashes or none.
 611         * configure: Regenerate.
 612
 613 2020-02-28  Vladimir Makarov  <vmakarov@redhat.com>
 614
 615         PR rtl-optimization/93564
 616         * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
 617         do not honor reg alloc order.
 618
 619 2020-02-27  Joel Hutton  <Joel.Hutton@arm.com>
 620
 621         PR target/87612
 622         * config/aarch64/aarch64.c (aarch64_override_options): Fix
 623         misleading warning string.
 624
 625 2020-02-27  Martin Sebor  <msebor@redhat.com>
 626
 627         * doc/invoke.texi (-Wbuiltin-declaration-mismatch): Fix a typo.
 628
 629 2020-02-27  Michael Meissner  <meissner@linux.ibm.com>
 630
 631         PR target/93932
 632         * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
 633         Split the insn into two parts.  This insn only does variable
 634         extract from a register.
 635         (vsx_extract_<mode>_var_load, VSX_D iterator): New insn, do
 636         variable extract from memory.
 637         (vsx_extract_v4sf_var): Split the insn into two parts.  This insn
 638         only does variable extract from a register.
 639         (vsx_extract_v4sf_var_load): New insn, do variable extract from
 640         memory.
 641         (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Split the insn
 642         into two parts.  This insn only does variable extract from a
 643         register.
 644         (vsx_extract_<mode>_var_load, VSX_EXTRACT_I iterator): New insn,
 645         do variable extract from memory.
 646
 647 2020-02-27  Martin Jambor  <mjambor@suse.cz>
 648             Feng Xue  <fxue@os.amperecomputing.com>
 649
 650         PR ipa/93707
 651         * ipa-cp.c (same_node_or_its_all_contexts_clone_p): Replaced with
 652         new function calls_same_node_or_its_all_contexts_clone_p.
 653         (cgraph_edge_brings_value_p): Use it.
 654         (cgraph_edge_brings_value_p): Likewise.
 655         (self_recursive_pass_through_p): Return false if caller is a clone.
 656         (self_recursive_agg_pass_through_p): Likewise.
 657
 658 2020-02-27  Jan Hubicka  <hubicka@ucw.cz>
 659
 660         PR middle-end/92152
 661         * alias.c (ends_tbaa_access_path_p): Break out from ...
 662         (component_uses_parent_alias_set_from): ... here.
 663         * alias.h (ends_tbaa_access_path_p): Declare.
 664         * tree-ssa-alias.c (access_path_may_continue_p): Break out from ...;
 665         handle trailing arrays past end of tbaa access path.
 666         (aliasing_component_refs_p): ... here; likewise.
 667         (nonoverlapping_refs_since_match_p): Track TBAA segment of the access
 668         path; disambiguate also past end of it.
 669         (nonoverlapping_component_refs_p): Use only TBAA segment of the access
 670         path.
 671
 672 2020-02-27  Mihail Ionescu  <mihail.ionescu@arm.com>
 673
 674         * (__ARM_NUM_LANES, __arm_lane, __arm_lane_q): Move to the
 675         beginning of the file.
 676         (vcreate_bf16, vcombine_bf16): New.
 677         (vdup_n_bf16, vdupq_n_bf16): New.
 678         (vdup_lane_bf16, vdup_laneq_bf16): New.
 679         (vdupq_lane_bf16, vdupq_laneq_bf16): New.
 680         (vduph_lane_bf16, vduph_laneq_bf16): New.
 681         (vset_lane_bf16, vsetq_lane_bf16): New.
 682         (vget_lane_bf16, vgetq_lane_bf16): New.
 683         (vget_high_bf16, vget_low_bf16): New.
 684         (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
 685         (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
 686         (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
 687         (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
 688         (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
 689         (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
 690         (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
 691         (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
 692         (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
 693         (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
 694         (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New.
 695         (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
 696         (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
 697         (vreinterpretq_bf16_p128): New.
 698         (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
 699         (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
 700         (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
 701         (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
 702         (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
 703         (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
 704         (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
 705         (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
 706         (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
 707         (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
 708         (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
 709         (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
 710         (vreinterpretq_p128_bf16): New.
 711         * config/arm/arm_neon_builtins.def (VDX): Add V4BF.
 712         (V_elem): Likewise.
 713         (V_elem_l): Likewise.
 714         (VD_LANE): Likewise.
 715         (VQX) Add V8BF.
 716         (V_DOUBLE): Likewise.
 717         (VDQX): Add V4BF and V8BF.
 718         (V_two_elem, V_three_elem, V_four_elem): Likewise.
 719         (V_reg): Likewise.
 720         (V_HALF): Likewise.
 721         (V_double_vector_mode): Likewise.
 722         (V_cmp_result): Likewise.
 723         (V_uf_sclr): Likewise.
 724         (V_sz_elem): Likewise.
 725         (Is_d_reg): Likewise.
 726         (V_mode_nunits): Likewise.
 727         * config/arm/neon.md (neon_vdup_lane): Enable for BFloat16.
 728
 729 2020-02-27  Andrew Stubbs  <ams@codesourcery.com>
 730
 731         * config/gcn/gcn-valu.md (VEC_SUBDWORD_MODE): New mode iterator.
 732         (<expander><mode>2<exec>): Change modes to VEC_ALL1REG_INT_MODE.
 733         (<expander><mode>3<exec>): Likewise.
 734         (<expander><mode>3): New.
 735         (v<expander><mode>3): New.
 736         (<expander><mode>3): New.
 737         (<expander><mode>3<exec>): Rename to ...
 738         (<expander>v64si3<exec>): ... this, and change modes to V64SI.
 739         * config/gcn/gcn.md (mnemonic): Use '%B' for not.
 740
 741 2020-02-27  Alexandre Oliva <oliva@adacore.com>
 742
 743         * config/vx-common.h (NO_DOLLAR_IN_LABEL, NO_DOT_IN_LABEL): Leave
 744         them alone on vx7.
 745
 746 2020-02-27  Richard Biener  <rguenther@suse.de>
 747
 748         PR tree-optimization/93508
 749         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle _CHK like
 750         non-_CHK variants.  Valueize their length arguments.
 751
 752 2020-02-27  Richard Biener  <rguenther@suse.de>
 753
 754         PR tree-optimization/93953
 755         * tree-vect-slp.c (slp_copy_subtree): Avoid keeping a reference
 756         to the hash-map entry.
 757
 758 2020-02-27  Andrew Stubbs  <ams@codesourcery.com>
 759
 760         * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
 761
 762 2020-02-27  Mark Williams  <mwilliams@fb.com>
 763
 764         * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
 765         * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
 766         -ffile-prefix-map and -fmacro-prefix-map.
 767         * lto-streamer-out.c: Include file-prefix-map.h.
 768         (lto_output_location): Remap the file part of locations.
 769
 770 2020-02-27  Jakub Jelinek  <jakub@redhat.com>
 771
 772         PR c/93949
 773         * gimplify.c (gimplify_init_constructor): Don't promote readonly
 774         DECL_REGISTER variables to TREE_STATIC.
 775
 776         PR tree-optimization/93582
 777         PR tree-optimization/93945
 778         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
 779         non-zero INTEGER_CST second argument and ref->offset or ref->size
 780         not a multiple of BITS_PER_UNIT.
 781
 782 2020-02-27  Jonathan Wakely  <jwakely@redhat.com>
 783
 784         * doc/install.texi (Binaries): Update description of BullFreeware.
 785
 786 2020-02-26  Sandra Loosemore  <sandra@codesourcery.com>
 787
 788         PR c++/90467
 789
 790         * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
 791         C++ Language Options, Warning Options, and Static Analyzer
 792         Options lists.  Document negative form of options enabled by
 793         default.  Move some things around to more accurately sort
 794         warnings by category.
 795         (C++ Dialect Options, Warning Options, Static Analyzer
 796         Options): Document negative form of options when enabled by
 797         default.  Move some things around to more accurately sort
 798         warnings by category.  Add some missing index entries.
 799         Light copy-editing.
 800
 801 2020-02-26  Carl Love  <cel@us.ibm.com>
 802
 803         PR target/91276
 804         * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
 805         ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
 806         for the vector unsigned short arguments.  It is also listed as the
 807         name of the built-in for arguments vector unsigned short,
 808         vector unsigned int and vector unsigned long long built-ins.  The
 809         name of the builtins for these arguments should be:
 810         __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
 811         __builtin_crypto_vpmsumd respectively.
 812
 813 2020-02-26  Richard Biener  <rguenther@suse.de>
 814
 815         * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
 816         and load permutation.
 817
 818 2020-02-26  Richard Sandiford  <richard.sandiford@arm.com>
 819
 820         PR middle-end/93843
 821         * optabs-tree.c (supportable_convert_operation): Reject types with
 822         scalar modes.
 823
 824 2020-02-26  David Malcolm  <dmalcolm@redhat.com>
 825
 826         * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
 827
 828 2020-02-26  Jakub Jelinek  <jakub@redhat.com>
 829
 830         PR tree-optimization/93820
 831         * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
 832         argument to ALL_INTEGER_CST_P boolean.
 833         (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
 834         (imm_store_chain_info::coalesce_immediate_stores): Likewise.  Handle
 835         adjacent INTEGER_CST store into merged_store->only_constants like
 836         overlapping one.
 837
 838 2020-02-25  Jakub Jelinek  <jakub@redhat.com>
 839
 840         PR other/93912
 841         * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
 842         -> probability.
 843         * cfghooks.c (verify_flow_info): Likewise.
 844         * predict.c (combine_predictions_for_bb): Likewise.
 845         * bb-reorder.c (connect_better_edge_p): Likewise.  Fix comment typo,
 846         sucessor -> successor.
 847         (find_traces_1_round): Fix comment typo, destinarion -> destination.
 848         * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
 849         successors.
 850         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
 851         message typo, sucessors -> successors.
 852
 853 2020-02-25  Martin Sebor  <msebor@redhat.com>
 854
 855         * doc/extend.texi (attribute access): Correct an example.
 856
 857 2020-02-25  Mihail Ionescu  <mihail.ionescu@arm.com>
 858
 859         * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
 860         Add simd_bf.
 861         (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
 862         (VAR15, VAR16): New.
 863         * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
 864         (VD): Enable for V4BF.
 865         (VDC): Likewise.
 866         (VQ): Enable for V8BF.
 867         (VQ2): Likewise.
 868         (VQ_NO2E): Likewise.
 869         (VDBL, Vdbl): Add V4BF.
 870         (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
 871         * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
 872         (bfloat16x8x2_t): Likewise.
 873         (bfloat16x4x3_t): Likewise.
 874         (bfloat16x8x3_t): Likewise.
 875         (bfloat16x4x4_t): Likewise.
 876         (bfloat16x8x4_t): Likewise.
 877         (vcombine_bf16): New.
 878         (vld1_bf16, vld1_bf16_x2): New.
 879         (vld1_bf16_x3, vld1_bf16_x4): New.
 880         (vld1q_bf16, vld1q_bf16_x2): New.
 881         (vld1q_bf16_x3, vld1q_bf16_x4): New.
 882         (vld1_lane_bf16): New.
 883         (vld1q_lane_bf16): New.
 884         (vld1_dup_bf16): New.
 885         (vld1q_dup_bf16): New.
 886         (vld2_bf16): New.
 887         (vld2q_bf16): New.
 888         (vld2_dup_bf16): New.
 889         (vld2q_dup_bf16): New.
 890         (vld3_bf16): New.
 891         (vld3q_bf16): New.
 892         (vld3_dup_bf16): New.
 893         (vld3q_dup_bf16): New.
 894         (vld4_bf16): New.
 895         (vld4q_bf16): New.
 896         (vld4_dup_bf16): New.
 897         (vld4q_dup_bf16): New.
 898         (vst1_bf16, vst1_bf16_x2): New.
 899         (vst1_bf16_x3, vst1_bf16_x4): New.
 900         (vst1q_bf16, vst1q_bf16_x2): New.
 901         (vst1q_bf16_x3, vst1q_bf16_x4): New.
 902         (vst1_lane_bf16): New.
 903         (vst1q_lane_bf16): New.
 904         (vst2_bf16): New.
 905         (vst2q_bf16): New.
 906         (vst3_bf16): New.
 907         (vst3q_bf16): New.
 908         (vst4_bf16): New.
 909         (vst4q_bf16): New.
 910
 911 2020-02-25  Mihail Ionescu  <mihail.ionescu@arm.com>
 912
 913         * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
 914         (VALL_F16): Likewise.
 915         (VALLDI_F16): Likewise.
 916         (Vtype): Likewise.
 917         (Vetype): Likewise.
 918         (vswap_width_name): Likewise.
 919         (VSWAP_WIDTH): Likewise.
 920         (Vel): Likewise.
 921         (VEL): Likewise.
 922         (q): Likewise.
 923         * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
 924         (vget_lane_bf16, vgetq_lane_bf16): New.
 925         (vcreate_bf16): New.
 926         (vdup_n_bf16, vdupq_n_bf16): New.
 927         (vdup_lane_bf16, vdup_laneq_bf16): New.
 928         (vdupq_lane_bf16, vdupq_laneq_bf16): New.
 929         (vduph_lane_bf16, vduph_laneq_bf16): New.
 930         (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
 931         (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
 932         (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
 933         (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
 934         (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
 935         (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
 936         (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
 937         (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
 938         (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
 939         (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
 940         (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
 941         (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
 942         (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
 943         (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
 944         (vreinterpretq_bf16_p128): New.
 945         (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
 946         (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
 947         (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
 948         (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
 949         (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
 950         (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
 951         (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
 952         (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
 953         (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
 954         (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
 955         (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
 956         (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
 957         (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
 958         (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
 959         (vreinterpretq_p128_bf16): New.
 960
 961 2020-02-25  Dennis Zhang  <dennis.zhang@arm.com>
 962
 963         * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
 964         (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
 965         (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
 966         * config/arm/arm_neon_builtins.def (vbfdot): New entry.
 967         (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
 968         * config/arm/iterators.md (VSF2BF): New attribute.
 969         * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
 970         (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
 971         (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
 972
 973 2020-02-25  Christophe Lyon  <christophe.lyon@linaro.org>
 974
 975         * config/arm/arm.md (required_for_purecode): New attribute.
 976         (enabled): Handle required_for_purecode.
 977         * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
 978         work with -mpure-code.
 979
 980 2020-02-25  Jakub Jelinek  <jakub@redhat.com>
 981
 982         PR rtl-optimization/93908
 983         * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
 984         with mask.
 985
 986 2019-02-25  Eric Botcazou  <ebotcazou@adacore.com>
 987
 988         * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
 989
 990 2020-02-25  Roman Zhuykov  <zhroma@ispras.ru>
 991
 992         * doc/install.texi (--enable-checking): Adjust wording.
 993
 994 2020-02-25  Richard Biener  <rguenther@suse.de>
 995
 996         PR tree-optimization/93868
 997         * tree-vect-slp.c (slp_copy_subtree): New function.
 998         (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
 999         re-arranging stmts in it.
1000
1001 2020-02-25  Jakub Jelinek  <jakub@redhat.com>
1002
1003         PR middle-end/93874
1004         * passes.c (pass_manager::dump_passes): Create a cgraph node for the
1005         dummy function and remove it at the end.
1006
1007         PR translation/93864
1008         * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
1009         paramter -> parameter.
1010         * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
1011         * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
1012
1013 2020-02-24  Roman Zhuykov  <zhroma@ispras.ru>
1014
1015         * doc/install.texi (--enable-checking): Properly document current
1016         behavior.
1017         (--enable-stage1-checking): Minor clarification about bootstrap.
1018
1019 2020-02-24  David Malcolm  <dmalcolm@redhat.com>
1020
1021         PR analyzer/93032
1022         * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
1023         -fanalyzer-checker=taint is also required.
1024         (-fanalyzer-checker=): Note that providing this option enables the
1025         given checker, and doing so may be required for checkers that are
1026         disabled by default.
1027
1028 2020-02-24  David Malcolm  <dmalcolm@redhat.com>
1029
1030         * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
1031         significant control flow events; add a "3" which shows all
1032         control flow events; the old "3" becomes "4".
1033
1034 2020-02-24  Jakub Jelinek  <jakub@redhat.com>
1035
1036         PR tree-optimization/93582
1037         * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
1038         pd.offset and pd.size to be counted in bits rather than bytes, add
1039         support for maxsizei that is not a multiple of BITS_PER_UNIT and
1040         handle bitfield stores and loads.
1041         (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
1042         uncomparable quantities - bytes vs. bits.  Allow push_partial_def
1043         on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
1044         pd.offset/pd.size to be counted in bits rather than bytes.
1045         Formatting fix.  Rename shadowed len variable to buflen.
1046
1047 2020-02-24  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
1048             Kugan Vivekandarajah  <kugan.vivekanandarajah@linaro.org>
1049
1050         PR driver/47785
1051         * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
1052         (driver::main): Call putenv_COLLECT_AS_OPTIONS.
1053         * opts-common.c (parse_options_from_collect_gcc_options): New function.
1054         (prepend_xassembler_to_collect_as_options): Likewise.
1055         * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
1056         (prepend_xassembler_to_collect_as_options): Likewise.
1057         * lto-opts.c (lto_write_options): Stream assembler options
1058         in COLLECT_AS_OPTIONS.
1059         * lto-wrapper.c (xassembler_options_error): New static variable.
1060         (get_options_from_collect_gcc_options): Move parsing options code to
1061         parse_options_from_collect_gcc_options and call it.
1062         (merge_and_complain): Validate -Xassembler options.
1063         (append_compiler_options): Handle OPT_Xassembler.
1064         (run_gcc): Append command line -Xassembler options to
1065         collect_gcc_options.
1066         * doc/invoke.texi: Add documentation about using Xassembler
1067         options with LTO.
1068
1069 2020-02-24  Kito Cheng  <kito.cheng@sifive.com>
1070
1071         * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
1072         for LTGT.
1073         (riscv_rtx_costs): Update cost model for LTGT.
1074
1075 2020-02-23  Vladimir Makarov  <vmakarov@redhat.com>
1076
1077         PR rtl-optimization/93564
1078         * ira-color.c (struct update_cost_queue_elem): New member start.
1079         (queue_update_cost, get_next_update_cost): Add new arg start.
1080         (allocnos_conflict_p): New function.
1081         (update_costs_from_allocno): Add new arg conflict_cost_update_p.
1082         Add checking conflicts with allocnos_conflict_p.
1083         (update_costs_from_prefs, restore_costs_from_copies): Adjust
1084         update_costs_from_allocno calls.
1085         (update_conflict_hard_regno_costs): Add checking conflicts with
1086         allocnos_conflict_p.  Adjust calls of queue_update_cost and
1087         get_next_update_cost.
1088         (assign_hard_reg): Adjust calls of queue_update_cost.  Add
1089         debugging print.
1090         (bucket_allocno_compare_func): Restore previous version.
1091
1092 2020-02-21  John David Anglin  <danglin@gcc.gnu.org>
1093
1094         * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
1095         double-word size when handling aggregate return values.
1096         * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
1097         that homogeneous SFmode and DFmode aggregates are passed and returned
1098         in general registers.
1099
1100 2020-02-21  Jakub Jelinek  <jakub@redhat.com>
1101
1102         PR translation/93759
1103         * opts.c (print_filtered_help): Translate help before appending
1104         messages to it rather than after that.
1105
1106 2020-02-19  Richard Sandiford  <richard.sandiford@arm.com>
1107
1108         PR rtl-optimization/PR92989
1109         * lra-lives.c (process_bb_lives): Restore the original order
1110         of the bb liveness update.  Call make_hard_regno_dead for each
1111         register clobbered at the start of an EH receiver.
1112
1113 2020-02-18  Feng Xue  <fxue@os.amperecomputing.com>
1114
1115         PR ipa/93763
1116         * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
1117         self-recursively generated.
1118
1119 2020-02-21  Iain Sandoe  <iain@sandoe.co.uk>
1120
1121         PR target/93860
1122         * config/darwin-c.c (pop_field_alignment): Adjust quoting of
1123         error string.
1124
1125 2020-02-21  Mihail Ionescu  <mihail.ionescu@arm.com>
1126
1127         * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
1128         Document new target supports option.
1129
1130 2020-02-21  Dennis Zhang  <dennis.zhang@arm.com>
1131
1132         * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
1133         * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
1134         * config/arm/iterators.md (MATMUL): New iterator.
1135         (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
1136         (mmla_sfx): New attribute.
1137         * config/arm/neon.md (neon_<sup>mmlav16qi): New.
1138         * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
1139         (UNSPEC_MATMUL_US): New.
1140
1141 2020-02-21  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
1142
1143         * config/arm/arm.md: Prevent scalar shifts from being used when big
1144         endian is enabled.
1145
1146 2020-02-21  Jan Hubicka  <hubicka@ucw.cz>
1147             Richard Biener  <rguenther@suse.de>
1148
1149         PR tree-optimization/93586
1150         * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
1151         after mismatched array refs; do not sure type size information to
1152         recover from unmatched referneces with !flag_strict_aliasing_p.
1153
1154 2020-02-21  Andrew Stubbs  <ams@codesourcery.com>
1155
1156         * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
1157         (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
1158         (scatter_store<mode>): Rename to ...
1159         (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
1160         (scatter<mode>_exec): Delete. Move contents ...
1161         (mask_scatter_store<mode>): ... here, and rename that to ...
1162         (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
1163         Remove mode conversion.
1164         (mask_gather_load<mode>): Rename to ...
1165         (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
1166         Remove mode conversion.
1167         * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
1168
1169 2020-02-21  Martin Jambor  <mjambor@suse.cz>
1170
1171         PR tree-optimization/93845
1172         * tree-sra.c (verify_sra_access_forest): Only test access size of
1173         scalar types.
1174
1175 2020-02-21  Andrew Stubbs  <ams@codesourcery.com>
1176
1177         * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
1178         * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
1179         (addv64di3_exec): Likewise.
1180         (subv64di3): Likewise.
1181         (subv64di3_exec): Likewise.
1182         (addv64di3_zext): Likewise.
1183         (addv64di3_zext_exec): Likewise.
1184         (addv64di3_zext_dup): Likewise.
1185         (addv64di3_zext_dup_exec): Likewise.
1186         (addv64di3_zext_dup2): Likewise.
1187         (addv64di3_zext_dup2_exec): Likewise.
1188         (addv64di3_sext_dup2): Likewise.
1189         (addv64di3_sext_dup2_exec): Likewise.
1190         (<expander>v64di3): Likewise.
1191         (<expander>v64di3_exec): Likewise.
1192         (*<reduc_op>_dpp_shr_v64di): Likewise.
1193         (*plus_carry_dpp_shr_v64di): Likewise.
1194         * config/gcn/gcn.md (adddi3): Likewise.
1195         (addptrdi3): Likewise.
1196         (<expander>di3): Likewise.
1197
1198 2020-02-21  Andrew Stubbs  <ams@codesourcery.com>
1199
1200         * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
1201
1202 2020-02-21  Richard Sandiford  <richard.sandiford@arm.com>
1203
1204         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
1205         support.  Use aarch64_emit_mult instead of emitting multiplication
1206         instructions directly.
1207         * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
1208         (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
1209
1210 2020-02-21  Richard Sandiford  <richard.sandiford@arm.com>
1211
1212         * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
1213         (aarch64_emit_approx_div): Add SVE support.  Use aarch64_emit_mult
1214         instead of emitting multiplication instructions directly.
1215         * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
1216         * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
1217         (@aarch64_frecps<mode>): New expanders.
1218
1219 2020-02-21  Richard Sandiford  <richard.sandiford@arm.com>
1220
1221         * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
1222         on and produce uint64_ts rather than ints.
1223         (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
1224         (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
1225
1226 2020-02-21  Richard Sandiford  <richard.sandiford@arm.com>
1227
1228         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
1229         an unused xmsk register when handling approximate rsqrt.
1230
1231 2020-02-21  Richard Sandiford  <richard.sandiford@arm.com>
1232
1233         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
1234         flag_finite_math_only condition.
1235
1236 2020-02-20  Uroš Bizjak  <ubizjak@gmail.com>
1237
1238         PR target/93828
1239         * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
1240         to destination operand for shufps alternative.
1241         (*vec_extractv2si_1): Ditto.
1242
1243 2020-02-20  Peter Bergner  <bergner@linux.ibm.com>
1244
1245         PR target/93658
1246         * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
1247         vector modes.
1248
1249 2020-02-20  Martin Liska  <mliska@suse.cz>
1250
1251         PR translation/93831
1252         * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
1253
1254 2020-02-20  Martin Liska  <mliska@suse.cz>
1255
1256         PR translation/93830
1257         * common/config/avr/avr-common.c: Remote trailing "|".
1258
1259 2020-02-19  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1260
1261         * collect2.c (maybe_run_lto_and_relink): Fix typo in
1262         comment.
1263
1264 2020-02-19  Richard Sandiford  <richard.sandiford@arm.com>
1265
1266         PR tree-optimization/93767
1267         * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
1268         access-size bias from the offset calculations for negative strides.
1269
1270 2020-02-19  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1271
1272         * collect2.c (c_file, o_file): Make const again.
1273         (ldout,lderrout, dump_ld_file): Remove.
1274         (tool_cleanup): Avoid calling not signal-safe functions.
1275         (maybe_run_lto_and_relink): Avoid possible signal handler
1276         access to unintialzed memory (lto_o_files).
1277         (main): Avoid leaking temp files in $TMPDIR.
1278         Initialize c_file/o_file with concat, which avoids exposing
1279         uninitialized memory to signal handler, which calls unlink(!).
1280         Avoid calling maybe_unlink when the main function returns,
1281         since the atexit handler is already doing this.
1282         * collect2.h (dump_ld_file, ldout, lderrout): Remove.
1283
1284 2020-02-19  Martin Jambor  <mjambor@suse.cz>
1285
1286         PR tree-optimization/93776
1287         * tree-sra.c (create_access): Do not create zero size accesses.
1288         (get_access_for_expr): Do not search for zero sized accesses.
1289
1290 2020-02-19  Martin Jambor  <mjambor@suse.cz>
1291
1292         PR tree-optimization/93667
1293         * tree-sra.c (scalarizable_type_p): Return false if record fields
1294         do not follow wach other.
1295
1296 2020-01-21  Kito Cheng  <kito.cheng@sifive.com>
1297
1298         * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
1299         rather than fmv.x.s/fmv.s.x.
1300
1301 2020-02-18  James Greenhalgh  <james.greenhalgh@arm.com>
1302
1303         * config/aarch64/aarch64-simd-builtins.def
1304         (intrinsic_vec_smult_lo_): New.
1305         (intrinsic_vec_umult_lo_): Likewise.
1306         (vec_widen_smult_hi_): Likewise.
1307         (vec_widen_umult_hi_): Likewise.
1308         * config/aarch64/aarch64-simd.md
1309         (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
1310         * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
1311         (vmull_high_s16): Likewise.
1312         (vmull_high_s32): Likewise.
1313         (vmull_high_u8): Likewise.
1314         (vmull_high_u16): Likewise.
1315         (vmull_high_u32): Likewise.
1316         (vmull_s8): Likewise.
1317         (vmull_s16): Likewise.
1318         (vmull_s32): Likewise.
1319         (vmull_u8): Likewise.
1320         (vmull_u16): Likewise.
1321         (vmull_u32): Likewise.
1322
1323 2020-02-18  Martin Liska  <mliska@suse.cz>
1324
1325         * value-prof.c (stream_out_histogram_value): Restore LTO PGO
1326         bootstrap by missing removal of invalid sanity check.
1327
1328 2020-02-18  Martin Liska  <mliska@suse.cz>
1329
1330         PR ipa/92518
1331         * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
1332         Always compare LHS of gimple_assign.
1333
1334 2020-02-18  Martin Liska  <mliska@suse.cz>
1335
1336         PR ipa/93583
1337         * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
1338         and return type of functions.
1339         * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
1340         Drop MALLOC attribute for void functions.
1341         * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
1342         malloc_state for a new VOID clone.
1343
1344 2020-02-18  Martin Liska  <mliska@suse.cz>
1345
1346         PR ipa/92924
1347         * common.opt: Add -fprofile-reproducibility.
1348         * doc/invoke.texi: Document it.
1349         * value-prof.c (dump_histogram_value):
1350         Document and support behavior for counters[0]
1351         being a negative value.
1352         (get_nth_most_common_value): Handle negative
1353         counters[0] in respect to flag_profile_reproducible.
1354
1355 2020-02-18  Jakub Jelinek  <jakub@redhat.com>
1356
1357         PR ipa/93797
1358         * cgraph.c (verify_speculative_call): Use speculative_id instead of
1359         speculative_uid in messages.  Remove trailing whitespace from error
1360         message.  Use num_speculative_call_targets instead of
1361         num_speculative_targets in a message.
1362         (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
1363         edge messages and stmt instead of cal_stmt in reference message.
1364
1365         PR tree-optimization/93780
1366         * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
1367         before calling build_vector_type.
1368         (execute_update_addresses_taken): Likewise.
1369
1370         PR driver/93796
1371         * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
1372         typo, functoin -> function.
1373         * tree.c (free_lang_data_in_decl): Fix comment typo,
1374         functoin -> function.
1375         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1376
1377 2020-02-17  David Malcolm  <dmalcolm@redhat.com>
1378
1379         * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
1380         won't be printed.
1381         (print_option_information): Don't call get_option_url if URLs
1382         won't be printed.
1383
1384 2020-02-17  Alexandre Oliva  <oliva@adacore.com>
1385
1386         * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
1387         handling of register_common-less targets.
1388
1389 2020-02-17  Martin Liska  <mliska@suse.cz>
1390
1391         PR ipa/93760
1392         * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
1393
1394 2020-02-17  Martin Liska  <mliska@suse.cz>
1395
1396         PR translation/93755
1397         * config/rs6000/rs6000.c (rs6000_option_override_internal):
1398         Fix double quotes.
1399
1400 2020-02-17  Martin Liska  <mliska@suse.cz>
1401
1402         PR other/93756
1403         * config/rx/elf.opt: Fix typo.
1404
1405 2020-02-17  Richard Biener  <rguenther@suse.de>
1406
1407         PR c/86134
1408         * opts-global.c (print_ignored_options): Use inform and
1409         amend message.
1410
1411 2020-02-17  Jiufu Guo  <guojiufu@linux.ibm.com>
1412
1413         PR target/93047
1414         * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
1415
1416 2020-02-16  Uroš Bizjak  <ubizjak@gmail.com>
1417
1418         PR target/93743
1419         * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
1420         (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
1421
1422 2020-02-15  Jason Merrill  <jason@redhat.com>
1423
1424         * doc/invoke.texi (C Dialect Options): Add -std=c++20.
1425
1426 2020-02-15  Jakub Jelinek  <jakub@redhat.com>
1427
1428         PR tree-optimization/93744
1429         * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
1430         A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
1431         A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
1432         sure @2 in the first and @1 in the other patterns has no side-effects.
1433
1434 2020-02-15  David Malcolm  <dmalcolm@redhat.com>
1435             Bernd Edlinger  <bernd.edlinger@hotmail.de>
1436
1437         PR 87488
1438         PR other/93168
1439         * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
1440         * configure.ac (--with-diagnostics-urls): New configuration
1441         option, based on --with-diagnostics-color.
1442         (DIAGNOSTICS_URLS_DEFAULT): New define.
1443         * config.h: Regenerate.
1444         * configure: Regenerate.
1445         * diagnostic.c (diagnostic_urls_init): Handle -1 for
1446         DIAGNOSTICS_URLS_DEFAULT from configure-time
1447         --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
1448         and TERM_URLS environment variable.
1449         * diagnostic-url.h (diagnostic_url_format): New enum type.
1450         (diagnostic_urls_enabled_p): rename to...
1451         (determine_url_format): ... this, and change return type.
1452         * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
1453         (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
1454         the linux console, and mingw.
1455         (diagnostic_urls_enabled_p): rename to...
1456         (determine_url_format): ... this, and adjust.
1457         * pretty-print.h (pretty_printer::show_urls): rename to...
1458         (pretty_printer::url_format): ... this, and change to enum.
1459         * pretty-print.c (pretty_printer::pretty_printer,
1460         pp_begin_url, pp_end_url, test_urls): Adjust.
1461         * doc/install.texi (--with-diagnostics-urls): Document the new
1462         configuration option.
1463         (--with-diagnostics-color): Document the existing interaction
1464         with GCC_COLORS better.
1465         * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
1466         vindex reference.  Update description of defaults based on the above.
1467         (-fdiagnostics-color): Update description of how -fdiagnostics-color
1468         interacts with GCC_COLORS.
1469
1470 2020-02-14  Eric Botcazou  <ebotcazou@adacore.com>
1471
1472         PR target/93704
1473         * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
1474         conjunction with TARGET_GNU_TLS in early return.
1475
1476 2020-02-14  Alexander Monakov  <amonakov@ispras.ru>
1477
1478         * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
1479         the mode is not wider than UNITS_PER_WORD.
1480
1481 2020-02-14  Martin Jambor  <mjambor@suse.cz>
1482
1483         PR tree-optimization/93516
1484         * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
1485         access of the same type as the parent.
1486         (propagate_subaccesses_from_lhs): Likewise.
1487
1488 2020-02-14 Hongtao Liu  <hongtao.liu@intel.com>
1489
1490         PR target/93724
1491         * config/i386/avx512vbmi2intrin.h
1492         (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
1493         _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
1494         _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
1495         _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
1496         _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
1497         _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
1498         _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
1499         _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
1500         _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
1501         of lacking a closing parenthesis.
1502         * config/i386/avx512vbmi2vlintrin.h
1503         (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
1504         _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
1505         _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
1506         _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
1507         _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
1508         _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
1509         _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
1510         _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
1511         _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
1512         _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
1513         _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
1514         _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
1515         _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
1516         _m_maskz_shrdi_epi64, _mm_shldi_epi16,
1517         _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
1518         _mm_shldi_epi32, _mm_mask_shldi_epi32,
1519         _mm_maskz_shldi_epi32, _mm_shldi_epi64,
1520         _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
1521
1522 2020-02-13  H.J. Lu  <hongjiu.lu@intel.com>
1523
1524         PR target/93656
1525         * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
1526         the target function entry.
1527
1528 2020-02-13  Claudiu Zissulescu  <claziss@synopsys.com>
1529
1530         * common/config/arc/arc-common.c (arc_option_optimization_table):
1531         Disable if-conversion step when optimized for size.
1532
1533 2020-02-13  Claudiu Zissulescu  <claziss@synopsys.com>
1534
1535         * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
1536         R12-R15 are always in ARCOMPACT16_REGS register class.
1537         * config/arc/arc.opt (mq-class): Deprecate.
1538         * config/arc/constraint.md ("q"): Remove dependency on mq-class
1539         option.
1540         * doc/invoke.texi (mq-class): Update text.
1541         * common/config/arc/arc-common.c (arc_option_optimization_table):
1542         Update list.
1543
1544 2020-02-13  Claudiu Zissulescu  <claziss@synopsys.com>
1545
1546         * config/arc/arc.c (arc_insn_cost): New function.
1547         (TARGET_INSN_COST): Define.
1548         * config/arc/arc.md (cost): New attribute.
1549         (add_n): Use arc_nonmemory_operand.
1550         (ashlsi3_insn): Likewise, also update constraints.
1551         (ashrsi3_insn): Likewise.
1552         (rotrsi3): Likewise.
1553         (add_shift): Likewise.
1554         * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
1555
1556 2020-02-13  Claudiu Zissulescu  <claziss@synopsys.com>
1557
1558         * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
1559         registers.
1560         (umulsidi_600): Likewise.
1561
1562 2020-02-13  Jakub Jelinek  <jakub@redhat.com>
1563
1564         PR target/93696
1565         * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
1566         _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
1567         _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
1568         _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
1569         pass __A to the builtin followed by __W instead of __A followed by
1570         __B.
1571         * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
1572         _mm512_mask_popcnt_epi64): Likewise.
1573         * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
1574         _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
1575         _mm256_mask_popcnt_epi64): Likewise.
1576
1577         PR tree-optimization/93582
1578         * fold-const.h (shift_bytes_in_array_left,
1579         shift_bytes_in_array_right): Declare.
1580         * fold-const.c (shift_bytes_in_array_left,
1581         shift_bytes_in_array_right): New function, moved from
1582         gimple-ssa-store-merging.c, no longer static.
1583         * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
1584         to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
1585         (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
1586         (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
1587         shift_bytes_in_array.
1588         (verify_shift_bytes_in_array): Rename to ...
1589         (verify_shift_bytes_in_array_left): ... this.  Use
1590         shift_bytes_in_array_left instead of shift_bytes_in_array.
1591         (store_merging_c_tests): Call verify_shift_bytes_in_array_left
1592         instead of verify_shift_bytes_in_array.
1593         * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
1594         / native_interpret_expr where the store covers all needed bits,
1595         punt on PDP-endian, otherwise allow all involved offsets and sizes
1596         not to be byte-aligned.
1597
1598         PR target/93673
1599         * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
1600         use const_0_to_255_operand predicate instead of immediate_operand.
1601         (avx512dq_fpclass<mode><mask_scalar_merge_name>,
1602         avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
1603         vgf2p8affineinvqb_<mode><mask_name>,
1604         vgf2p8affineqb_<mode><mask_name>): Drop mode from
1605         const_0_to_255_operand predicated operands.
1606
1607 2020-02-12  Jeff Law  <law@redhat.com>
1608
1609         * config/h8300/h8300.md (comparison shortening peepholes): Use
1610         a mode iterator to merge the HImode and SImode peepholes.
1611
1612 2020-02-12  Jakub Jelinek  <jakub@redhat.com>
1613
1614         PR middle-end/93663
1615         * real.c (is_even): Make static.  Function comment fix.
1616         (is_halfway_below): Make static, don't assert R is not inf/nan,
1617         instead return false for those.  Small formatting fixes.
1618
1619 2020-02-12  Martin Sebor  <msebor@redhat.com>
1620
1621         PR middle-end/93646
1622         * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
1623         (handle_builtin_stxncpy_strncat): ...to this.  Change first argument.
1624         Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
1625         (strlen_check_and_optimize_call): Adjust callee name.
1626
1627 2020-02-12  Jeff Law  <law@redhat.com>
1628
1629         * config/h8300/h8300.md (comparison shortening peepholes): Drop
1630         (and (xor)) variant.  Combine other two into single peephole.
1631
1632 2020-02-12  Wilco Dijkstra  <wdijkstr@arm.com>
1633
1634         PR rtl-optimization/93565
1635         * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
1636
1637 2020-02-12  Wilco Dijkstra  <wdijkstr@arm.com>
1638
1639         * config/aarch64/aarch64-simd.md
1640         (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
1641         * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
1642         generating separate ADDV and zero_extend patterns.
1643         * config/aarch64/iterators.md (VDQV_E): New iterator.
1644
1645 2020-02-12  Jeff Law  <law@redhat.com>
1646
1647         * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
1648         expanders, splits, etc.
1649         (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
1650         (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
1651         (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
1652         * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
1653         (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
1654         * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
1655         function prototype.
1656         (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
1657
1658 2020-02-12  Jakub Jelinek  <jakub@redhat.com>
1659
1660         PR target/93670
1661         * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
1662         (avx512vl_vextractf128<mode>): Use it instead of VI48F_256.  Remove
1663         TARGET_AVX512DQ from condition.
1664         (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
1665         instead of <mask_mode512bit_condition> in condition.  If
1666         TARGET_AVX512DQ is false, emit vextract*64x4 instead of
1667         vextract*32x8.
1668         (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
1669         from condition.
1670
1671 2020-02-12  Kewen Lin  <linkw@gcc.gnu.org>
1672
1673         PR target/91052
1674         * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
1675
1676 2020-02-12  Segher Boessenkool  <segher@kernel.crashing.org>
1677
1678         * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
1679         where strlen is more legible.
1680         (rs6000_builtin_vectorized_libmass): Ditto.
1681         (rs6000_print_options_internal): Ditto.
1682
1683 2020-02-11  Martin Sebor  <msebor@redhat.com>
1684
1685         PR tree-optimization/93683
1686         * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
1687
1688 2020-02-11  Michael Meissner  <meissner@linux.ibm.com>
1689
1690         * config/rs6000/predicates.md (cint34_operand): Rename the
1691         -mprefixed-addr option to be -mprefixed.
1692         * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
1693         the -mprefixed-addr option to be -mprefixed.
1694         (OTHER_FUTURE_MASKS): Likewise.
1695         (POWERPC_MASKS): Likewise.
1696         * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
1697         the -mprefixed-addr option to be -mprefixed.  Change error
1698         messages to refer to -mprefixed.
1699         (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
1700         -mprefixed.
1701         (rs6000_legitimate_offset_address_p): Likewise.
1702         (rs6000_mode_dependent_address): Likewise.
1703         (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
1704         "-mprefixed" for target attributes and pragmas.
1705         (address_to_insn_form): Rename the -mprefixed-addr option to be
1706         -mprefixed.
1707         (rs6000_adjust_insn_length): Likewise.
1708         * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
1709         -mprefixed-addr option to be -mprefixed.
1710         (ASM_OUTPUT_OPCODE): Likewise.
1711         * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
1712         -mprefixed-addr option to be -mprefixed.
1713         * config/rs6000/rs6000.opt (-mprefixed): Rename the
1714         -mprefixed-addr option to be prefixed.  Change the option from
1715         being undocumented to being documented.
1716         * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
1717         -mprefixed option.  Update the -mpcrel documentation to mention
1718         -mprefixed.
1719
1720 2020-02-11  Hans-Peter Nilsson  <hp@axis.com>
1721
1722         * ira-conflicts.c (print_hard_reg_set): Correct output for sets
1723         including FIRST_PSEUDO_REGISTER - 1.
1724         * ira-color.c (print_hard_reg_set): Ditto.
1725
1726 2020-02-11  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
1727
1728         * config/arm/arm-builtins.c (enum arm_type_qualifiers):
1729         (USTERNOP_QUALIFIERS): New define.
1730         (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
1731         (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
1732         (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
1733         (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
1734         * config/arm/arm_neon.h (vusdot_s32): New.
1735         (vusdot_lane_s32): New.
1736         (vusdotq_lane_s32): New.
1737         (vsudot_lane_s32): New.
1738         (vsudotq_lane_s32): New.
1739         * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
1740         * config/arm/iterators.md (DOTPROD_I8MM): New.
1741         (sup, opsuffix): Add <us/su>.
1742         * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
1743         * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
1744
1745 2020-02-11  Richard Biener  <rguenther@suse.de>
1746
1747         PR tree-optimization/93661
1748         PR tree-optimization/93662
1749         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
1750         tree_to_poly_int64.
1751         * tree-sra.c (get_access_for_expr): Likewise.
1752
1753 2020-02-10  Jakub Jelinek  <jakub@redhat.com>
1754
1755         PR target/93637
1756         * config/i386/sse.md (VI_256_AVX2): New mode iterator.
1757         (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
1758         Change condition from TARGET_AVX2 to TARGET_AVX.
1759
1760 2020-02-10  Iain Sandoe  <iain@sandoe.co.uk>
1761
1762         PR other/93641
1763         * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
1764         argument of strncmp.
1765
1766 2020-02-10  Hans-Peter Nilsson  <hp@axis.com>
1767
1768         Try to generate zero-based comparisons.
1769         * config/cris/cris.c (cris_reduce_compare): New function.
1770         * config/cris/cris-protos.h  (cris_reduce_compare): Add prototype.
1771         * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
1772         (cstore<mode>4"): Apply cris_reduce_compare in expanders.
1773
1774 2020-02-10  Richard Earnshaw  <rearnsha@arm.com>
1775
1776         PR target/91913
1777         * config/arm/arm.md (movsi_compare0): Allow SP as a source register
1778         in Thumb state and also as a destination in Arm state.  Add T16
1779         variants.
1780
1781 2020-02-10  Hans-Peter Nilsson  <hp@axis.com>
1782
1783         * md.texi (Define Subst): Match closing paren in example.
1784
1785 2020-02-10  Jakub Jelinek  <jakub@redhat.com>
1786
1787         PR target/58218
1788         PR other/93641
1789         * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
1790         arguments of strncmp.
1791
1792 2020-02-10  Feng Xue  <fxue@os.amperecomputing.com>
1793
1794         PR ipa/93203
1795         * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
1796         but different source value.
1797         (adjust_callers_for_value_intersection): New function.
1798         (gather_edges_for_value): Adjust order of callers to let a
1799         non-self-recursive caller be the first element.
1800         (self_recursive_pass_through_p): Add a new parameter "simple", and
1801         check generalized self-recursive pass-through jump function.
1802         (self_recursive_agg_pass_through_p): Likewise.
1803         (find_more_scalar_values_for_callers_subset): Compute value from
1804         pass-through jump function for self-recursive.
1805         (intersect_with_plats): Cleanup previous implementation code for value
1806         itersection with self-recursive call edge.
1807         (intersect_with_agg_replacements): Likewise.
1808         (intersect_aggregates_with_edge): Deduce value from pass-through jump
1809         function for self-recursive call edge.  Cleanup previous implementation
1810         code for value intersection with self-recursive call edge.
1811         (decide_whether_version_node): Remove dead callers and adjust order
1812         to let a non-self-recursive caller be the first element.
1813
1814 2020-02-09  Uroš Bizjak  <ubizjak@gmail.com>
1815
1816         * recog.c: Move pass_split_before_sched2 code in front of
1817         pass_split_before_regstack.
1818         (pass_data_split_before_sched2): Rename pass to split3 from split4.
1819         (pass_data_split_before_regstack): Rename pass to split4 from split3.
1820         (rest_of_handle_split_before_sched2): Remove.
1821         (pass_split_before_sched2::execute): Unconditionally call
1822         split_all_insns.
1823         (enable_split_before_sched2): New function.
1824         (pass_split_before_sched2::gate): Use enable_split_before_sched2.
1825         (pass_split_before_regstack::gate): Ditto.
1826         * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1827         Update name check for renamed split4 pass.
1828         * config/sh/sh.c (register_sh_passes): Update pass insertion
1829         point for renamed split4 pass.
1830
1831 2020-02-09  Jakub Jelinek  <jakub@redhat.com>
1832
1833         * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1834         DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1835         copying them around between host and target.
1836
1837 2020-02-08  Andrew Pinski  <apinski@marvell.com>
1838
1839         PR target/91927
1840         * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1841         STRICT_ALIGNMENT also.
1842
1843 2020-02-08  Jim Wilson  <jimw@sifive.com>
1844
1845         PR target/93532
1846         * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1847
1848 2020-02-08  Uroš Bizjak  <ubizjak@gmail.com>
1849             Jakub Jelinek  <jakub@redhat.com>
1850
1851         PR target/65782
1852         * config/i386/i386.h (CALL_USED_REGISTERS): Make
1853         xmm16-xmm31 call-used even in 64-bit ms-abi.
1854
1855 2020-02-07  Dennis Zhang  <dennis.zhang@arm.com>
1856
1857         * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1858         (simd_ummla, simd_usmmla): Likewise.
1859         * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1860         * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1861         (vusmmlaq_s32): New.
1862
1863 2020-02-07  Richard Biener  <rguenther@suse.de>
1864
1865         PR middle-end/93519
1866         * tree-inline.c (fold_marked_statements): Do a PRE walk,
1867         skipping unreachable regions.
1868         (optimize_inline_calls): Skip folding stmts when we didn't
1869         inline.
1870
1871 2020-02-07  H.J. Lu  <hongjiu.lu@intel.com>
1872
1873         PR target/85667
1874         * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1875         Don't return aggregates with only SFmode and DFmode in SSE
1876         register.
1877         (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1878
1879 2020-02-07  Jakub Jelinek  <jakub@redhat.com>
1880
1881         PR target/93122
1882         * config/rs6000/rs6000-logue.c
1883         (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1884         if it fails, move rs into end_addr and retry.  Add
1885         REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1886         the insn pattern doesn't describe well what exactly happens to
1887         dwarf2cfi.c.
1888
1889         PR target/93594
1890         * config/i386/predicates.md (avx_identity_operand): Remove.
1891         * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1892         (avx_<castmode><avxsizesuffix>_<castmode>,
1893         avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1894         a VEC_CONCAT of the operand and UNSPEC_CAST.
1895         (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1896         a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1897         UNSPEC_CAST.
1898
1899         PR target/93611
1900         * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1901         recog_data.insn if distance_non_agu_define changed it.
1902
1903 2020-02-06  Michael Meissner  <meissner@linux.ibm.com>
1904
1905         PR target/93569
1906         * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1907         we only had X-FORM (reg+reg) addressing for vectors.  Also before
1908         ISA 3.0, we only had X-FORM addressing for scalars in the
1909         traditional Altivec registers.
1910
1911 2020-02-06  <zhongyunde@huawei.com>
1912             Vladimir Makarov  <vmakarov@redhat.com>
1913
1914         PR rtl-optimization/93561
1915         * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1916         hard register range.
1917
1918 2020-02-06  Richard Sandiford  <richard.sandiford@arm.com>
1919
1920         * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1921         attribute.
1922
1923 2020-02-06  Segher Boessenkool  <segher@kernel.crashing.org>
1924
1925         * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1926         where the low and the high 32 bits are equal to each other specially,
1927         with an rldimi instruction.
1928
1929 2020-02-06  Mihail Ionescu  <mihail.ionescu@arm.com>
1930
1931         * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1932
1933 2020-02-06  Mihail Ionescu  <mihail.ionescu@arm.com>
1934
1935         * config/arm/arm-tables.opt: Regenerate.
1936
1937 2020-02-06  Richard Sandiford  <richard.sandiford@arm.com>
1938
1939         PR target/87763
1940         * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1941         * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1942         * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1943
1944 2020-02-06  Richard Sandiford  <richard.sandiford@arm.com>
1945
1946         PR rtl-optimization/87763
1947         * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1948
1949 2020-02-06  Delia Burduv  <delia.burduv@arm.com>
1950
1951         * config/aarch64/aarch64-simd-builtins.def
1952         (bfmlaq): New built-in function.
1953         (bfmlalb): New built-in function.
1954         (bfmlalt): New built-in function.
1955         (bfmlalb_lane): New built-in function.
1956         (bfmlalt_lane): New built-in function.
1957         * config/aarch64/aarch64-simd.md
1958         (aarch64_bfmmlaqv4sf): New pattern.
1959         (aarch64_bfmlal<bt>v4sf): New pattern.
1960         (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1961         * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1962         (vbfmlalbq_f32): New intrinsic.
1963         (vbfmlaltq_f32): New intrinsic.
1964         (vbfmlalbq_lane_f32): New intrinsic.
1965         (vbfmlaltq_lane_f32): New intrinsic.
1966         (vbfmlalbq_laneq_f32): New intrinsic.
1967         (vbfmlaltq_laneq_f32): New intrinsic.
1968         * config/aarch64/iterators.md (BF_MLA): New int iterator.
1969         (bt): New int attribute.
1970
1971 2020-02-06  Uroš Bizjak  <ubizjak@gmail.com>
1972
1973         * config/i386/i386.md (*pushtf): Emit "#" instead of
1974         calling gcc_unreachable in insn output.
1975         (*pushxf): Ditto.
1976         (*pushdf): Ditto.
1977         (*pushsf_rex64): Ditto for alternatives other than 1.
1978         (*pushsf): Ditto for alternatives other than 1.
1979
1980 2020-02-06  Martin Liska  <mliska@suse.cz>
1981
1982         PR gcov-profile/91971
1983         PR gcov-profile/93466
1984         * coverage.c (coverage_init): Revert mangling of
1985         path into filename.  It can lead to huge filename length.
1986         Creation of subfolders seem more natural.
1987
1988 2020-02-06  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
1989
1990         PR target/93300
1991         * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1992         (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1993         Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1994
1995 2020-02-06  Jakub Jelinek  <jakub@redhat.com>
1996
1997         PR target/93594
1998         * config/i386/predicates.md (avx_identity_operand): New predicate.
1999         * config/i386/sse.md (*avx_vec_concat<mode>_1): New
2000         define_insn_and_split.
2001
2002         PR libgomp/93515
2003         * omp-low.c (use_pointer_for_field): For nested constructs, also
2004         look for map clauses on target construct.
2005         (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
2006         taskreg_nesting_level.
2007
2008         PR libgomp/93515
2009         * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
2010         shared clause, call omp_notice_variable on outer context if any.
2011
2012 2020-02-05  Jason Merrill  <jason@redhat.com>
2013
2014         PR c++/92003
2015         * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
2016         non-zero address even if weak and not yet defined.
2017
2018 2020-02-05  Martin Sebor  <msebor@redhat.com>
2019
2020         PR tree-optimization/92765
2021         * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
2022         * tree-ssa-strlen.c (compute_string_length): Remove.
2023         (determine_min_objsize): Remove.
2024         (get_len_or_size): Add an argument.  Call get_range_strlen_dynamic.
2025         Avoid using type size as the upper bound on string length.
2026         (handle_builtin_string_cmp): Add an argument.  Adjust.
2027         (strlen_check_and_optimize_call): Pass additional argument to
2028         handle_builtin_string_cmp.
2029
2030 2020-02-05  Uroš Bizjak  <ubizjak@gmail.com>
2031
2032         * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
2033         (*pushdi2_rex64 peephole2): Unconditionally split after
2034         epilogue_completed.
2035         (*ashl<mode>3_doubleword): Ditto.
2036         (*<shift_insn><mode>3_doubleword): Ditto.
2037
2038 2020-02-05  Michael Meissner  <meissner@linux.ibm.com>
2039
2040         PR target/93568
2041         * config/rs6000/rs6000.c (get_vector_offset): Fix
2042
2043 2020-02-05  Andrew Stubbs  <ams@codesourcery.com>
2044
2045         * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
2046
2047 2020-02-05  David Malcolm  <dmalcolm@redhat.com>
2048
2049         * doc/analyzer.texi
2050         (Special Functions for Debugging the Analyzer): Update description
2051         of __analyzer_dump_exploded_nodes.
2052
2053 2020-02-05  Jakub Jelinek  <jakub@redhat.com>
2054
2055         PR target/92190
2056         * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
2057         include sets and not clobbers in the vzeroupper pattern.
2058         * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
2059         the parallel has 17 (64-bit) or 9 (32-bit) elts.
2060         (*avx_vzeroupper_1): New define_insn_and_split.
2061
2062         PR target/92190
2063         * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
2064         don't run when !optimize.
2065         (pass_split_before_regstack::gate): For STACK_REGS targets, run even
2066         when !optimize.
2067
2068 2020-02-05  Richard Biener  <rguenther@suse.de>
2069
2070         PR middle-end/90648
2071         * genmatch.c (dt_node::gen_kids_1): Emit number of argument
2072         checks before matching calls.
2073
2074 2020-02-05  Jakub Jelinek  <jakub@redhat.com>
2075
2076         * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
2077         function comment typo.
2078
2079         PR middle-end/93555
2080         * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
2081         simd_clone_create failed when i == 0, adjust clone->nargs by
2082         clone->inbranch.
2083
2084 2020-02-05  Martin Liska  <mliska@suse.cz>
2085
2086         PR c++/92717
2087         * doc/invoke.texi: Document that one should
2088         not combine ASLR and -fpch.
2089
2090 2020-02-04  Richard Biener  <rguenther@suse.de>
2091
2092         PR tree-optimization/93538
2093         * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
2094
2095 2020-02-04  Richard Biener  <rguenther@suse.de>
2096
2097         PR tree-optimization/91123
2098         * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
2099         (vn_walk_cb_data::last_vuse): New member.
2100         (vn_walk_cb_data::saved_operands): Likewsie.
2101         (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
2102         (vn_walk_cb_data::push_partial_def): Use finish.
2103         (vn_reference_lookup_2): Update last_vuse and use finish if
2104         we've saved operands.
2105         (vn_reference_lookup_3): Use finish and update calls to
2106         push_partial_defs everywhere.  When translating through
2107         memcpy or aggregate copies save off operands and alias-set.
2108         (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
2109         operation for redundant store removal.
2110
2111 2020-02-04  Richard Biener  <rguenther@suse.de>
2112
2113         PR tree-optimization/92819
2114         * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
2115         generating more stmts than before.
2116
2117 2020-02-04  Martin Liska  <mliska@suse.cz>
2118
2119         * config/arm/arm.c (arm_gen_far_branch): Move the function
2120         outside of selftests.
2121
2122 2020-02-03  Michael Meissner  <meissner@linux.ibm.com>
2123
2124         * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
2125         function to adjust PC-relative vector addresses.
2126         (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
2127         handle vectors with PC-relative addresses.
2128
2129 2020-02-03  Michael Meissner  <meissner@linux.ibm.com>
2130
2131         * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
2132         reference.
2133         (hard_reg_and_mode_to_addr_mask): Delete.
2134         (rs6000_adjust_vec_address): If the original vector address
2135         was REG+REG or REG+OFFSET and the element is not zero, do the add
2136         of the elements in the original address before adding the offset
2137         for the vector element.  Use address_to_insn_form to validate the
2138         address using the register being loaded, rather than guessing
2139         whether the address is a DS-FORM or DQ-FORM address.
2140
2141 2020-02-03  Michael Meissner  <meissner@linux.ibm.com>
2142
2143         * config/rs6000/rs6000.c (get_vector_offset): New helper function
2144         to calculate the offset in memory from the start of a vector of a
2145         particular element.  Add code to keep the element number in
2146         bounds if the element number is variable.
2147         (rs6000_adjust_vec_address): Move calculation of offset of the
2148         vector element to get_vector_offset.
2149         (rs6000_split_vec_extract_var): Do not do the initial AND of
2150         element here, move the code to get_vector_offset.
2151
2152 2020-02-03  Michael Meissner  <meissner@linux.ibm.com>
2153
2154         * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
2155         gcc_asserts.
2156
2157 2020-02-03  Segher Boessenkool  <segher@kernel.crashing.org>
2158
2159         * config/rs6000/constraints.md: Improve documentation.
2160
2161 2020-02-03  Richard Earnshaw  <rearnsha@arm.com>
2162
2163         PR target/93548
2164         * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
2165         ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
2166
2167 2020-02-03  Andrew Stubbs  <ams@codesourcery.com>
2168
2169         * config.gcc: Remove "carrizo" support.
2170         * config/gcn/gcn-opts.h (processor_type): Likewise.
2171         * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
2172         * config/gcn/gcn.opt (gpu_type): Likewise.
2173         * config/gcn/t-omp-device: Likewise.
2174
2175 2020-02-03  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
2176
2177         PR target/91816
2178         * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
2179         * config/arm/arm.c (arm_gen_far_branch): New function
2180         arm_gen_far_branch.
2181         * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
2182
2183 2020-02-03  Julian Brown  <julian@codesourcery.com>
2184             Tobias Burnus  <tobias@codesourcery.com>
2185
2186         * doc/invoke.texi: Update mention of OpenACC version to 2.6.
2187
2188 2020-02-03  Jakub Jelinek  <jakub@redhat.com>
2189
2190         PR target/93533
2191         * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
2192         valid RTL to sum up the lowest and second lowest bytes of the popcnt
2193         result.
2194
2195 2020-02-02  Vladimir Makarov  <vmakarov@redhat.com>
2196
2197         PR rtl-optimization/91333
2198         * ira-color.c (struct allocno_color_data): Add member
2199         hard_reg_prefs.
2200         (init_allocno_threads): Set the member up.
2201         (bucket_allocno_compare_func): Add compare hard reg
2202         prefs.
2203
2204 2020-01-31  Sandra Loosemore  <sandra@codesourcery.com>
2205
2206         nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
2207
2208         * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
2209         * config.in: Regenerated.
2210         * configure: Regenerated.
2211         * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
2212         for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
2213         (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
2214
2215 2020-02-01  Andrew Burgess  <andrew.burgess@embecosm.com>
2216
2217         * configure: Regenerate.
2218
2219 2020-01-31  Vladimir Makarov  <vmakarov@redhat.com>
2220
2221         PR rtl-optimization/91333
2222         * ira-color.c (bucket_allocno_compare_func): Move conflict hard
2223         reg preferences comparison up.
2224
2225 2020-01-31  Richard Sandiford  <richard.sandiford@arm.com>
2226
2227         * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
2228         * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
2229         aarch64-sve-builtins-base.h.
2230         * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
2231         aarch64-sve-builtins-base.cc.
2232         * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
2233         (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2234         (svcvtnt): Declare.
2235         * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
2236         (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2237         (svcvtnt): New functions.
2238         * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
2239         (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2240         (svcvtnt): New functions.
2241         (svcvt): Add a form that converts f32 to bf16.
2242         * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
2243         (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
2244         Declare.
2245         * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
2246         Treat B as bfloat16_t.
2247         (ternary_bfloat_lane_base): New class.
2248         (ternary_bfloat_def): Likewise.
2249         (ternary_bfloat): New shape.
2250         (ternary_bfloat_lane_def): New class.
2251         (ternary_bfloat_lane): New shape.
2252         (ternary_bfloat_lanex2_def): New class.
2253         (ternary_bfloat_lanex2): New shape.
2254         (ternary_bfloat_opt_n_def): New class.
2255         (ternary_bfloat_opt_n): New shape.
2256         * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
2257         * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
2258         (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
2259         (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
2260         (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
2261         (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
2262         (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
2263         * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
2264         the pattern off the narrow mode instead of the wider one.
2265         * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
2266         (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
2267         (sve_fp_op): Handle them.
2268         (SVE_BFLOAT_TERNARY_LONG): New int itertor.
2269         (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
2270
2271 2020-01-31  Richard Sandiford  <richard.sandiford@arm.com>
2272
2273         * config/aarch64/arm_sve.h: Include arm_bf16.h.
2274         * config/aarch64/aarch64-modes.def (BF): Move definition before
2275         VECTOR_MODES.  Remove separate VECTOR_MODES for V4BF and V8BF.
2276         (SVE_MODES): Handle BF modes.
2277         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
2278         BF modes.
2279         (aarch64_full_sve_mode): Likewise.
2280         * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
2281         and VNx32BF.
2282         (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
2283         (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
2284         (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
2285         (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
2286         new SVE BF modes.
2287         * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
2288         type_class_index.
2289         * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
2290         (TYPES_all_data): Add bf16.
2291         (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
2292         (register_tuple_type): Increase buffer size.
2293         * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
2294         (bf16): New type suffix.
2295         * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
2296         (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
2297         (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
2298         Change type from all_data to all_arith.
2299         * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
2300         (svminp): Likewise.
2301
2302 2020-01-31  Dennis Zhang  <dennis.zhang@arm.com>
2303             Matthew Malcomson  <matthew.malcomson@arm.com>
2304             Richard Sandiford  <richard.sandiford@arm.com>
2305
2306         * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
2307         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
2308         __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
2309         __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate.  Don't define
2310         __ARM_FEATURE_MATMUL_FP64.
2311         * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
2312         (sve): Add AARCH64_FL_F32MM to the list of extensions that should
2313         be disabled at the same time.
2314         (f32mm): New extension.
2315         * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
2316         (AARCH64_FL_F64MM): Bump to the next bit up.
2317         (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
2318         (TARGET_SVE_F64MM): New macros.
2319         * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
2320         (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
2321         (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
2322         (UNSPEC_ZIP2Q): New unspeccs.
2323         (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
2324         (optab, sur, perm_insn): Handle the new unspecs.
2325         (sve_fp_op): Handle UNSPEC_FMMLA.  Resort.
2326         * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
2327         TARGET_SVE_F64MM instead of separate tests.
2328         (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
2329         (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
2330         (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
2331         (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
2332         (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
2333         * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
2334         (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
2335         (TYPES_s_signed): New macro.
2336         (TYPES_s_integer): Use it.
2337         (TYPES_d_float): New macro.
2338         (TYPES_d_data): Use it.
2339         * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
2340         (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
2341         (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
2342         * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
2343         (svmmla): New shape.
2344         (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
2345         template parameters.
2346         (ternary_resize2_lane_base): Likewise.
2347         (ternary_resize2_base): New class.
2348         (ternary_qq_lane_base): Likewise.
2349         (ternary_intq_uintq_lane_def): Likewise.
2350         (ternary_intq_uintq_lane): New shape.
2351         (ternary_intq_uintq_opt_n_def): New class
2352         (ternary_intq_uintq_opt_n): New shape.
2353         (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
2354         (ternary_uintq_intq_def): New class.
2355         (ternary_uintq_intq): New shape.
2356         (ternary_uintq_intq_lane_def): New class.
2357         (ternary_uintq_intq_lane): New shape.
2358         (ternary_uintq_intq_opt_n_def): New class.
2359         (ternary_uintq_intq_opt_n): New shape.
2360         * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
2361         (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
2362         (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
2363         * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
2364         Generalize to...
2365         (svdotprod_lane_impl): ...this new class.
2366         (svmmla_impl, svusdot_impl): New classes.
2367         (svdot_lane): Update to use svdotprod_lane_impl.
2368         (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
2369         (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
2370         functions.
2371         * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
2372         function, with no types defined.
2373         (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
2374         AARCH64_FL_I8MM functions.
2375         (svmmla): New AARCH64_FL_F32MM function.
2376         (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
2377         (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
2378         AARCH64_FL_F64MM function.
2379         (REQUIRED_EXTENSIONS):
2380
2381 2020-01-31  Andrew Stubbs  <ams@codesourcery.com>
2382
2383         * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
2384         alternative only.
2385
2386 2020-01-31  Uroš Bizjak  <ubizjak@gmail.com>
2387
2388         * config/i386/i386.md (*movoi_internal_avx): Do not check for
2389         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.  Remove MODE_V8SF handling.
2390         (*movti_internal): Do not check for
2391         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
2392         (*movtf_internal): Move check for TARGET_SSE2 and size optimization
2393         just after check for TARGET_AVX.
2394         (*movdf_internal): Ditto.
2395         * config/i386/mmx.md (*mov<mode>_internal): Do not check for
2396         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
2397         * config/i386/sse.md (mov<mode>_internal): Only check
2398         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode.  Move check
2399         for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
2400         (<sse>_andnot<mode>3<mask_name>): Move check for
2401         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
2402         (<code><mode>3<mask_name>): Ditto.
2403         (*andnot<mode>3): Ditto.
2404         (*andnottf3): Ditto.
2405         (*<code><mode>3): Ditto.
2406         (*<code>tf3): Ditto.
2407         (*andnot<VI:mode>3): Remove
2408         TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
2409         (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
2410         (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
2411         (sse4_1_blendv<ssemodesuffix>): Ditto.
2412         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
2413         Explain that tune applies to 128bit instructions only.
2414
2415 2020-01-31  Kwok Cheung Yeung  <kcy@codesourcery.com>
2416
2417         * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
2418         to definition of hsa_kernel_description.  Parse assembly to find SGPR
2419         and VGPR count of kernel and store in hsa_kernel_description.
2420
2421 2020-01-31  Tamar Christina  <tamar.christina@arm.com>
2422
2423         PR rtl-optimization/91838
2424         * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
2425         to truncate if allowed or reject combination.
2426
2427 2020-01-31  Andrew Stubbs  <ams@codesourcery.com>
2428
2429         * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
2430         (find_inv_vars_cb): Likewise.
2431
2432 2020-01-31  David Malcolm  <dmalcolm@redhat.com>
2433
2434         * calls.c (special_function_p): Split out the check for DECL_NAME
2435         being non-NULL and fndecl being extern at file scope into a
2436         new maybe_special_function_p and call it.  Drop check for fndecl
2437         being non-NULL that was after a usage of DECL_NAME (fndecl).
2438         * tree.h (maybe_special_function_p): New inline function.
2439
2440 2020-01-30  Andrew Stubbs  <ams@codesourcery.com>
2441
2442         * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
2443         (mask_gather_load<mode>): ... here, and zero-initialize the
2444         destination.
2445         (maskload<mode>di): Zero-initialize the destination.
2446         * config/gcn/gcn.c:
2447
2448 2020-01-30  David Malcolm  <dmalcolm@redhat.com>
2449
2450         PR analyzer/93356
2451         * doc/analyzer.texi (Limitations): Note that constraints on
2452         floating-point values are currently ignored.
2453
2454 2020-01-30  Jakub Jelinek  <jakub@redhat.com>
2455
2456         PR lto/93384
2457         * symtab.c (symtab_node::noninterposable_alias): If localalias
2458         already exists, but is not usable, append numbers after it until
2459         a unique name is found.  Formatting fix.
2460
2461         PR middle-end/93505
2462         * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
2463         rotate counts.
2464
2465 2020-01-30  Andrew Stubbs  <ams@codesourcery.com>
2466
2467         * config/gcn/gcn.c (print_operand): Handle LTGT.
2468         * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
2469
2470 2020-01-30  Richard Biener  <rguenther@suse.de>
2471
2472         * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
2473         and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
2474
2475 2020-01-30  John David Anglin  <danglin@gcc.gnu.org>
2476
2477         * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
2478         without a DECL in .data.rel.ro.local.
2479
2480 2020-01-30  Jakub Jelinek  <jakub@redhat.com>
2481
2482         PR target/93494
2483         * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
2484         returned.
2485
2486         PR target/91824
2487         * config/i386/sse.md
2488         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
2489         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this.  Use
2490         any_extend code iterator instead of always zero_extend.
2491         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
2492         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
2493         Use any_extend code iterator instead of always zero_extend.
2494         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
2495         (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
2496         Use any_extend code iterator instead of always zero_extend.
2497         (*sse2_pmovmskb_ext): New define_insn.
2498         (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
2499
2500         PR target/91824
2501         * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
2502         (*popcountsi2_zext_falsedep): New define_insn.
2503
2504 2020-01-30  Dragan Mladjenovic  <dmladjenovic@wavecomp.com>
2505
2506         * config.in: Regenerated.
2507         * configure: Regenerated.
2508
2509 2020-01-29  Tobias Burnus  <tobias@codesourcery.com>
2510
2511         PR bootstrap/93409
2512         * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
2513         LLVM's assembler changed the default in version 9.
2514
2515 2020-01-24  Jeff Law  <law@redhat.com>
2516
2517         PR tree-optimization/89689
2518         * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
2519
2520 2020-01-29  Richard Sandiford  <richard.sandiford@arm.com>
2521
2522         Revert:
2523
2524         2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
2525
2526         PR rtl-optimization/87763
2527         * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2528         simplification to handle subregs as well as bare regs.
2529         * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2530
2531 2020-01-29  Joel Hutton  <Joel.Hutton@arm.com>
2532
2533         PR target/93221
2534         * ira.c (ira): Revert use of simplified LRA algorithm.
2535
2536 2020-01-29  Martin Jambor  <mjambor@suse.cz>
2537
2538         PR tree-optimization/92706
2539         * tree-sra.c (struct access): Fields first_link, last_link,
2540         next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
2541         next_rhs_queued and grp_rhs_queued respectively, new fields
2542         first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
2543         (struct assign_link): Field next renamed to next_rhs, new field
2544         next_lhs.  Updated comment.
2545         (work_queue_head): Renamed to rhs_work_queue_head.
2546         (lhs_work_queue_head): New variable.
2547         (add_link_to_lhs): New function.
2548         (relink_to_new_repr): Also relink LHS lists.
2549         (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
2550         (add_access_to_lhs_work_queue): New function.
2551         (pop_access_from_work_queue): Renamed to
2552         pop_access_from_rhs_work_queue.
2553         (pop_access_from_lhs_work_queue): New function.
2554         (build_accesses_from_assign): Also add links to LHS lists and to LHS
2555         work_queue.
2556         (child_would_conflict_in_lacc): Renamed to
2557         child_would_conflict_in_acc.  Adjusted parameter names.
2558         (create_artificial_child_access): New parameter set_grp_read, use it.
2559         (subtree_mark_written_and_enqueue): Renamed to
2560         subtree_mark_written_and_rhs_enqueue.
2561         (propagate_subaccesses_across_link): Renamed to
2562         propagate_subaccesses_from_rhs.
2563         (propagate_subaccesses_from_lhs): New function.
2564         (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
2565         RHSs.
2566
2567 2020-01-29  Martin Jambor  <mjambor@suse.cz>
2568
2569         PR tree-optimization/92706
2570         * tree-sra.c (struct access): Adjust comment of
2571         grp_total_scalarization.
2572         (find_access_in_subtree): Look for single children spanning an entire
2573         access.
2574         (scalarizable_type_p): Allow register accesses, adjust callers.
2575         (completely_scalarize): Remove function.
2576         (scalarize_elem): Likewise.
2577         (create_total_scalarization_access): Likewise.
2578         (sort_and_splice_var_accesses): Do not track total scalarization
2579         flags.
2580         (analyze_access_subtree): New parameter totally, adjust to new meaning
2581         of grp_total_scalarization.
2582         (analyze_access_trees): Pass new parameter to analyze_access_subtree.
2583         (can_totally_scalarize_forest_p): New function.
2584         (create_total_scalarization_access): Likewise.
2585         (create_total_access_and_reshape): Likewise.
2586         (total_should_skip_creating_access): Likewise.
2587         (totally_scalarize_subtree): Likewise.
2588         (analyze_all_variable_accesses): Perform total scalarization after
2589         subaccess propagation using the new functions above.
2590         (initialize_constant_pool_replacements): Output initializers by
2591         traversing the access tree.
2592
2593 2020-01-29  Martin Jambor  <mjambor@suse.cz>
2594
2595         * tree-sra.c (verify_sra_access_forest): New function.
2596         (verify_all_sra_access_forests): Likewise.
2597         (create_artificial_child_access): Set parent.
2598         (analyze_all_variable_accesses): Call the verifier.
2599
2600 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
2601
2602         * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
2603         if called on indirect edge.
2604         (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
2605         speculative call if needed.
2606
2607 2020-01-29  Richard Biener  <rguenther@suse.de>
2608
2609         PR tree-optimization/93428
2610         * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
2611         permutation when the load node is created.
2612         (vect_analyze_slp_instance): Re-use it here.
2613
2614 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
2615
2616         * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
2617
2618 2020-01-28  Vladimir Makarov  <vmakarov@redhat.com>
2619
2620         PR rtl-optimization/93272
2621         * ira-lives.c (process_out_of_region_eh_regs): New function.
2622         (process_bb_node_lives): Call it.
2623
2624 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
2625
2626         * coverage.c (read_counts_file): Make error message lowercase.
2627
2628 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
2629
2630         * profile-count.c (profile_quality_display_names): Fix ordering.
2631
2632 2020-01-28  Jan Hubicka  <hubicka@ucw.cz>
2633
2634         PR lto/93318
2635         * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
2636         hash only when edge is first within the sequence.
2637         (cgraph_edge::set_call_stmt): Update handling of speculative calls.
2638         (symbol_table::create_edge): Do not set target_prob.
2639         (cgraph_edge::remove_caller): Watch for speculative calls when updating
2640         the call site hash.
2641         (cgraph_edge::make_speculative): Drop target_prob parameter.
2642         (cgraph_edge::speculative_call_info): Remove.
2643         (cgraph_edge::first_speculative_call_target): New member function.
2644         (update_call_stmt_hash_for_removing_direct_edge): New function.
2645         (cgraph_edge::resolve_speculation): Rewrite to new API.
2646         (cgraph_edge::speculative_call_for_target): New member function.
2647         (cgraph_edge::make_direct): Rewrite to new API; fix handling of
2648         multiple speculation targets.
2649         (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
2650         of profile.
2651         (verify_speculative_call): Verify that targets form an interval.
2652         * cgraph.h (cgraph_edge::speculative_call_info): Remove.
2653         (cgraph_edge::first_speculative_call_target): New member function.
2654         (cgraph_edge::next_speculative_call_target): New member function.
2655         (cgraph_edge::speculative_call_target_ref): New member function.
2656         (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
2657         (cgraph_edge): Remove target_prob.
2658         * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
2659         Fix handling of speculative calls.
2660         * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
2661         * ipa-fnsummary.c (analyze_function_body): Likewise.
2662         * ipa-inline.c (speculation_useful_p): Use new speculative call API.
2663         * ipa-profile.c (dump_histogram): Fix formating.
2664         (ipa_profile_generate_summary): Watch for overflows.
2665         (ipa_profile): Do not require probablity to be 1/2; update to new API.
2666         * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
2667         (update_indirect_edges_after_inlining): Update to new API.
2668         * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
2669         profiles.
2670         * profile-count.h: (profile_probability::adjusted): New.
2671         * tree-inline.c (copy_bb): Update to new speculative call API; fix
2672         updating of profile.
2673         * value-prof.c (gimple_ic_transform): Rename to ...
2674         (dump_ic_profile): ... this one; update dumping.
2675         (stream_in_histogram_value): Fix formating.
2676         (gimple_value_profile_transformations): Update.
2677
2678 2020-01-28  H.J. Lu  <hongjiu.lu@intel.com>
2679
2680         PR target/91461
2681         * config/i386/i386.md (*movoi_internal_avx): Remove
2682         TARGET_SSE_TYPELESS_STORES check.
2683         (*movti_internal): Prefer TARGET_AVX over
2684         TARGET_SSE_TYPELESS_STORES.
2685         (*movtf_internal): Likewise.
2686         * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
2687         TARGET_SSE_TYPELESS_STORES.  Remove "<MODE_SIZE> == 16" check
2688         from TARGET_SSE_TYPELESS_STORES.
2689
2690 2020-01-28  David Malcolm  <dmalcolm@redhat.com>
2691
2692         * diagnostic-core.h (warning_at): Rename overload to...
2693         (warning_meta): ...this.
2694         (emit_diagnostic_valist): Delete decl of overload taking
2695         diagnostic_metadata.
2696         * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
2697         (warning_at): Rename overload taking diagnostic_metadata to...
2698         (warning_meta): ...this.
2699
2700 2020-01-28  Richard Biener  <rguenther@suse.de>
2701
2702         PR tree-optimization/93439
2703         * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
2704         * tree-cfg.c (move_sese_region_to_fn): ... here.
2705         (verify_types_in_gimple_reference): Verify used cliques are
2706         tracked.
2707
2708 2020-01-28  H.J. Lu  <hongjiu.lu@intel.com>
2709
2710         PR target/91399
2711         * config/i386/i386-options.c (set_ix86_tune_features): Add an
2712         argument of a pointer to struct gcc_options and pass it to
2713         parse_mtune_ctrl_str.
2714         (ix86_function_specific_restore): Pass opts to
2715         set_ix86_tune_features.
2716         (ix86_option_override_internal): Likewise.
2717         (parse_mtune_ctrl_str): Add an argument of a pointer to struct
2718         gcc_options and use it for x_ix86_tune_ctrl_string.
2719
2720 2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
2721
2722         PR rtl-optimization/87763
2723         * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2724         simplification to handle subregs as well as bare regs.
2725         * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2726
2727 2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
2728
2729         * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
2730         for reduction chains that (now) include a call.
2731
2732 2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
2733
2734         PR tree-optimization/92822
2735         * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
2736         out the don't-care elements of a vector whose significant elements
2737         are duplicates, make the don't-care elements duplicates too.
2738
2739 2020-01-28  Richard Sandiford  <richard.sandiford@arm.com>
2740
2741         PR tree-optimization/93434
2742         * tree-predcom.c (split_data_refs_to_components): Record which
2743         components have had aliasing loads removed.  Prevent store-store
2744         commoning for all such components.
2745
2746 2020-01-28  Jakub Jelinek  <jakub@redhat.com>
2747
2748         PR target/93418
2749         * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
2750         -1 or is_vshift is true, use new_vector with number of elts npatterns
2751         rather than new_unary_operation.
2752
2753         PR tree-optimization/93454
2754         * gimple-fold.c (fold_array_ctor_reference): Perform
2755         elt_size.to_uhwi () just once, instead of calling it in every
2756         iteration.  Punt if that value is above size of the temporary
2757         buffer.  Decrease third native_encode_expr argument when
2758         bufoff + elt_sz is above size of buf.
2759
2760 2020-01-27  Joseph Myers  <joseph@codesourcery.com>
2761
2762         * config/mips/mips.c (mips_declare_object_name)
2763         [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
2764
2765 2020-01-27  Martin Liska  <mliska@suse.cz>
2766
2767         PR gcov-profile/93403
2768         * tree-profile.c (gimple_init_gcov_profiler): Generate
2769         both __gcov_indirect_call_profiler_v4 and
2770         __gcov_indirect_call_profiler_v4_atomic.
2771
2772 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
2773
2774         PR target/92822
2775         * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
2776         expander.
2777         (@aarch64_split_simd_mov<mode>): Use it.
2778         (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
2779         Leave the vec_extract patterns to handle 2-element vectors.
2780         (aarch64_simd_mov_from_<mode>high): Likewise.
2781         (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
2782         (vec_extractv2dfv1df): Likewise.
2783
2784 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
2785
2786         * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
2787         jump conditions for *compare_condjump<GPI:mode>.
2788
2789 2020-01-27  David Malcolm  <dmalcolm@redhat.com>
2790
2791         PR analyzer/93276
2792         * digraph.cc (test_edge::test_edge): Specify template for base
2793         class initializer.
2794
2795 2020-01-27  Claudiu Zissulescu  <claziss@synopsys.com>
2796
2797         * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
2798
2799 2020-01-27  Claudiu Zissulescu  <claziss@synopsys.com>
2800
2801         * config/arc/arc-protos.h (gen_mlo): Remove.
2802         (gen_mhi): Likewise.
2803         * config/arc/arc.c (AUX_MULHI): Define.
2804         (arc_must_save_reister): Special handling for r58/59.
2805         (arc_compute_frame_size): Consider mlo/mhi registers.
2806         (arc_save_callee_saves): Emit fp/sp move only when emit_move
2807         paramter is true.
2808         (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
2809         mlo/mhi name selection.
2810         (arc_restore_callee_saves): Don't early restore blink when ISR.
2811         (arc_expand_prologue): Add mlo/mhi saving.
2812         (arc_expand_epilogue): Add mlo/mhi restoring.
2813         (gen_mlo): Remove.
2814         (gen_mhi): Remove.
2815         * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
2816         numbering when MUL64 option is used.
2817         (DWARF2_FRAME_REG_OUT): Define.
2818         * config/arc/arc.md (arc600_stall): New pattern.
2819         (VUNSPEC_ARC_ARC600_STALL): Define.
2820         (mulsi64): Use correct mlo/mhi registers.
2821         (mulsi_600): Clean it up.
2822         * config/arc/predicates.md (mlo_operand): Remove any dependency on
2823         TARGET_BIG_ENDIAN.
2824         (mhi_operand): Likewise.
2825
2826 2020-01-27  Claudiu Zissulescu  <claziss@synopsys.com>
2827             Petro Karashchenko  <petro.karashchenko@ring.com>
2828
2829         * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2830         attributes if needed.
2831         (prepare_move_operands): Generate special unspec instruction for
2832         direct access.
2833         (arc_isuncached_mem_p): Propagate uncached attribute to each
2834         structure member.
2835         * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2836         (VUNSPEC_ARC_STDI): Likewise.
2837         (ALLI): New mode iterator.
2838         (mALLI): New mode attribute.
2839         (lddi): New instruction pattern.
2840         (stdi): Likewise.
2841         (stdidi_split): Split instruction for architectures which are not
2842         supporting ll64 option.
2843         (lddidi_split): Likewise.
2844
2845 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
2846
2847         PR rtl-optimization/92989
2848         * lra-lives.c (process_bb_lives): Update the live-in set before
2849         processing additional clobbers.
2850
2851 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
2852
2853         PR rtl-optimization/93170
2854         * cselib.c (cselib_invalidate_regno_val): New function, split out
2855         from...
2856         (cselib_invalidate_regno): ...here.
2857         (cselib_invalidated_by_call_p): New function.
2858         (cselib_process_insn): Iterate over all the hard-register entries in
2859         REG_VALUES and invalidate any that cross call-clobbered registers.
2860
2861 2020-01-27  Richard Sandiford  <richard.sandiford@arm.com>
2862
2863         * dojump.c (split_comparison): Use HONOR_NANS rather than
2864         HONOR_SNANS when splitting LTGT.
2865
2866 2020-01-27  Martin Liska  <mliska@suse.cz>
2867
2868         PR driver/91220
2869         * opts.c (print_filtered_help): Exclude language-specific
2870         options from --help=common unless enabled in all FEs.
2871
2872 2020-01-27  Martin Liska  <mliska@suse.cz>
2873
2874         * opts.c (print_help): Exclude params from
2875         all except --help=param.
2876
2877 2020-01-27  Martin Liska  <mliska@suse.cz>
2878
2879         PR target/93274
2880         * config/i386/i386-features.c (make_resolver_func):
2881         Align the code with ppc64 target implementation.
2882         Do not generate a unique name for resolver function.
2883
2884 2020-01-27  Richard Biener  <rguenther@suse.de>
2885
2886         PR tree-optimization/93397
2887         * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2888         converted reduction chain SLP graph adjustment.
2889
2890 2020-01-26  Marek Polacek  <polacek@redhat.com>
2891
2892         PR sanitizer/93436
2893         * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2894         null DECL_NAME.
2895
2896 2020-01-26  Jason Merrill  <jason@redhat.com>
2897
2898         PR c++/92601
2899         * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2900         of complete types.
2901
2902 2020-01-26  Darius Galis  <darius.galis@cyberthorstudios.com>
2903
2904         * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2905         (rx_setmem): Likewise.
2906
2907 2020-01-26  Jakub Jelinek  <jakub@redhat.com>
2908
2909         PR target/93412
2910         * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2911         Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2912         drop <di> from constraint of last operand.
2913
2914         PR target/93430
2915         * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2916         TARGET_AVX2 and V4DFmode not in the split condition, but in the
2917         pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2918
2919 2020-01-25  Feng Xue  <fxue@os.amperecomputing.com>
2920
2921         PR ipa/93166
2922         * ipa-cp.c (get_info_about_necessary_edges): Remove value
2923         check assertion.
2924
2925 2020-01-24  Jeff Law  <law@redhat.com>
2926
2927         PR tree-optimization/92788
2928         * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2929         not EDGE_ABNORMAL.
2930
2931 2020-01-24  Jakub Jelinek  <jakub@redhat.com>
2932
2933         PR target/93395
2934         * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2935         *avx_vperm_broadcast_<mode>,
2936         <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2937         *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2938         Move before avx2_perm<mode>/avx512f_perm<mode>.
2939
2940         PR target/93376
2941         * simplify-rtx.c (simplify_const_unary_operation,
2942         simplify_const_binary_operation): Punt for mode precision above
2943         MAX_BITSIZE_MODE_ANY_INT.
2944
2945 2020-01-24  Andrew Pinski  <apinski@marvell.com>
2946
2947         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2948         alu.shift_reg to 0.
2949
2950 2020-01-24  Jeff Law  <law@redhat.com>
2951
2952         PR target/13721
2953         * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2954         for REGs.  Call output_operand_lossage to get more reasonable
2955         diagnostics.
2956
2957 2020-01-24  Andrew Stubbs  <ams@codesourcery.com>
2958
2959         * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2960         gcn_fp_compare_operator.
2961         (vec_cmpu<mode>di): Use gcn_compare_operator.
2962         (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2963         (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2964         (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2965         (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2966         (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2967         (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2968         (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2969         gcn_fp_compare_operator.
2970         (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2971         gcn_fp_compare_operator.
2972         (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2973         gcn_fp_compare_operator.
2974         (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2975         gcn_fp_compare_operator.
2976
2977 2020-01-24  Maciej W. Rozycki  <macro@wdc.com>
2978
2979         * doc/install.texi (Cross-Compiler-Specific Options): Document
2980         `--with-toolexeclibdir' option.
2981
2982 2020-01-24  Hans-Peter Nilsson  <hp@axis.com>
2983
2984         * target.def (flags_regnum): Also mention effect on delay slot filling.
2985         * doc/tm.texi: Regenerate.
2986
2987 2020-01-23  Jeff Law  <law@redhat.com>
2988
2989         PR translation/90162
2990         * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2991
2992 2020-01-23  Mikael Tillenius  <mti-1@tillenius.com>
2993
2994         PR target/92269
2995         * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2996         profiling label
2997
2998 2020-01-23  Jakub Jelinek  <jakub@redhat.com>
2999
3000         PR rtl-optimization/93402
3001         * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
3002         USE insns.
3003
3004 2020-01-23  Dragan Mladjenovic  <dmladjenovic@wavecomp.com>
3005
3006         * config.in: Regenerated.
3007         * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
3008         for TARGET_LIBC_GNUSTACK.
3009         * configure: Regenerated.
3010         * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
3011         found to be 2.31 or greater.
3012
3013 2020-01-23  Dragan Mladjenovic  <dmladjenovic@wavecomp.com>
3014
3015         * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
3016         TARGET_SOFT_FLOAT.
3017         * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
3018         (mips_asm_file_end): New function. Delegate to
3019         file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
3020         * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3021
3022 2020-01-23  Jakub Jelinek  <jakub@redhat.com>
3023
3024         PR target/93376
3025         * config/i386/i386-modes.def (POImode): New mode.
3026         (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
3027         * config/i386/i386.md (DPWI): New mode attribute.
3028         (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
3029         (QWI): Rename to...
3030         (QPWI): ... this.  Use POI instead of OI for TImode.
3031         (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
3032         *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
3033         instead of <QWI>.
3034
3035 2020-01-23  Richard Sandiford  <richard.sandiford@arm.com>
3036
3037         PR target/93341
3038         * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
3039         unspec.
3040         (speculation_tracker_rev): New pattern.
3041         * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
3042         Use speculation_tracker_rev to track the inverse condition.
3043
3044 2020-01-23  Richard Biener  <rguenther@suse.de>
3045
3046         PR tree-optimization/93381
3047         * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
3048         alias-set of the def as argument and record the first one.
3049         (vn_walk_cb_data::first_set): New member.
3050         (vn_reference_lookup_3): Pass the alias-set of the current def
3051         to push_partial_def.  Fix alias-set used in the aggregate copy
3052         case.
3053         (vn_reference_lookup): Consistently set *last_vuse_ptr.
3054         * real.c (clear_significand_below): Fix out-of-bound access.
3055
3056 2020-01-23  Jakub Jelinek  <jakub@redhat.com>
3057
3058         PR target/93346
3059         * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
3060         New define_insn patterns.
3061
3062 2020-01-23  Richard Sandiford  <richard.sandiford@arm.com>
3063
3064         * doc/sourcebuild.texi (check-function-bodies): Add an
3065         optional target/xfail selector.
3066
3067 2020-01-23  Richard Sandiford  <richard.sandiford@arm.com>
3068
3069         PR rtl-optimization/93124
3070         * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
3071         bare USE and CLOBBER insns.
3072
3073 2020-01-22  Andrew Pinski  <apinski@marvell.com>
3074
3075         * config/arc/arc.c (output_short_suffix): Check insn for nullness.
3076
3077 2020-01-22  David Malcolm  <dmalcolm@redhat.com>
3078
3079         PR analyzer/93307
3080         * gdbinit.in (break-on-saved-diagnostic): Update for move of
3081         diagnostic_manager into "ana" namespace.
3082         * selftest-run-tests.c (selftest::run_tests): Update for move of
3083         selftest::run_analyzer_selftests to
3084         ana::selftest::run_analyzer_selftests.
3085
3086 2020-01-22  Richard Sandiford  <richard.sandiford@arm.com>
3087
3088         * cfgexpand.c (union_stack_vars): Update the size.
3089
3090 2020-01-22  Richard Biener  <rguenther@suse.de>
3091
3092         PR tree-optimization/93381
3093         * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
3094         throughout, handle all conversions the same.
3095
3096 2020-01-22  Jakub Jelinek  <jakub@redhat.com>
3097
3098         PR target/93335
3099         * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
3100         gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
3101         predicate, not whenever it is CONST_INT.  Otherwise, force_reg it.
3102         Call force_reg on high_in2 unconditionally.
3103
3104 2020-01-22  Martin Liska  <mliska@suse.cz>
3105
3106         PR tree-optimization/92924
3107         * profile.c (compute_value_histograms): Divide
3108         all counter values.
3109
3110 2020-01-22  Jakub Jelinek  <jakub@redhat.com>
3111
3112         PR target/91298
3113         * output.h (assemble_name_resolve): Declare.
3114         * varasm.c (assemble_name_resolve): New function.
3115         (assemble_name): Use it.
3116         * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
3117
3118 2020-01-22  Joseph Myers  <joseph@codesourcery.com>
3119
3120         * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
3121         update_web_docs_git instead of update_web_docs_svn.
3122
3123 2020-01-21  Andrew Pinski  <apinski@marvell.com>
3124
3125         PR target/9311
3126         * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
3127         as PTR mode. Have operand 1 as being modeless, it can be P mode.
3128         (*tlsgd_small_<mode>): Likewise.
3129         * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
3130         <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
3131         register.  Convert that register back to dest using convert_mode.
3132
3133 2020-01-21  Jim Wilson  <jimw@sifive.com>
3134
3135         * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
3136         instead of XINT.
3137
3138 2020-01-21  H.J. Lu  <hongjiu.lu@intel.com>
3139             Uros Bizjak    <ubizjak@gmail.com>
3140
3141         PR target/93319
3142         * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
3143         with ptr_mode.
3144         (legitimize_tls_address): Do GNU2 TLS address computation in
3145         ptr_mode and zero-extend result to Pmode.
3146         *  config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
3147         :P with :PTR and Pmode with ptr_mode.
3148         (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
3149         (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
3150         (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
3151
3152 2020-01-21  Jakub Jelinek  <jakub@redhat.com>
3153
3154         PR target/93333
3155         * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
3156         the last two operands are CONST_INT_P before using them as such.
3157
3158 2020-01-21  Richard Sandiford  <richard.sandiford@arm.com>
3159
3160         * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
3161         to get the integer element types.
3162
3163 2020-01-21  Richard Sandiford  <richard.sandiford@arm.com>
3164
3165         * config/aarch64/aarch64-sve-builtins.h
3166         (function_expander::convert_to_pmode): Declare.
3167         * config/aarch64/aarch64-sve-builtins.cc
3168         (function_expander::convert_to_pmode): New function.
3169         (function_expander::get_contiguous_base): Use it.
3170         (function_expander::prepare_gather_address_operands): Likewise.
3171         * config/aarch64/aarch64-sve-builtins-sve2.cc
3172         (svwhilerw_svwhilewr_impl::expand): Likewise.
3173
3174 2020-01-21  Szabolcs Nagy  <szabolcs.nagy@arm.com>
3175
3176         PR target/92424
3177         * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
3178         cfun->machine->label_is_assembled.
3179         (aarch64_print_patchable_function_entry): New.
3180         (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
3181         * config/aarch64/aarch64.h (struct machine_function): New field,
3182         label_is_assembled.
3183
3184 2020-01-21  David Malcolm  <dmalcolm@redhat.com>
3185
3186         PR ipa/93315
3187         * ipa-profile.c (ipa_profile): Delete call_sums and set it to
3188         NULL on exit.
3189
3190 2020-01-18  Jan Hubicka  <hubicka@ucw.cz>
3191
3192         PR lto/93318
3193         * cgraph.c (cgraph_edge::resolve_speculation,
3194         cgraph_edge::redirect_call_stmt_to_callee): Fix update of
3195         call_stmt_site_hash.
3196
3197 2020-01-21  Martin Liska  <mliska@suse.cz>
3198
3199         * config/rs6000/rs6000.c (common_mode_defined): Remove
3200         unused variable.
3201
3202 2020-01-21  Richard Biener  <rguenther@suse.de>
3203
3204         PR tree-optimization/92328
3205         * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
3206         type when value-numbering same-sized store by inserting a
3207         VIEW_CONVERT_EXPR.
3208         (eliminate_dom_walker::eliminate_stmt): When eliminating
3209         a redundant store handle bit-reinterpretation of the same value.
3210
3211 2020-01-21  Andrew Pinski  <apinski@marvel.com>
3212
3213         PR tree-opt/93321
3214         * tree-into-ssa.c (prepare_block_for_update_1): Split out
3215         from ...
3216         (prepare_block_for_update): This.  Use a worklist instead of
3217         recursing.
3218
3219 2020-01-21  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3220
3221         * gcc/config/arm/arm.c (clear_operation_p):
3222         Initialise last_regno, skip first iteration
3223         based on the first_set value and use ints instead
3224         of the unnecessary HOST_WIDE_INTs.
3225
3226 2020-01-21  Jakub Jelinek  <jakub@redhat.com>
3227
3228         PR target/93073
3229         * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
3230         compare_mode other than SFmode or DFmode.
3231
3232 2020-01-21  Kito Cheng  <kito.cheng@sifive.com>
3233
3234         PR target/93304
3235         * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
3236         * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
3237         * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
3238
3239 2020-01-20  Wilco Dijkstra  <wdijkstr@arm.com>
3240
3241         * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
3242
3243 2020-01-20  Andrew Pinski  <apinski@marvell.com>
3244
3245         PR middle-end/93242
3246         * targhooks.c (default_print_patchable_function_entry): Use
3247         output_asm_insn to emit the nop instruction.
3248
3249 2020-01-20  Fangrui Song  <maskray@google.com>
3250
3251         PR middle-end/93194
3252         * targhooks.c (default_print_patchable_function_entry): Align to
3253         POINTER_SIZE.
3254
3255 2020-01-20  H.J. Lu  <hongjiu.lu@intel.com>
3256
3257         PR target/93319
3258         * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
3259         gen_tls_dynamic_gnu2_64.  Compute GNU2 TLS address in ptr_mode.
3260         * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
3261         (@tls_dynamic_gnu2_64_<mode>): This.  Replace DI with P.
3262         (*tls_dynamic_gnu2_lea_64): Renamed to ...
3263         (*tls_dynamic_gnu2_lea_64_<mode>): This.  Replace DI with P.
3264         Remove the {q} suffix from lea.
3265         (*tls_dynamic_gnu2_call_64): Renamed to ...
3266         (*tls_dynamic_gnu2_call_64_<mode>): This.  Replace DI with P.
3267         (*tls_dynamic_gnu2_combine_64): Renamed to ...
3268         (*tls_dynamic_gnu2_combine_64_<mode>): This.  Replace DI with P.
3269         Pass Pmode to gen_tls_dynamic_gnu2_64.
3270
3271 2020-01-20  Wilco Dijkstra  <wdijkstr@arm.com>
3272
3273         * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
3274
3275 2020-01-20  Richard Sandiford  <richard.sandiford@arm.com>
3276
3277         * config/aarch64/aarch64-sve-builtins-base.cc
3278         (svld1ro_impl::memory_vector_mode): Remove parameter name.
3279
3280 2020-01-20  Richard Biener  <rguenther@suse.de>
3281
3282         PR debug/92763
3283         * dwarf2out.c (prune_unused_types): Unconditionally mark
3284         called function DIEs.
3285
3286 2020-01-20  Martin Liska  <mliska@suse.cz>
3287
3288         PR tree-optimization/93199
3289         * tree-eh.c (struct leh_state): Add
3290         new field outer_non_cleanup.
3291         (cleanup_is_dead_in): Pass leh_state instead
3292         of eh_region.  Add a checking that state->outer_non_cleanup
3293         points to outer non-clean up region.
3294         (lower_try_finally): Record outer_non_cleanup
3295         for this_state.
3296         (lower_catch): Likewise.
3297         (lower_eh_filter): Likewise.
3298         (lower_eh_must_not_throw): Likewise.
3299         (lower_cleanup): Likewise.
3300
3301 2020-01-20  Richard Biener  <rguenther@suse.de>
3302
3303         PR tree-optimization/93094
3304         * tree-vectorizer.h (vect_loop_versioning): Adjust.
3305         (vect_transform_loop): Likewise.
3306         * tree-vectorizer.c (try_vectorize_loop_1): Pass down
3307         loop_vectorized_call to vect_transform_loop.
3308         * tree-vect-loop.c (vect_transform_loop): Pass down
3309         loop_vectorized_call to vect_loop_versioning.
3310         * tree-vect-loop-manip.c (vect_loop_versioning): Use
3311         the earlier discovered loop_vectorized_call.
3312
3313 2020-01-19  Eric S. Raymond <esr@thyrsus.com>
3314
3315         * doc/contribute.texi: Update for SVN -> Git transition.
3316         * doc/install.texi: Likewise.
3317
3318 2020-01-18  Jan Hubicka  <hubicka@ucw.cz>
3319
3320         PR lto/93318
3321         * cgraph.c (cgraph_edge::make_speculative): Increase number of
3322         speculative targets.
3323         (verify_speculative_call): New function
3324         (cgraph_node::verify_node): Use it.
3325         * ipa-profile.c (ipa_profile): Fix formating; do not set number of
3326         speculations.
3327
3328 2020-01-18  Jan Hubicka  <hubicka@ucw.cz>
3329
3330         PR lto/93318
3331         * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
3332         (cgraph_edge::make_direct): Remove all indirect targets.
3333         (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
3334         (cgraph_node::verify_node): Verify that only one call_stmt or
3335         lto_stmt_uid is set.
3336         * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
3337         lto_stmt_uid.
3338         * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
3339         (lto_output_ref): Simplify streaming of stmt.
3340         * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
3341
3342 2020-01-18  Tamar Christina  <tamar.christina@arm.com>
3343
3344         * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
3345         Mark parameter unused.
3346
3347 2020-01-18  Hans-Peter Nilsson  <hp@axis.com>
3348
3349         * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
3350
3351 2019-01-18  Gerald Pfeifer  <gerald@pfeifer.com>
3352
3353         * varpool.c (ctor_useable_for_folding_p): Fix grammar.
3354
3355 2020-01-18  Iain Sandoe  <iain@sandoe.co.uk>
3356
3357         * Makefile.in: Add coroutine-passes.o.
3358         * builtin-types.def (BT_CONST_SIZE): New.
3359         (BT_FN_BOOL_PTR): New.
3360         (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
3361         * builtins.def (DEF_COROUTINE_BUILTIN): New.
3362         * coroutine-builtins.def: New file.
3363         * coroutine-passes.cc: New file.
3364         * function.h (struct GTY function): Add a bit to indicate that the
3365         function is a coroutine component.
3366         * internal-fn.c (expand_CO_FRAME): New.
3367         (expand_CO_YIELD): New.
3368         (expand_CO_SUSPN): New.
3369         (expand_CO_ACTOR): New.
3370         * internal-fn.def (CO_ACTOR): New.
3371         (CO_YIELD): New.
3372         (CO_SUSPN): New.
3373         (CO_FRAME): New.
3374         * passes.def: Add pass_coroutine_lower_builtins,
3375         pass_coroutine_early_expand_ifns.
3376         * tree-pass.h (make_pass_coroutine_lower_builtins): New.
3377         (make_pass_coroutine_early_expand_ifns): New.
3378         * doc/invoke.texi: Document the fcoroutines command line
3379         switch.
3380
3381 2020-01-18  Jakub Jelinek  <jakub@redhat.com>
3382
3383         * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
3384
3385         PR target/93312
3386         * config/arm/arm.c (clear_operation_p): Don't use REGNO until
3387         after checking the argument is a REG.  Don't use REGNO (reg)
3388         again to set last_regno, reuse regno variable instead.
3389
3390 2020-01-17  David Malcolm  <dmalcolm@redhat.com>
3391
3392         * doc/analyzer.texi (Limitations): Add note about NaN.
3393
3394 2020-01-17  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3395             Sudakshina Das  <sudi.das@arm.com>
3396
3397         * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
3398         and valid immediate.
3399         (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
3400         (lshrdi3): Generate thumb2_lsrl for valid immediates.
3401         * config/arm/constraints.md (Pg): New.
3402         * config/arm/predicates.md (long_shift_imm): New.
3403         (arm_reg_or_long_shift_imm): Likewise.
3404         * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
3405         (thumb2_lsll): Likewise.
3406         (thumb2_lsrl): New.
3407
3408 2020-01-17  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3409             Sudakshina Das  <sudi.das@arm.com>
3410
3411         * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
3412         (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
3413         * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
3414         register pairs for doubleword quantities for ARMv8.1M-Mainline.
3415         * config/arm/thumb2.md (thumb2_asrl): New.
3416         (thumb2_lsll): Likewise.
3417
3418 2020-01-17  Jakub Jelinek  <jakub@redhat.com>
3419
3420         * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
3421         unused variable.
3422
3423 2020-01-17  Alexander Monakov  <amonakov@ispras.ru>
3424
3425         * gdbinit.in (help-gcc-hooks): New command.
3426         (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
3427         pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
3428         documentation.
3429
3430 2020-01-17  Matthew Malcomson  <matthew.malcomson@arm.com>
3431
3432         * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
3433         correct target macro.
3434
3435 2020-01-17  Matthew Malcomson  <matthew.malcomson@arm.com>
3436
3437         * config/aarch64/aarch64-protos.h
3438         (aarch64_sve_ld1ro_operand_p): New.
3439         * config/aarch64/aarch64-sve-builtins-base.cc
3440         (class load_replicate): New.
3441         (class svld1ro_impl): New.
3442         (class svld1rq_impl): Change to inherit from load_replicate.
3443         (svld1ro): New sve intrinsic function base.
3444         * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
3445         New DEF_SVE_FUNCTION.
3446         * config/aarch64/aarch64-sve-builtins-base.h
3447         (svld1ro): New decl.
3448         * config/aarch64/aarch64-sve-builtins.cc
3449         (function_expander::add_mem_operand): Modify assert to allow
3450         OImode.
3451         * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
3452         pattern.
3453         * config/aarch64/aarch64.c
3454         (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
3455         (aarch64_sve_ld1rq_ld1ro_operand_p): This.
3456         (aarch64_sve_ld1ro_operand_p): New.
3457         * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
3458         * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
3459         * config/aarch64/predicates.md
3460         (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
3461
3462 2020-01-17  Matthew Malcomson  <matthew.malcomson@arm.com>
3463
3464         * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
3465         Introduce this ACLE specified predefined macro.
3466         * config/aarch64/aarch64-option-extensions.def (f64mm): New.
3467         (fp): Disabling this disables f64mm.
3468         (simd): Disabling this disables f64mm.
3469         (fp16): Disabling this disables f64mm.
3470         (sve): Disabling this disables f64mm.
3471         * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
3472         (AARCH64_ISA_F64MM): New.
3473         (TARGET_F64MM): New.
3474         * doc/invoke.texi (f64mm): Document new option.
3475
3476 2020-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
3477
3478         * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
3479         (neoversen1_tunings): Likewise.
3480
3481 2020-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
3482
3483         PR target/92692
3484         * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
3485         Add assert to ensure prolog has been emitted.
3486         (aarch64_split_atomic_op): Likewise.
3487         * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
3488         Use epilogue_completed rather than reload_completed.
3489         (aarch64_atomic_exchange<mode>): Likewise.
3490         (aarch64_atomic_<atomic_optab><mode>): Likewise.
3491         (atomic_nand<mode>): Likewise.
3492         (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
3493         (atomic_fetch_nand<mode>): Likewise.
3494         (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
3495         (atomic_nand_fetch<mode>): Likewise.
3496
3497 2020-01-17  Richard Sandiford  <richard.sandiford@arm.com>
3498
3499         PR target/93133
3500         * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
3501         for FP modes.
3502         (REVERSE_CONDITION): Delete.
3503         * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
3504         (CCFP_CCFPE): Likewise.
3505         (e): New mode attribute.
3506         * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
3507         (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
3508         (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
3509         (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
3510         (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
3511         (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
3512         * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
3513         name of generator from gen_ccmpdi to gen_ccmpccdi.
3514         (aarch64_gen_ccmp_next): Use code_for_ccmp.  If we want to reverse
3515         the previous comparison but aren't able to, use the new ccmp_rev
3516         patterns instead.
3517
3518 2020-01-17  Richard Sandiford  <richard.sandiford@arm.com>
3519
3520         * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
3521         than testing directly for INTEGER_CST.
3522         (gimplify_target_expr, gimplify_omp_depend): Likewise.
3523
3524 2020-01-17  Jakub Jelinek  <jakub@redhat.com>
3525
3526         PR tree-optimization/93292
3527         * tree-vect-stmts.c (vectorizable_comparison): Punt also if
3528         get_vectype_for_scalar_type returns NULL.
3529
3530 2020-01-16  Jan Hubicka  <hubicka@ucw.cz>
3531
3532         * params.opt (-param=max-predicted-iterations): Increase range from 0.
3533         * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
3534
3535 2020-01-16  Jan Hubicka  <hubicka@ucw.cz>
3536
3537         * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
3538         dump.
3539         * params.opt: (max-predicted-iterations): Set bounds.
3540         * predict.c (real_almost_one, real_br_prob_base,
3541         real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
3542         (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
3543         probabilities; do not truncate to reg_br_prob_bases.
3544         (estimate_loops_at_level): Pass max_cyclic_prob.
3545         (estimate_loops): Compute max_cyclic_prob.
3546         (estimate_bb_frequencies): Do not initialize real_*; update calculation
3547         of back edge prob.
3548         * profile-count.c (profile_probability::to_sreal): New.
3549         * profile-count.h (class sreal): Move up in file.
3550         (profile_probability::to_sreal): Declare.
3551
3552 2020-01-16  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
3553
3554         * config/arm/arm.c
3555         (arm_invalid_conversion): New function for target hook.
3556         (arm_invalid_unary_op): New function for target hook.
3557         (arm_invalid_binary_op): New function for target hook.
3558
3559 2020-01-16  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
3560
3561         * config.gcc: Add arm_bf16.h.
3562         * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
3563         (arm_simd_builtin_std_type): Add BFmode.
3564         (arm_init_simd_builtin_types): Define element types for vector types.
3565         (arm_init_bf16_types): New function.
3566         (arm_init_builtins): Add arm_init_bf16_types function call.
3567         * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
3568         * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
3569         * config/arm/arm.c (aapcs_vfp_sub_candidate):  Add BFmode.
3570         (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
3571         (arm_vector_mode_supported_p): Add V4BF, V8BF.
3572         (arm_mangle_type):  Add __bf16.
3573         * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
3574         VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
3575         arm_bf16_ptr_type_node.
3576         * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
3577         define_split between ARM registers.
3578         * config/arm/arm_bf16.h: New file.
3579         * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3580         * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
3581         (VQXMOV): Add V8BF.
3582         * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
3583         * config/arm/vfp.md: Add BFmode to movhf patterns.
3584
3585 2020-01-16  Mihail Ionescu  <mihail.ionescu@arm.com>
3586             Andre Vieira  <andre.simoesdiasvieira@arm.com>
3587
3588         * config/arm/arm-cpus.in (mve, mve_float): New features.
3589         (dsp, mve, mve.fp): New options.
3590         * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
3591         * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
3592         * doc/invoke.texi: Document the armv8.1-m mve and dps options.
3593
3594 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3595             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3596
3597         * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
3598         Armv8-M Mainline.
3599         * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
3600         error for using -mcmse when targeting Armv8.1-M Mainline.
3601
3602 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3603             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3604
3605         * config/arm/arm.md (nonsecure_call_internal): Do not force memory
3606         address in r4 when targeting Armv8.1-M Mainline.
3607         (nonsecure_call_value_internal): Likewise.
3608         * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
3609         a register match_operand again.  Emit BLXNS when targeting
3610         Armv8.1-M Mainline.
3611         (nonsecure_call_value_reg_thumb2): Likewise.
3612
3613 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3614             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3615
3616         * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
3617         (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
3618         variable as true when floating-point ABI is not hard.  Replace
3619         check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
3620         Generate VLSTM and VLLDM instruction respectively before and
3621         after a function call to cmse_nonsecure_call function.
3622         * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
3623         (VUNSPEC_VLLDM): Likewise.
3624         * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
3625         (lazy_load_multiple_insn): Likewise.
3626
3627 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3628             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3629
3630         * config/arm/arm.c (vfp_emit_fstmd): Declare early.
3631         (arm_emit_vfp_multi_reg_pop): Likewise.
3632         (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
3633         registers to clear in max_fp_regno.  Emit VPUSH and VPOP to save and
3634         restore callee-saved VFP registers.
3635
3636 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3637             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3638
3639         * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
3640         (cmse_nonsecure_call_clear_caller_saved): Rename into ...
3641         (cmse_nonsecure_call_inline_register_clear): This.  Save and clear
3642         callee-saved GPRs as well as clear ip register before doing a nonsecure
3643         call then restore callee-saved GPRs after it when targeting
3644         Armv8.1-M Mainline.
3645         (arm_reorg): Adapt to function rename.
3646
3647 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3648             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3649
3650         * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
3651         * config/arm/arm.c (clear_operation_p): Extend to be able to check a
3652         clear_vfp_multiple pattern based on a new vfp parameter.
3653         (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
3654         targeting Armv8.1-M Mainline.
3655         (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
3656         unconditionally when targeting Armv8.1-M Mainline architecture.  Check
3657         whether VFP registers are available before looking call_used_regs for a
3658         VFP register.
3659         * config/arm/predicates.md (clear_multiple_operation): Adapt to change
3660         of prototype of clear_operation_p.
3661         (clear_vfp_multiple_operation): New predicate.
3662         * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
3663         * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
3664
3665 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3666             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3667
3668         * config/arm/arm-protos.h (clear_operation_p): Declare.
3669         * config/arm/arm.c (clear_operation_p): New function.
3670         (cmse_clear_registers): Generate clear_multiple instruction pattern if
3671         targeting Armv8.1-M Mainline or successor.
3672         (output_return_instruction): Only output APSR register clearing if
3673         Armv8.1-M Mainline instructions not available.
3674         (thumb_exit): Likewise.
3675         * config/arm/predicates.md (clear_multiple_operation): New predicate.
3676         * config/arm/thumb2.md (clear_apsr): New define_insn.
3677         (clear_multiple): Likewise.
3678         * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
3679
3680 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3681             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3682
3683         * config/arm/arm.c (fp_sysreg_names): Declare and define.
3684         (use_return_insn): Also return false for Armv8.1-M Mainline.
3685         (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
3686         Mainline instructions are available.
3687         (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
3688         when targeting Armv8.1-M Mainline Security Extensions.
3689         (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
3690         Mainline entry function.
3691         (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
3692         targeting Armv8.1-M Mainline or successor.
3693         (arm_expand_epilogue): Fix indentation of caller-saved register
3694         clearing.  Restore FPCXTNS if this is an Armv8.1-M Mainline
3695         entry function.
3696         * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
3697         (FP_SYSREGS): Likewise.
3698         (enum vfp_sysregs_encoding): Define enum.
3699         (fp_sysreg_names): Declare.
3700         * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
3701         * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
3702         (pop_fpsysreg_insn): Likewise.
3703
3704 2020-01-16  Mihail-Calin Ionescu  <mihail.ionescu@arm.com>
3705             Thomas Preud'homme  <thomas.preudhomme@arm.com>
3706
3707         * config/arm/arm-cpus.in (armv8_1m_main): New feature.
3708         (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
3709         ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
3710         ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
3711         ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
3712         (ARMv8_1m_main): New feature group.
3713         (armv8.1-m.main): New architecture.
3714         * config/arm/arm-tables.opt: Regenerate.
3715         * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
3716         (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
3717         (arm_options_perform_arch_sanity_checks): Error out when targeting
3718         Armv8.1-M Mainline Security Extensions.
3719         * config/arm/arm.h (arm_arch8_1m_main): Declare.
3720
3721 2020-01-16  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
3722
3723         * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
3724         aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
3725         * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
3726         aarch64_bfdot_laneq): New.
3727         * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
3728         vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
3729         vbfdotq_laneq_f32): New.
3730         * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
3731         VBFMLA_W, VBF): New.
3732         (isquadop): Add V4BF, V8BF.
3733
3734 2020-01-16  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
3735
3736         * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
3737         New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
3738         TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
3739         (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
3740         (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
3741         * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
3742         usdot_laneq, sudot_lane,sudot_laneq): New.
3743         * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
3744         (aarch64_<sur>dot_lane): New.
3745         * config/aarch64/arm_neon.h (vusdot_s32): New.
3746         (vusdotq_s32): New.
3747         (vusdot_lane_s32): New.
3748         (vsudot_lane_s32): New.
3749         * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
3750         (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
3751
3752 2020-01-16  Martin Liska  <mliska@suse.cz>
3753
3754         * value-prof.c (dump_histogram_value): Fix
3755         obvious spacing issue.
3756
3757 2020-01-16  Andrew Pinski  <apinski@marvell.com>
3758
3759         * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
3760         !storage_order_barrier_p.
3761
3762 2020-01-16  Andrew Pinski  <apinski@marvell.com>
3763
3764         * sched-int.h (_dep): Add unused bit-field field for the padding.
3765         * sched-deps.c (init_dep_1): Init unused field.
3766
3767 2020-01-16  Andrew Pinski  <apinski@marvell.com>
3768
3769         * optabs.h (create_expand_operand): Initialize target field also.
3770
3771 2020-01-16  Andre Vieira  <andre.simoesdiasvieira@arm.com>
3772
3773         PR tree-optimization/92429
3774         * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
3775         * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
3776         control folding.
3777         * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
3778         tree.
3779
3780 2020-01-16  Richard Sandiford  <richard.sandiford@arm.com>
3781
3782         * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
3783         aarch64_sve_int_mode to each mode.
3784
3785 2020-01-15  David Malcolm  <dmalcolm@redhat.com>
3786
3787         * doc/analyzer.texi (Overview): Add note about
3788         -fdump-ipa-analyzer.
3789
3790 2020-01-15  Wilco Dijkstra  <wdijkstr@arm.com>
3791
3792         PR tree-optimization/93231
3793         * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
3794         input_type is unsigned.  Use tree_to_shwi for shift constant.
3795         Check CST_STRING element size is CHAR_TYPE_SIZE bits.
3796         (simplify_count_trailing_zeroes): Add test to handle known non-zero
3797         inputs more efficiently.
3798
3799 2020-01-15  Uroš Bizjak  <ubizjak@gmail.com>
3800
3801         * config/i386/i386.md (*movsf_internal): Do not require
3802         SSE2 ISA for alternatives 14 and 15.
3803
3804 2020-01-15  Richard Biener  <rguenther@suse.de>
3805
3806         PR middle-end/93273
3807         * tree-eh.c (sink_clobbers): If we already visited the destination
3808         block do not defer insertion.
3809         (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
3810         the purpose of defered insertion.
3811
3812 2020-01-15  Jakub Jelinek  <jakub@redhat.com>
3813
3814         * BASE-VER: Bump to 10.0.1.
3815
3816 2020-01-15  Richard Sandiford  <richard.sandiford@arm.com>
3817
3818         PR tree-optimization/93247
3819         * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
3820         type of the stmt that we're going to vectorize.
3821
3822 2020-01-15  Richard Sandiford  <richard.sandiford@arm.com>
3823
3824         * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
3825         VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3826         type from the lhs.
3827
3828 2020-01-15  Martin Liska  <mliska@suse.cz>
3829
3830         * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3831         2 calls of streamer_read_hwi in a function call.
3832
3833 2020-01-15  Richard Biener  <rguenther@suse.de>
3834
3835         * alias.c (record_alias_subset): Avoid redundant work when
3836         subset is already recorded.
3837
3838 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
3839
3840         * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3841         the analyzer options provide CWE identifiers.
3842
3843 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
3844
3845         * tree-diagnostic-path.cc (path_summary::event_range::print):
3846         When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3847         using get_pure_location.
3848
3849 2020-01-15  Jakub Jelinek  <jakub@redhat.com>
3850
3851         PR tree-optimization/93262
3852         * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3853         perform head trimming only if the last argument is constant,
3854         either all ones, or larger or equal to head trim, in the latter
3855         case decrease the last argument by head_trim.
3856
3857         PR tree-optimization/93249
3858         * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3859         (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3860         function body scope, reindent.  For BUILTIN_IN_STRNCPY*, don't
3861         perform head trim unless we can prove there are no '\0' chars
3862         from the source among the first head_trim chars.
3863
3864 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
3865
3866         * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3867
3868 2020-01-15  Jakub Jelinek  <jakub@redhat.com>
3869
3870         PR target/93009
3871         * config/i386/sse.md
3872         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3873         *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3874         *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3875         *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3876         just a single alternative instead of two, make operands 1 and 2
3877         commutative.
3878
3879 2020-01-14  Jan Hubicka  <hubicka@ucw.cz>
3880
3881         PR lto/91576
3882         * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3883         TYPE_MODE.
3884
3885 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
3886
3887         * Makefile.in (lang_opt_files): Add analyzer.opt.
3888         (ANALYZER_OBJS): New.
3889         (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3890         tristate.o and ANALYZER_OBJS.
3891         (TEXI_GCCINT_FILES): Add analyzer.texi.
3892         * common.opt (-fanalyzer): New driver option.
3893         * config.in: Regenerate.
3894         * configure: Regenerate.
3895         * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3896         (gccdepdir): Also create depdir for "analyzer" subdir.
3897         * digraph.cc: New file.
3898         * digraph.h: New file.
3899         * doc/analyzer.texi: New file.
3900         * doc/gccint.texi ("Static Analyzer") New menu item.
3901         (analyzer.texi): Include it.
3902         * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3903         ("Warning Options"): Add static analysis warnings to the list.
3904         (-Wno-analyzer-double-fclose): New option.
3905         (-Wno-analyzer-double-free): New option.
3906         (-Wno-analyzer-exposure-through-output-file): New option.
3907         (-Wno-analyzer-file-leak): New option.
3908         (-Wno-analyzer-free-of-non-heap): New option.
3909         (-Wno-analyzer-malloc-leak): New option.
3910         (-Wno-analyzer-possible-null-argument): New option.
3911         (-Wno-analyzer-possible-null-dereference): New option.
3912         (-Wno-analyzer-null-argument): New option.
3913         (-Wno-analyzer-null-dereference): New option.
3914         (-Wno-analyzer-stale-setjmp-buffer): New option.
3915         (-Wno-analyzer-tainted-array-index): New option.
3916         (-Wno-analyzer-use-after-free): New option.
3917         (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3918         (-Wno-analyzer-use-of-uninitialized-value): New option.
3919         (-Wanalyzer-too-complex): New option.
3920         (-fanalyzer-call-summaries): New warning.
3921         (-fanalyzer-checker=): New warning.
3922         (-fanalyzer-fine-grained): New warning.
3923         (-fno-analyzer-state-merge): New warning.
3924         (-fno-analyzer-state-purge): New warning.
3925         (-fanalyzer-transitivity): New warning.
3926         (-fanalyzer-verbose-edges): New warning.
3927         (-fanalyzer-verbose-state-changes): New warning.
3928         (-fanalyzer-verbosity=): New warning.
3929         (-fdump-analyzer): New warning.
3930         (-fdump-analyzer-callgraph): New warning.
3931         (-fdump-analyzer-exploded-graph): New warning.
3932         (-fdump-analyzer-exploded-nodes): New warning.
3933         (-fdump-analyzer-exploded-nodes-2): New warning.
3934         (-fdump-analyzer-exploded-nodes-3): New warning.
3935         (-fdump-analyzer-supergraph): New warning.
3936         * doc/sourcebuild.texi (dg-require-dot): New.
3937         (dg-check-dot): New.
3938         * gdbinit.in (break-on-saved-diagnostic): New command.
3939         * graphviz.cc: New file.
3940         * graphviz.h: New file.
3941         * ordered-hash-map-tests.cc: New file.
3942         * ordered-hash-map.h: New file.
3943         * passes.def (pass_analyzer): Add before
3944         pass_ipa_whole_program_visibility.
3945         * selftest-run-tests.c (selftest::run_tests): Call
3946         selftest::ordered_hash_map_tests_cc_tests.
3947         * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3948         decl.
3949         * shortest-paths.h: New file.
3950         * timevar.def (TV_ANALYZER): New timevar.
3951         (TV_ANALYZER_SUPERGRAPH): Likewise.
3952         (TV_ANALYZER_STATE_PURGE): Likewise.
3953         (TV_ANALYZER_PLAN): Likewise.
3954         (TV_ANALYZER_SCC): Likewise.
3955         (TV_ANALYZER_WORKLIST): Likewise.
3956         (TV_ANALYZER_DUMP): Likewise.
3957         (TV_ANALYZER_DIAGNOSTICS): Likewise.
3958         (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3959         * tree-pass.h (make_pass_analyzer): New decl.
3960         * tristate.cc: New file.
3961         * tristate.h: New file.
3962
3963 2020-01-14  Uroš Bizjak  <ubizjak@gmail.com>
3964
3965         PR target/93254
3966         * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3967         alternatives 9 and 10.
3968
3969 2020-01-14  David Malcolm  <dmalcolm@redhat.com>
3970
3971         * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3972         * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3973         * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3974         * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3975         (selftest::hash_map_tests_c_tests): Call it.
3976         * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3977         New static constant, using the value of = H::empty_zero_p.
3978         (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3979         from default_hash_traits <Value>.
3980         * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3981         from Traits.
3982         * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3983         * hash-table.h (hash_table::alloc_entries): Guard the loop of
3984         calls to mark_empty with !Descriptor::empty_zero_p.
3985         (hash_table::empty_slow): Conditionalize the memset call with a
3986         check that Descriptor::empty_zero_p; otherwise, loop through the
3987         entries calling mark_empty on them.
3988         * hash-traits.h (int_hash::empty_zero_p): New static constant.
3989         (pointer_hash::empty_zero_p): Likewise.
3990         (pair_hash::empty_zero_p): Likewise.
3991         * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3992         Likewise.
3993         * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3994         (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3995         * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3996         * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3997         (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3998         * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3999         * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
4000         * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
4001         * tree-vectorizer.h
4002         (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
4003         Likewise.
4004
4005 2020-01-14  Kewen Lin  <linkw@gcc.gnu.org>
4006
4007         * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
4008         fix typo on return value.
4009
4010 2020-01-14  Xiong Hu Luo  <luoxhu@linux.ibm.com>
4011
4012         PR ipa/69678
4013         * cgraph.c (symbol_table::create_edge): Init speculative_id and
4014         target_prob.
4015         (cgraph_edge::make_speculative): Add param for setting speculative_id
4016         and target_prob.
4017         (cgraph_edge::speculative_call_info): Update comments and find reference
4018         by speculative_id for multiple indirect targets.
4019         (cgraph_edge::resolve_speculation): Decrease the speculations
4020         for indirect edge, drop it's speculative if not direct target
4021         left. Update comments.
4022         (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
4023         (cgraph_node::dump): Print num_speculative_call_targets.
4024         (cgraph_node::verify_node): Don't report error if speculative
4025         edge not include statement.
4026         (cgraph_edge::num_speculative_call_targets_p): New function.
4027         * cgraph.h (int common_target_id): Remove.
4028         (int common_target_probability): Remove.
4029         (num_speculative_call_targets): New variable.
4030         (make_speculative): Add param for setting speculative_id.
4031         (cgraph_edge::num_speculative_call_targets_p): New declare.
4032         (target_prob): New variable.
4033         (speculative_id): New variable.
4034         * ipa-fnsummary.c (analyze_function_body): Create and duplicate
4035           call summaries for multiple speculative call targets.
4036         * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
4037         * ipa-profile.c (struct speculative_call_target): New struct.
4038         (class speculative_call_summary): New class.
4039         (class speculative_call_summaries): New class.
4040         (call_sums): New variable.
4041         (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
4042         (ipa_profile_write_edge_summary): New function.
4043         (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
4044         (ipa_profile_dump_all_summaries): New function.
4045         (ipa_profile_read_edge_summary): New function.
4046         (ipa_profile_read_summary_section): New function.
4047         (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
4048         (ipa_profile): Generate num_speculative_call_targets from
4049         profile summaries.
4050         * ipa-ref.h (speculative_id): New variable.
4051         * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
4052         * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
4053         common_target_probability.   Stream out speculative_id and
4054         num_speculative_call_targets.
4055         (input_edge): Likewise.
4056         * predict.c (dump_prediction): Remove edges count assert to be
4057         precise.
4058         * symtab.c (symtab_node::create_reference): Init speculative_id.
4059         (symtab_node::clone_references): Clone speculative_id.
4060         (symtab_node::clone_referring): Clone speculative_id.
4061         (symtab_node::clone_reference): Clone speculative_id.
4062         (symtab_node::clear_stmts_in_references): Clear speculative_id.
4063         * tree-inline.c (copy_bb): Duplicate all the speculative edges
4064         if indirect call contains multiple speculative targets.
4065         * value-prof.h  (check_ic_target): Remove.
4066         * value-prof.c  (gimple_value_profile_transformations):
4067         Use void function gimple_ic_transform.
4068         * value-prof.c  (gimple_ic_transform): Handle topn case.
4069         Fix comment typos.  Change it to a void function.
4070
4071 2020-01-13  Andrew Pinski  <apinski@marvell.com>
4072
4073         * config/aarch64/aarch64-cores.def (octeontx2): New define.
4074         (octeontx2t98): New define.
4075         (octeontx2t96): New define.
4076         (octeontx2t93): New define.
4077         (octeontx2f95): New define.
4078         (octeontx2f95n): New define.
4079         (octeontx2f95mm): New define.
4080         * config/aarch64/aarch64-tune.md: Regenerate.
4081         * doc/invoke.texi (-mcpu=): Document the new cpu types.
4082
4083 2020-01-13  Jason Merrill  <jason@redhat.com>
4084
4085         PR c++/33799 - destroy return value if local cleanup throws.
4086         * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
4087
4088 2020-01-13  Martin Liska  <mliska@suse.cz>
4089
4090         * ipa-cp.c (get_max_overall_size): Use newly
4091         renamed param param_ipa_cp_unit_growth.
4092         * params.opt: Remove legacy param name.
4093
4094 2020-01-13  Martin Sebor  <msebor@redhat.com>
4095
4096         PR tree-optimization/93213
4097         * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
4098         stores to be eliminated.
4099
4100 2020-01-13  Martin Liska  <mliska@suse.cz>
4101
4102         * opts.c (print_help): Do not print CL_PARAM
4103         and CL_WARNING for CL_OPTIMIZATION.
4104
4105 2020-01-13  Jonathan Wakely  <jwakely@redhat.com>
4106
4107         PR driver/92757
4108         * doc/invoke.texi (Warning Options): Add caveat about some warnings
4109         depending on optimization settings.
4110
4111 2020-01-13  Jakub Jelinek  <jakub@redhat.com>
4112
4113         PR tree-optimization/90838
4114         * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
4115         SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
4116         argument rather than to initialize temporary for targets that
4117         don't use the mode argument at all.  Initialize ctzval to avoid
4118         warning at -O0.
4119
4120 2020-01-10  Thomas Schwinge  <thomas@codesourcery.com>
4121
4122         * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
4123         * tree-core.h: Document it.
4124         * gimplify.c (gimplify_omp_workshare): Set it.
4125         * omp-low.c (lower_omp_target): Use it.
4126         * tree-pretty-print.c (dump_omp_clause): Print it.
4127
4128         * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
4129         Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
4130
4131 2020-01-10  David Malcolm  <dmalcolm@redhat.com>
4132
4133         * Makefile.in (OBJS): Add tree-diagnostic-path.o.
4134         * common.opt (fdiagnostics-path-format=): New option.
4135         (diagnostic_path_format): New enum.
4136         (fdiagnostics-show-path-depths): New option.
4137         * coretypes.h (diagnostic_event_id_t): New forward decl.
4138         * diagnostic-color.c (color_dict): Add "path".
4139         * diagnostic-event-id.h: New file.
4140         * diagnostic-format-json.cc (json_from_expanded_location): Make
4141         non-static.
4142         (json_end_diagnostic): Call context->make_json_for_path if it
4143         exists and the diagnostic has a path.
4144         (diagnostic_output_format_init): Clear context->print_path.
4145         * diagnostic-path.h: New file.
4146         * diagnostic-show-locus.c (colorizer::set_range): Special-case
4147         when printing a run of events in a diagnostic_path so that they
4148         all get the same color.
4149         (layout::m_diagnostic_path_p): New field.
4150         (layout::layout): Initialize it.
4151         (layout::print_any_labels): Don't colorize the label text for an
4152         event in a diagnostic_path.
4153         (gcc_rich_location::add_location_if_nearby): Add
4154         "restrict_to_current_line_spans" and "label" params.  Pass the
4155         former to layout.maybe_add_location_range; pass the latter
4156         when calling add_range.
4157         * diagnostic.c: Include "diagnostic-path.h".
4158         (diagnostic_initialize): Initialize context->path_format and
4159         context->show_path_depths.
4160         (diagnostic_show_any_path): New function.
4161         (diagnostic_path::interprocedural_p): New function.
4162         (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
4163         (simple_diagnostic_path::num_events): New function.
4164         (simple_diagnostic_path::get_event): New function.
4165         (simple_diagnostic_path::add_event): New function.
4166         (simple_diagnostic_event::simple_diagnostic_event): New ctor.
4167         (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
4168         (debug): New overload taking a diagnostic_path *.
4169         * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
4170         * diagnostic.h (enum diagnostic_path_format): New enum.
4171         (json::value): New forward decl.
4172         (diagnostic_context::path_format): New field.
4173         (diagnostic_context::show_path_depths): New field.
4174         (diagnostic_context::print_path): New callback field.
4175         (diagnostic_context::make_json_for_path): New callback field.
4176         (diagnostic_show_any_path): New decl.
4177         (json_from_expanded_location): New decl.
4178         * doc/invoke.texi (-fdiagnostics-path-format=): New option.
4179         (-fdiagnostics-show-path-depths): New option.
4180         (-fdiagnostics-color): Add "path" to description of default
4181         GCC_COLORS; describe it.
4182         (-fdiagnostics-format=json): Document how diagnostic paths are
4183         represented in the JSON output format.
4184         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
4185         Add optional params "restrict_to_current_line_spans" and "label".
4186         * opts.c (common_handle_option): Handle
4187         OPT_fdiagnostics_path_format_ and
4188         OPT_fdiagnostics_show_path_depths.
4189         * pretty-print.c: Include "diagnostic-event-id.h".
4190         (pp_format): Implement "%@" format code for printing
4191         diagnostic_event_id_t *.
4192         (selftest::test_pp_format): Add tests for "%@".
4193         * selftest-run-tests.c (selftest::run_tests): Call
4194         selftest::tree_diagnostic_path_cc_tests.
4195         * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
4196         * toplev.c (general_init): Initialize global_dc->path_format and
4197         global_dc->show_path_depths.
4198         * tree-diagnostic-path.cc: New file.
4199         * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
4200         non-static.  Drop "diagnostic" param in favor of storing the
4201         original value of "where" and re-using it.
4202         (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
4203         maybe_unwind_expanded_macro_loc.
4204         (tree_diagnostics_defaults): Initialize context->print_path and
4205         context->make_json_for_path.
4206         * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
4207         decl.
4208         (default_tree_make_json_for_path): New decl.
4209         (maybe_unwind_expanded_macro_loc): New decl.
4210
4211 2020-01-10  Jakub Jelinek  <jakub@redhat.com>
4212
4213         PR tree-optimization/93210
4214         * fold-const.h (native_encode_initializer,
4215         can_native_interpret_type_p): Declare.
4216         * fold-const.c (native_encode_string): Fix up handling with off != -1,
4217         simplify.
4218         (native_encode_initializer): New function, moved from dwarf2out.c.
4219         Adjust to native_encode_expr compatible arguments, including dry-run
4220         and partial extraction modes.  Don't handle STRING_CST.
4221         (can_native_interpret_type_p): No longer static.
4222         * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
4223         offset / BITS_PER_UNIT fits into int and don't call it if
4224         can_native_interpret_type_p fails.  If suboff is NULL and for
4225         CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
4226         native_encode_initializer.
4227         (fold_const_aggregate_ref_1): Formatting fix.
4228         * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
4229         (tree_add_const_value_attribute): Adjust caller.
4230
4231         PR tree-optimization/90838
4232         * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
4233         SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
4234         CTZ_DEFINED_VALUE_AT_ZERO.
4235
4236 2020-01-10  Vladimir Makarov  <vmakarov@redhat.com>
4237
4238         PR inline-asm/93027
4239         * lra-constraints.c (match_reload): Permit input operands have the
4240         same mode as output while other input operands have a different
4241         mode.
4242
4243 2020-01-10  Wilco Dijkstra  <wdijkstr@arm.com>
4244
4245         PR tree-optimization/90838
4246         * tree-ssa-forwprop.c (check_ctz_array): Add new function.
4247         (check_ctz_string): Likewise.
4248         (optimize_count_trailing_zeroes): Likewise.
4249         (simplify_count_trailing_zeroes): Likewise.
4250         (pass_forwprop::execute): Try ctz simplification.
4251         * match.pd: Add matching for ctz idioms.
4252
4253 2020-01-10  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
4254
4255         * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
4256         for target hook.
4257         (aarch64_invalid_unary_op): New function for target hook.
4258         (aarch64_invalid_binary_op): New function for target hook.
4259
4260 2020-01-10  Stam Markianos-Wright  <stam.markianos-wright@arm.com>
4261
4262         * config.gcc: Add arm_bf16.h.
4263         * config/aarch64/aarch64-builtins.c
4264         (aarch64_simd_builtin_std_type): Add BFmode.
4265         (aarch64_init_simd_builtin_types): Define element types for vector
4266         types.
4267         (aarch64_init_bf16_types): New function.
4268         (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
4269         * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
4270         modes.
4271         * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
4272         * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
4273         patterns.
4274         * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
4275         (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
4276         * config/aarch64/aarch64.c
4277         (aarch64_classify_vector_mode): Add support for BF types.
4278         (aarch64_gimplify_va_arg_expr): Add support for BF types.
4279         (aarch64_vq_mode): Add support for BF types.
4280         (aarch64_simd_container_mode): Add support for BF types.
4281         (aarch64_mangle_type): Add support for BF scalar type.
4282         * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
4283         * config/aarch64/arm_bf16.h: New file.
4284         * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
4285         * config/aarch64/iterators.md: Add BF types to mode attributes.
4286         (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
4287
4288 2020-01-10  Jason Merrill  <jason@redhat.com>
4289
4290         PR c++/93173 - incorrect tree sharing.
4291         * gimplify.c (copy_if_shared): No longer static.
4292         * gimplify.h: Declare it.
4293
4294 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
4295
4296         * doc/invoke.texi (-msve-vector-bits=): Document that
4297         -msve-vector-bits=128 now generates VL-specific code for
4298         little-endian targets.
4299         * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
4300         build_vector_type_for_mode to construct the data vector types.
4301         * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
4302         VL-specific code for -msve-vector-bits=128 on little-endian targets.
4303         (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
4304         for 128-bit vectors.
4305
4306 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
4307
4308         * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
4309         invocation.
4310
4311 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
4312
4313         * config/aarch64/aarch64-builtins.c
4314         (aarch64_builtin_vectorized_function): Check for specific vector modes,
4315         rather than checking the number of elements and the element mode.
4316
4317 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
4318
4319         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
4320         get_related_vectype_for_scalar_type rather than build_vector_type
4321         to create the index type for a conditional reduction.
4322
4323 2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>
4324
4325         * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
4326         for any type of gather or scatter, including strided accesses.
4327
4328 2020-01-10  Andre Vieira  <andre.simoesdiasvieira@arm.com>
4329
4330         * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
4331          comment.
4332
4333 2020-01-10  Andre Vieira  <andre.simoesdiasvieira@arm.com>
4334
4335         * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
4336         get_dr_vinfo_offset
4337         * tree-vect-loop.c (update_epilogue_loop_vinfo):  Remove orig_drs_init
4338         parameter and its use to reset DR_OFFSET's.
4339         (vect_transform_loop): Remove orig_drs_init argument.
4340         * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
4341         member of dr_vec_info rather than the offset of the associated
4342         data_reference's innermost_loop_behavior.
4343         (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
4344         (vect_do_peeling): Remove orig_drs_init parameter and its construction.
4345         * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
4346         get_dr_vinfo_offset.
4347         (vectorizable_store): Likewise.
4348         (vectorizable_load): Likewise.
4349
4350 2020-01-10  Richard Biener  <rguenther@suse.de>
4351
4352         * gimple-ssa-store-merging
4353         (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
4354
4355 2020-01-10  Martin Liska  <mliska@suse.cz>
4356
4357         PR ipa/93217
4358         * ipa-inline-analysis.c (offline_size): Make proper parenthesis
4359         encapsulation that was there before r280040.
4360
4361 2020-01-10  Richard Biener  <rguenther@suse.de>
4362
4363         PR middle-end/93199
4364         * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
4365         sequences to avoid walking them again for secondary opportunities.
4366         (pass_lower_eh_dispatch::execute): Instead actually insert
4367         them here.
4368
4369 2020-01-10  Richard Biener  <rguenther@suse.de>
4370
4371         PR middle-end/93199
4372         * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
4373         (cleanup_all_empty_eh): Walk landing pads in reverse order to
4374         avoid quadraticness.
4375
4376 2020-01-10  Martin Jambor  <mjambor@suse.cz>
4377
4378         * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
4379         * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
4380         to get param_ipa_sra_max_replacements.
4381         (param_splitting_across_edge): Pass the caller to
4382         pull_accesses_from_callee.
4383
4384 2020-01-10  Martin Jambor  <mjambor@suse.cz>
4385
4386         * params.opt (param_ipcp_unit_growth): Mark as Optimization.
4387         * ipa-cp.c (max_new_size): Removed.
4388         (orig_overall_size): New variable.
4389         (get_max_overall_size): New function.
4390         (estimate_local_effects): Use it.  Adjust dump.
4391         (decide_about_value): Likewise.
4392         (ipcp_propagate_stage): Do not calculate max_new_size, just store
4393         orig_overall_size.  Adjust dump.
4394         (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
4395
4396 2020-01-10  Martin Jambor  <mjambor@suse.cz>
4397
4398         * params.opt (param_ipa_max_agg_items): Mark as Optimization
4399         * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
4400         instead of param_ipa_max_agg_items.
4401         (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
4402         optimization info for the callee.
4403
4404 2020-01-09  Kwok Cheung Yeung  <kcy@codesourcery.com>
4405
4406         * lto-streamer-in.c (input_function): Remove streamed-in inline debug
4407         markers if debug_inline_points is false.
4408
4409 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4410
4411         * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
4412         extra_objs.
4413         * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
4414         aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
4415         aarch64-sve-builtins-sve2.h.
4416         (aarch64-sve-builtins-sve2.o): New rule.
4417         * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
4418         (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
4419         (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
4420         (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
4421         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4422         TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
4423         TARGET_SVE2_SM4.
4424         * config/aarch64/aarch64-sve.md: Update comments with SVE2
4425         instructions that are handled here.
4426         (@cond_asrd<mode>): Generalize to...
4427         (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
4428         (*cond_asrd<mode>_2): Generalize to...
4429         (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
4430         (*cond_asrd<mode>_z): Generalize to...
4431         (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
4432         * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
4433         (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
4434         (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
4435         * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
4436         pattern.
4437         (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
4438         (@aarch64_scatter_stnt<mode>): Likewise.
4439         (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
4440         (@aarch64_mul_lane_<mode>): Likewise.
4441         (@aarch64_sve_suqadd<mode>_const): Likewise.
4442         (*<sur>h<addsub><mode>): Generalize to...
4443         (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
4444         new pattern.
4445         (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
4446         (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
4447         (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
4448         (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
4449         (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
4450         (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
4451         (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
4452         (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
4453         (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
4454         (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
4455         (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
4456         (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
4457         (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
4458         (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
4459         (@aarch64_sve_add_mul_lane_<mode>): Likewise.
4460         (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
4461         (@aarch64_sve2_xar<mode>): Likewise.
4462         (@aarch64_sve2_bcax<mode>): Likewise.
4463         (*aarch64_sve2_eor3<mode>): Rename to...
4464         (@aarch64_sve2_eor3<mode>): ...this.
4465         (@aarch64_sve2_bsl<mode>): New expander.
4466         (@aarch64_sve2_nbsl<mode>): Likewise.
4467         (@aarch64_sve2_bsl1n<mode>): Likewise.
4468         (@aarch64_sve2_bsl2n<mode>): Likewise.
4469         (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
4470         (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
4471         (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
4472         (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
4473         (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
4474         (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
4475         (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
4476         (<su>mull<bt><Vwide>): Generalize to...
4477         (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
4478         pattern.
4479         (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
4480         (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
4481         (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
4482         (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4483         (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
4484         (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4485         (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
4486         (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4487         (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
4488         (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4489         (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
4490         (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
4491         (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
4492         (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
4493         (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
4494         (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
4495         (<SHRNB:r>shrnb<mode>): Generalize to...
4496         (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
4497         new pattern.
4498         (<SHRNT:r>shrnt<mode>): Generalize to...
4499         (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
4500         new pattern.
4501         (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
4502         (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
4503         (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
4504         (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
4505         (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
4506         (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
4507         (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
4508         (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
4509         (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
4510         (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
4511         (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
4512         (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
4513         (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
4514         (@aarch64_sve2_cvtnt<mode>): Likewise.
4515         (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
4516         (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
4517         (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
4518         (@aarch64_sve2_cvtxnt<mode>): Likewise.
4519         (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
4520         (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
4521         (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
4522         (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
4523         (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
4524         (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
4525         (@aarch64_sve2_pmul<mode>): Likewise.
4526         (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
4527         (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
4528         (@aarch64_sve2_tbl2<mode>): Likewise.
4529         (@aarch64_sve2_tbx<mode>): Likewise.
4530         (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
4531         (@aarch64_sve2_histcnt<mode>): Likewise.
4532         (@aarch64_sve2_histseg<mode>): Likewise.
4533         (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
4534         (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
4535         (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
4536         (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
4537         (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
4538         (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
4539         (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
4540         (<su>mulh<r>s<mode>3): Update after above pattern name changes.
4541         * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
4542         (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
4543         (SVE2_PMULL_PAIR_I): New mode iterators.
4544         (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
4545         (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
4546         (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
4547         (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
4548         (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
4549         (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
4550         (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
4551         (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
4552         (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
4553         (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
4554         (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
4555         (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
4556         (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
4557         (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
4558         (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
4559         (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
4560         (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
4561         (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
4562         (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
4563         (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
4564         (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
4565         (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
4566         (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
4567         (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
4568         (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
4569         (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
4570         (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
4571         (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
4572         (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
4573         (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
4574         (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
4575         further down file.
4576         (VNARROW, Ventype): New mode attributes.
4577         (Vewtype): Handle VNx2DI.  Fix typo in comment.
4578         (VDOUBLE): New mode attribute.
4579         (sve_lane_con): Handle VNx8HI.
4580         (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
4581         (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
4582         (sve_int_op, sve_int_op_rev): Handle the above codes.
4583         (sve_pred_int_rhs2_operand): Likewise.
4584         (MULLBT, SHRNB, SHRNT): Delete.
4585         (SVE_INT_SHIFT_IMM): New int iterator.
4586         (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
4587         and UNSPEC_WHILEHS for TARGET_SVE2.
4588         (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
4589         (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
4590         (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
4591         (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
4592         (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
4593         (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
4594         (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
4595         (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
4596         (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
4597         (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
4598         (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
4599         (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
4600         (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
4601         (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
4602         (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
4603         (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
4604         (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
4605         (optab): Handle the new unspecs.
4606         (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
4607         and UNSPEC_RSHRNT.
4608         (lr): Handle the new unspecs.
4609         (bt): Delete.
4610         (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
4611         (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
4612         (sve_int_qsub_op): New int attributes.
4613         (sve_fp_op, rot): Handle the new unspecs.
4614         * config/aarch64/aarch64-sve-builtins.h
4615         (function_resolver::require_matching_pointer_type): Declare.
4616         (function_resolver::resolve_unary): Add an optional boolean argument.
4617         (function_resolver::finish_opt_n_resolution): Add an optional
4618         type_suffix_index argument.
4619         (gimple_folder::redirect_call): Declare.
4620         (gimple_expander::prepare_gather_address_operands): Add an optional
4621         bool parameter.
4622         * config/aarch64/aarch64-sve-builtins.cc: Include
4623         aarch64-sve-builtins-sve2.h.
4624         (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
4625         (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
4626         (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
4627         (TYPES_hsd_integer): Use TYPES_hsd_signed.
4628         (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
4629         (TYPES_s_unsigned): Likewise.
4630         (TYPES_s_integer): Use TYPES_s_unsigned.
4631         (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
4632         (TYPES_sd_integer): Use them.
4633         (TYPES_d_unsigned): New macro.
4634         (TYPES_d_integer): Use it.
4635         (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
4636         (TYPES_cvt_narrow): Likewise.
4637         (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
4638         (preds_mx): New variable.
4639         (function_builder::add_overloaded_function): Allow the new feature
4640         set to be more restrictive than the original one.
4641         (function_resolver::infer_pointer_type): Remove qualifiers from
4642         the pointer type before printing it.
4643         (function_resolver::require_matching_pointer_type): New function.
4644         (function_resolver::resolve_sv_displacement): Handle functions
4645         that don't support 32-bit vector indices or svint32_t vector offsets.
4646         (function_resolver::finish_opt_n_resolution): Take the inferred type
4647         as a separate argument.
4648         (function_resolver::resolve_unary): Optionally treat all forms in
4649         the same way as normal merging functions.
4650         (gimple_folder::redirect_call): New function.
4651         (function_expander::prepare_gather_address_operands): Add an argument
4652         that says whether scaled forms are available.  If they aren't,
4653         handle scaling of vector indices and don't add the extension and
4654         scaling operands.
4655         (function_expander::map_to_unspecs): If aarch64_sve isn't available,
4656         fall back to using cond_* instead.
4657         * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
4658         Split out the member variables into...
4659         (rtx_code_function_base): ...this new base class.
4660         (rtx_code_function_rotated): Inherit rtx_code_function_base.
4661         (unspec_based_function): Split out the member variables into...
4662         (unspec_based_function_base): ...this new base class.
4663         (unspec_based_function_rotated): Inherit unspec_based_function_base.
4664         (unspec_based_function_exact_insn): New class.
4665         (unspec_based_add_function, unspec_based_add_lane_function)
4666         (unspec_based_lane_function, unspec_based_pred_function)
4667         (unspec_based_qadd_function, unspec_based_qadd_lane_function)
4668         (unspec_based_qsub_function, unspec_based_qsub_lane_function)
4669         (unspec_based_sub_function, unspec_based_sub_lane_function): New
4670         typedefs.
4671         (unspec_based_fused_function): New class.
4672         (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
4673         (unspec_based_fused_lane_function): New class.
4674         (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
4675         typedefs.
4676         (CODE_FOR_MODE1): New macro.
4677         (fixed_insn_function): New class.
4678         (while_comparison): Likewise.
4679         * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
4680         (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
4681         (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
4682         (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
4683         (load_gather_sv_restricted, shift_left_imm_long): Declare.
4684         (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
4685         (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
4686         (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
4687         (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
4688         (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
4689         (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
4690         (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
4691         (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
4692         * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
4693         Also add an initial argument for unary_convert_narrowt, regardless
4694         of the predication type.
4695         (build_32_64): Allow loads and stores to specify MODE_none.
4696         (build_sv_index64, build_sv_uint_offset): New functions.
4697         (long_type_suffix): New function.
4698         (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
4699         (binary_imm_long_base, load_gather_sv_base): Likewise.
4700         (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
4701         (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
4702         (unary_narrowb_base, unary_narrowt_base): Likewise.
4703         (binary_long_lane_def, binary_long_lane): New shape.
4704         (binary_long_opt_n_def, binary_long_opt_n): Likewise.
4705         (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
4706         (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
4707         (binary_to_uint_def, binary_to_uint): Likewise.
4708         (binary_wide_def, binary_wide): Likewise.
4709         (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
4710         (compare_def, compare): Likewise.
4711         (compare_ptr_def, compare_ptr): Likewise.
4712         (load_ext_gather_index_restricted_def,
4713         load_ext_gather_index_restricted): Likewise.
4714         (load_ext_gather_offset_restricted_def,
4715         load_ext_gather_offset_restricted): Likewise.
4716         (load_gather_sv_def): Inherit from load_gather_sv_base.
4717         (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
4718         (shift_left_imm_def, shift_left_imm): Likewise.
4719         (shift_left_imm_long_def, shift_left_imm_long): Likewise.
4720         (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
4721         (store_scatter_index_restricted_def,
4722         store_scatter_index_restricted): Likewise.
4723         (store_scatter_offset_restricted_def,
4724         store_scatter_offset_restricted): Likewise.
4725         (tbl_tuple_def, tbl_tuple): Likewise.
4726         (ternary_long_lane_def, ternary_long_lane): Likewise.
4727         (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
4728         (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
4729         (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
4730         (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
4731         (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
4732         (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
4733         (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
4734         (ternary_uint_def, ternary_uint): Likewise.
4735         (unary_convert): Fix typo in comment.
4736         (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
4737         (unary_long_def, unary_long): Likewise.
4738         (unary_narrowb_def, unary_narrowb): Likewise.
4739         (unary_narrowt_def, unary_narrowt): Likewise.
4740         (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
4741         (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
4742         (unary_to_int_def, unary_to_int): Likewise.
4743         * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
4744         (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
4745         (svasrd_impl): Delete.
4746         (svcadd_impl::expand): Handle integer operations too.
4747         (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
4748         new functions to derive the unspec numbers.
4749         (svmla_svmls_lane_impl): Replace with...
4750         (svmla_lane_impl, svmls_lane_impl): ...these new classes.  Handle
4751         integer operations too.
4752         (svwhile_impl): Rename to...
4753         (svwhilelx_impl): ...this and inherit from while_comparison.
4754         (svasrd): Use unspec_based_function.
4755         (svmla_lane): Use svmla_lane_impl.
4756         (svmls_lane): Use svmls_lane_impl.
4757         (svrecpe, svrsqrte): Handle unsigned integer operations too.
4758         (svwhilele, svwhilelt): Use svwhilelx_impl.
4759         * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
4760         * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
4761         * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
4762         * config/aarch64/aarch64-sve-builtins.def: Include
4763         aarch64-sve-builtins-sve2.def.
4764
4765 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4766
4767         * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
4768         (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
4769         * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
4770         (aarch64_sve_sqadd_sqsub_immediate_p): Likewise.  Handle scalar
4771         immediates as well as vector ones.
4772         * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
4773         (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
4774         (aarch64_sve_qsub_immediate): Update calls accordingly.
4775
4776 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4777
4778         * config/aarch64/aarch64-sve2.md: Add banner comments.
4779         (<su>mulh<r>s<mode>3): Move further up file.
4780         (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
4781         (*aarch64_sve2_sra<mode>): Move further down file.
4782         * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
4783
4784 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4785
4786         * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
4787         and UNSPEC_WHILEWR.
4788         (while_optab_cmp): Handle them.
4789         * config/aarch64/aarch64-sve.md
4790         (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
4791         and add a "@" marker.
4792         * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
4793         instead of gen_aarch64_sve2_while_ptest.
4794         (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
4795
4796 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4797
4798         * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
4799         (UNSPEC_WHILELE): ...this.
4800         (UNSPEC_WHILE_LO): Rename to...
4801         (UNSPEC_WHILELO): ...this.
4802         (UNSPEC_WHILE_LS): Rename to...
4803         (UNSPEC_WHILELS): ...this.
4804         (UNSPEC_WHILE_LT): Rename to...
4805         (UNSPEC_WHILELT): ...this.
4806         * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
4807         (cmp_op, while_optab_cmp): Likewise.
4808         * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
4809         * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
4810         (svwhilelt): Likewise.
4811
4812 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4813
4814         * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
4815         (unary_to_uint): Define.
4816         * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
4817         (unary_count): Rename to...
4818         (unary_to_uint_def, unary_to_uint): ...this.
4819         * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
4820
4821 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4822
4823         * config/aarch64/aarch64-sve-builtins-functions.h
4824         (code_for_mode_function): New class.
4825         (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
4826         * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4827         (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4828         (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4829         (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4830
4831 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4832
4833         * config/aarch64/iterators.md (addsub): New code attribute.
4834         * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4835         Re-express as...
4836         (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4837         in the asm string and attributes.  Fix indentation.
4838         * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4839         Re-express as...
4840         (@aarch64_sve_<optab><mode>): ...this.
4841         * config/aarch64/aarch64-sve-builtins.h
4842         (function_expander::expand_signed_unpred_op): Delete.
4843         * config/aarch64/aarch64-sve-builtins.cc
4844         (function_expander::expand_signed_unpred_op): Likewise.
4845         (function_expander::map_to_rtx_codes): If the optab isn't defined,
4846         try using code_for_aarch64_sve instead.
4847         * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4848         (svqsub_impl): Likewise.
4849         (svqadd, svqsub): Use rtx_code_function instead.
4850
4851 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4852
4853         * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4854         (HADDSUB, sur, addsub): Remove them.
4855
4856 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4857
4858         * tree-nrv.c (pass_return_slot::execute): Handle all internal
4859         functions the same way, rather than singling out those that
4860         aren't mapped directly to optabs.
4861
4862 2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>
4863
4864         * target.def (compatible_vector_types_p): New target hook.
4865         * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4866         * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4867         * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4868         * doc/tm.texi: Regenerate.
4869         * gimple-expr.c: Include target.h.
4870         (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4871         * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4872         function.
4873         (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4874         * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4875         Use the original predicate if it already has a suitable type.
4876
4877 2020-01-09  Martin Jambor  <mjambor@suse.cz>
4878
4879         * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4880         resolve_speculation and redirect_call_stmt_to_callee static.  Change
4881         return type of set_call_stmt to cgraph_edge *.
4882         * auto-profile.c (afdo_indirect_call): Adjust call to
4883         redirect_call_stmt_to_callee.
4884         * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4885         make the this pointer explicit, adjust self-recursive calls and the
4886         call top make_direct.  Return the resulting edge.
4887         (cgraph_edge::remove): Make this pointer explicit.
4888         (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4889         (cgraph_edge::make_direct): Likewise, adjust call to
4890         resolve_speculation.
4891         (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4892         call to set_call_stmt.
4893         (cgraph_update_edges_for_call_stmt_node): Update call to
4894         set_call_stmt and remove.
4895         * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4896         Renamed edge to master_edge.  Adjusted calls to set_call_stmt.
4897         (cgraph_node::create_edge_including_clones): Moved "first" definition
4898         of edge to the block where it was used.  Adjusted calls to
4899         set_call_stmt.
4900         (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4901         cgraph_edge::remove.
4902         * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4903         make_direct and redirect_call_stmt_to_callee.
4904         * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4905         resolve_speculation and make_direct.
4906         * ipa-inline-transform.c (inline_transform): Adjust call to
4907         redirect_call_stmt_to_callee.
4908         (check_speculations_1):: Adjust call to resolve_speculation.
4909         * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4910         resolve-speculation.
4911         (inline_small_functions): Adjust call to resolve_speculation.
4912         (ipa_inline): Likewise.
4913         * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4914         make_direct.
4915         * ipa-visibility.c (function_and_variable_visibility): Make iteration
4916         safe with regards to edge removal, adjust calls to
4917         redirect_call_stmt_to_callee.
4918         * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4919         and redirect_call_stmt_to_callee.
4920         * multiple_target.c (create_dispatcher_calls): Adjust call to
4921         redirect_call_stmt_to_callee
4922         (redirect_to_specific_clone): Likewise.
4923         * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4924         Adjust calls to cgraph_edge::remove.
4925         * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4926         (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4927         (expand_call_inline): Adjust call to cgraph_edge::remove.
4928
4929 2020-01-09  Martin Liska  <mliska@suse.cz>
4930
4931         * params.opt: Set Optimization for
4932         param_max_speculative_devirt_maydefs.
4933
4934 2020-01-09  Martin Sebor  <msebor@redhat.com>
4935
4936         PR middle-end/93200
4937         PR fortran/92956
4938         * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4939
4940 2020-01-09  Martin Liska  <mliska@suse.cz>
4941
4942         * auto-profile.c (auto_profile): Use opt_for_fn
4943         for a parameter.
4944         * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4945         (propagate_vals_across_arith_jfunc): Likewise.
4946         (hint_time_bonus): Likewise.
4947         (incorporate_penalties): Likewise.
4948         (good_cloning_opportunity_p): Likewise.
4949         (perform_estimation_of_a_value): Likewise.
4950         (estimate_local_effects): Likewise.
4951         (ipcp_propagate_stage): Likewise.
4952         * ipa-fnsummary.c (decompose_param_expr): Likewise.
4953         (set_switch_stmt_execution_predicate): Likewise.
4954         (analyze_function_body): Likewise.
4955         * ipa-inline-analysis.c (offline_size): Likewise.
4956         * ipa-inline.c (early_inliner): Likewise.
4957         * ipa-prop.c (ipa_analyze_node): Likewise.
4958         (ipcp_transform_function): Likewise.
4959         * ipa-sra.c (process_scan_results): Likewise.
4960         (ipa_sra_summarize_function): Likewise.
4961         * params.opt: Rename ipcp-unit-growth to
4962         ipa-cp-unit-growth.  Add Optimization for various
4963         IPA-related parameters.
4964
4965 2020-01-09  Richard Biener  <rguenther@suse.de>
4966
4967         PR middle-end/93054
4968         * gimplify.c (gimplify_expr): Deal with NOP definitions.
4969
4970 2020-01-09  Richard Biener  <rguenther@suse.de>
4971
4972         PR tree-optimization/93040
4973         * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4974
4975 2020-01-09  Georg-Johann Lay  <avr@gjlay.de>
4976
4977         * common/config/avr/avr-common.c (avr_option_optimization_table)
4978         [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4979
4980 2020-01-09  Martin Liska  <mliska@suse.cz>
4981
4982         * cgraphclones.c (symbol_table::materialize_all_clones):
4983         Use cgraph_node::dump_name.
4984
4985 2020-01-09  Jakub Jelinek  <jakub@redhat.com>
4986
4987         PR inline-asm/93202
4988         * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4989         output_operand_lossage instead of gcc_unreachable.
4990         * doc/md.texi (riscv f constraint): Fix typo.
4991
4992         PR target/93141
4993         * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4994         SWI.  Use <general_hilo_operand> instead of <general_operand>.  Use
4995         CONST_SCALAR_INT_P instead of CONST_INT_P.
4996         (*subv<mode>4_1): Rename to ...
4997         (subv<mode>4_1): ... this.
4998         (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4999         define_insn_and_split patterns.
5000         (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
5001         patterns.
5002
5003 2020-01-08  David Malcolm  <dmalcolm@redhat.com>
5004
5005         * vec.c (class selftest::count_dtor): New class.
5006         (selftest::test_auto_delete_vec): New test.
5007         (selftest::vec_c_tests): Call it.
5008         * vec.h (class auto_delete_vec): New class template.
5009         (auto_delete_vec<T>::~auto_delete_vec): New dtor.
5010
5011 2020-01-08  David Malcolm  <dmalcolm@redhat.com>
5012
5013         * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
5014
5015 2020-01-08  Jim Wilson  <jimw@sifive.com>
5016
5017         * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
5018         use of TLS_MODEL_LOCAL_EXEC when not pic.
5019
5020 2020-01-08  David Malcolm  <dmalcolm@redhat.com>
5021
5022         * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
5023         memory leak.
5024
5025 2020-01-08  Jakub Jelinek  <jakub@redhat.com>
5026
5027         PR target/93187
5028         * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
5029         *stack_protect_set_3 peephole2): Also check that the second
5030         insns source is general_operand.
5031
5032         PR target/93174
5033         * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
5034         predicate for output operand instead of register_operand.
5035         (addcarry<mode>, addcarry<mode>_1): Likewise.  Add alternative with
5036         memory destination and non-memory operands[2].
5037
5038 2020-01-08  Martin Liska  <mliska@suse.cz>
5039
5040         * cgraph.c (cgraph_node::dump): Use ::dump_name or
5041         ::dump_asm_name instead of (::name or ::asm_name).
5042         * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
5043         * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
5044         (analyze_functions): Likewise.
5045         (expand_all_functions): Likewise.
5046         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
5047         (propagate_bits_across_jump_function): Likewise.
5048         (dump_profile_updates): Likewise.
5049         (ipcp_store_bits_results): Likewise.
5050         (ipcp_store_vr_results): Likewise.
5051         * ipa-devirt.c (dump_targets): Likewise.
5052         * ipa-fnsummary.c (analyze_function_body): Likewise.
5053         * ipa-hsa.c (check_warn_node_versionable): Likewise.
5054         (process_hsa_functions): Likewise.
5055         * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
5056         (set_alias_uids): Likewise.
5057         * ipa-inline-transform.c (save_inline_function_body): Likewise.
5058         * ipa-inline.c (recursive_inlining): Likewise.
5059         (inline_to_all_callers_1): Likewise.
5060         (ipa_inline): Likewise.
5061         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
5062         (ipa_propagate_frequency): Likewise.
5063         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5064         (remove_described_reference): Likewise.
5065         * ipa-pure-const.c (worse_state): Likewise.
5066         (check_retval_uses): Likewise.
5067         (analyze_function): Likewise.
5068         (propagate_pure_const): Likewise.
5069         (propagate_nothrow): Likewise.
5070         (dump_malloc_lattice): Likewise.
5071         (propagate_malloc): Likewise.
5072         (pass_local_pure_const::execute): Likewise.
5073         * ipa-visibility.c (optimize_weakref): Likewise.
5074         (function_and_variable_visibility): Likewise.
5075         * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
5076         (ipa_discover_variable_flags): Likewise.
5077         * lto-streamer-out.c (output_function): Likewise.
5078         (output_constructor): Likewise.
5079         * tree-inline.c (copy_bb): Likewise.
5080         * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
5081         * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
5082
5083 2020-01-08  Richard Biener  <rguenther@suse.de>
5084
5085         PR middle-end/93199
5086         * tree-eh.c (sink_clobbers): Update virtual operands for
5087         the first and last stmt only.  Add a dry-run capability.
5088         (pass_lower_eh_dispatch::execute): Perform clobber sinking
5089         after CFG manipulations and in RPO order to catch all
5090         secondary opportunities reliably.
5091
5092 2020-01-08  Georg-Johann Lay  <avr@gjlay.de>
5093
5094         PR target/93182
5095         * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
5096
5097 2019-01-08  Richard Biener  <rguenther@suse.de>
5098
5099         PR middle-end/93199
5100         * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
5101         * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
5102         virtual operand, also updating SSA use.
5103         * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
5104         Update stmt after resetting virtual operand.
5105         (tree_loop_interchange::move_code_to_inner_loop): Likewise.
5106         * gimple-iterator.c (gsi_remove): When not removing the stmt
5107         permanently do not delink immediate uses or mark the stmt modified.
5108
5109 2020-01-08  Martin Liska  <mliska@suse.cz>
5110
5111         * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
5112         (ipa_call_context::estimate_size_and_time): Likewise.
5113         (inline_analyze_function): Likewise.
5114
5115 2020-01-08  Martin Liska  <mliska@suse.cz>
5116
5117         * cgraph.c (cgraph_node::dump): Use systematically
5118         dump_asm_name.
5119
5120 2020-01-08  Georg-Johann Lay  <avr@gjlay.de>
5121
5122         Add -nodevicespecs option for avr.
5123
5124         PR target/93182
5125         * config/avr/avr.opt (-nodevicespecs): New driver option.
5126         * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
5127         "-specs=device-specs/..." if that option is not set.
5128         * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
5129
5130 2020-01-08  Georg-Johann Lay  <avr@gjlay.de>
5131
5132         Implement 64-bit double functions for avr.
5133
5134         PR target/92055
5135         * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
5136         --with-double-comparison.
5137         * doc/install.texi: Document them.
5138         * config/avr/avr-c.c (avr_cpu_cpp_builtins)
5139         <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
5140         <WITH_DOUBLE_COMPARISON>: New built-in defines.
5141         * doc/invoke.texi (AVR Built-in Macros): Document them.
5142         * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
5143         * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
5144         * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
5145
5146 2020-01-08  Richard Earnshaw  <rearnsha@arm.com>
5147
5148         PR target/93188
5149         * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
5150         armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
5151         when only building rm-profile multilibs.
5152
5153 2020-01-08  Feng Xue  <fxue@os.amperecomputing.com>
5154
5155         PR ipa/93084
5156         * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
5157         lattice for a value to check.
5158         (propagate_vals_across_arith_jfunc): Add an assertion to ensure
5159         finite propagation in self-recursive scc.
5160
5161 2020-01-08  Luo Xiong Hu  <luoxhu@linux.ibm.com>
5162
5163         * ipa-inline.c (caller_growth_limits): Restore the AND.
5164
5165 2020-01-07  Andrew Stubbs  <ams@codesourcery.com>
5166
5167         * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
5168         (VEC_ALLREG_ALT): New iterator.
5169         (VEC_ALLREG_INT_MODE): New iterator.
5170         (VCMP_MODE): New iterator.
5171         (VCMP_MODE_INT): New iterator.
5172         (vec_cmpu<mode>di): Use VCMP_MODE_INT.
5173         (vec_cmp<u>v64qidi): New define_expand.
5174         (vec_cmp<mode>di_exec): Use VCMP_MODE.
5175         (vec_cmpu<mode>di_exec): New define_expand.
5176         (vec_cmp<u>v64qidi_exec): New define_expand.
5177         (vec_cmp<mode>di_dup): Use VCMP_MODE.
5178         (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
5179         (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
5180         (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
5181         (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
5182         (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
5183         (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
5184         (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
5185         (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
5186         (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
5187         this.
5188         * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
5189         * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
5190
5191 2020-01-07  Andrew Stubbs  <ams@codesourcery.com>
5192
5193         * config/gcn/constraints.md (DA): Update description and match.
5194         (DB): Likewise.
5195         (Db): New constraint.
5196         * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
5197         parameter.
5198         * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
5199         Implement 'Db' mixed immediate type.
5200         * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
5201         (addcv64si3_dup<exec_vcc>): Delete.
5202         (subcv64si3<exec_vcc>): Rework constraints.
5203         (addv64di3): Rework constraints.
5204         (addv64di3_exec): Rework constraints.
5205         (subv64di3): Rework constraints.
5206         (addv64di3_dup): Delete.
5207         (addv64di3_dup_exec): Delete.
5208         (addv64di3_zext): Rework constraints.
5209         (addv64di3_zext_exec): Rework constraints.
5210         (addv64di3_zext_dup): Rework constraints.
5211         (addv64di3_zext_dup_exec): Rework constraints.
5212         (addv64di3_zext_dup2): Rework constraints.
5213         (addv64di3_zext_dup2_exec): Rework constraints.
5214         (addv64di3_sext_dup2): Rework constraints.
5215         (addv64di3_sext_dup2_exec): Rework constraints.
5216
5217 2020-01-07  Andre Vieira  <andre.simoesdiasvieira@arm.com>
5218
5219         * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
5220         existing target checks.
5221
5222 2020-01-07  Richard Biener  <rguenther@suse.de>
5223
5224         * doc/install.texi: Bump minimal supported MPC version.
5225
5226 2020-01-07  Richard Sandiford  <richard.sandiford@arm.com>
5227
5228         * langhooks-def.h (lhd_simulate_enum_decl): Declare.
5229         (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
5230         * langhooks.c: Include stor-layout.h.
5231         (lhd_simulate_enum_decl): New function.
5232         * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
5233         handle_arm_sve_h for the LTO frontend.
5234         (register_vector_type): Cope with null returns from pushdecl.
5235
5236 2020-01-07  Richard Sandiford  <richard.sandiford@arm.com>
5237
5238         * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
5239         (aarch64_sve::nvectors_if_data_type): Replace with...
5240         (aarch64_sve::builtin_type_p): ...this.
5241         * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
5242         (find_vector_type): Delete.
5243         (add_sve_type_attribute): New function.
5244         (lookup_sve_type_attribute): Likewise.
5245         (register_builtin_types): Add an "SVE type" attribute to each type.
5246         (register_tuple_type): Likewise.
5247         (svbool_type_p, nvectors_if_data_type): Delete.
5248         (mangle_builtin_type): Use lookup_sve_type_attribute.
5249         (builtin_type_p): Likewise.  Add an overload that returns the
5250         number of constituent vector and predicate registers.
5251         * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
5252         (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
5253         instead of aarch64_sve_argument_p.
5254         (aarch64_takes_arguments_in_sve_regs_p): Likewise.
5255         (aarch64_pass_by_reference): Likewise.
5256         (aarch64_function_value_1): Likewise.
5257         (aarch64_return_in_memory): Likewise.
5258         (aarch64_layout_arg): Likewise.
5259
5260 2020-01-07  Jakub Jelinek  <jakub@redhat.com>
5261
5262         PR tree-optimization/93156
5263         * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
5264         least significant bit is always clear.
5265
5266         PR tree-optimization/93118
5267         * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?.  Add new
5268         simplifier with two intermediate conversions.
5269
5270 2020-01-07  Martin Liska  <mliska@suse.cz>
5271
5272         * params.opt: Add Optimization for various parameters.
5273
5274 2020-01-07  Martin Liska  <mliska@suse.cz>
5275
5276         PR ipa/83411
5277         * doc/extend.texi: Explain cloning for target_clone
5278         attribute.
5279
5280 2020-01-07  Martin Liska  <mliska@suse.cz>
5281
5282         PR tree-optimization/92860
5283         * common.opt: Make in Optimization option
5284         as it is affected by -O0, which is an Optimization
5285         option.
5286         * tree-inline.c (tree_inlinable_function_p):
5287         Use opt_for_fn for warn_inline.
5288         (expand_call_inline): Likewise.
5289
5290 2020-01-07  Martin Liska  <mliska@suse.cz>
5291
5292         PR tree-optimization/92860
5293         * common.opt: Make flag_ree as optimization
5294         attribute.
5295
5296 2020-01-07  Martin Liska  <mliska@suse.cz>
5297
5298         PR optimization/92860
5299         * params.opt: Mark param_min_crossjump_insns with Optimization
5300         keyword.
5301
5302 2020-01-07  Luo Xiong Hu  <luoxhu@linux.ibm.com>
5303
5304         * ipa-inline-analysis.c (estimate_growth): Fix typo.
5305         * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
5306
5307 2020-01-06  Michael Meissner  <meissner@linux.ibm.com>
5308
5309         * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
5310         helper function to return the valid addressing formats for a given
5311         hard register and mode.
5312         (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
5313
5314         * config/rs6000/constraints.md (Q constraint): Update
5315         documentation.
5316         * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
5317         documentation.
5318
5319         * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
5320         Use 'Q' for doing vector extract from memory.
5321         (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
5322         memory.
5323         (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
5324         doing vector extract from memory.
5325         (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
5326         extract from memory.
5327
5328         * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
5329         for the offset being 34-bits when -mcpu=future is used.
5330
5331 2020-01-06  John David Anglin  <danglin@gcc.gnu.org>
5332
5333         * config/pa/pa.md: Revert change to use ordered_comparison_operator
5334         instead of cmpib_comparison_operator in cmpib patterns.
5335         * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
5336         of cmpib_comparison_operator.  Revise comment.
5337
5338 2020-01-06  Richard Sandiford  <richard.sandiford@arm.com>
5339
5340         * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
5341         in an IFN_DIV_POW2 node to be equal.
5342
5343 2020-01-06  Richard Sandiford  <richard.sandiford@arm.com>
5344
5345         * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
5346         (vect_check_scalar_mask): ...this.
5347         (vectorizable_store, vectorizable_load): Update call accordingly.
5348         (vectorizable_call): Use vect_check_scalar_mask to check the mask
5349         argument in calls to conditional internal functions.
5350
5351 2020-01-06  Andrew Stubbs  <ams@codesourcery.com>
5352
5353         * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
5354         '0' matching inputs.
5355         (subv64di3_exec): Likewise.
5356
5357 2020-01-06  Bryan Stenson  <bryan@siliconvortex.com>
5358
5359         * config/mips/mips.c (vr4130_align_insns): Fix typo.
5360         * doc/md.texi (movstr): Likewise.
5361
5362 2020-01-06  Andrew Stubbs  <ams@codesourcery.com>
5363
5364         * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
5365         clobber.
5366
5367 2020-01-06  Richard Sandiford  <richard.sandiford@arm.com>
5368
5369         * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
5370         Depend on...
5371         (s-aarch64-tune-md): ...this new stamp file.  Pipe the new contents
5372         to a temporary file and use move-if-change to update the real
5373         file where necessary.
5374
5375 2020-01-06  Richard Sandiford  <richard.sandiford@arm.com>
5376
5377         * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
5378         rather than Upa for CPY /M.
5379
5380 2020-01-06  Andrew Stubbs  <ams@codesourcery.com>
5381
5382         * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
5383         immediate.
5384
5385 2020-01-06  Martin Liska  <mliska@suse.cz>
5386
5387     PR tree-optimization/92860
5388     * params.opt: Mark param_max_combine_insns with Optimization
5389     keyword.
5390
5391 2020-01-05  Jakub Jelinek  <jakub@redhat.com>
5392
5393         PR target/93141
5394         * config/i386/i386.md (SWIDWI): New mode iterator.
5395         (DWI, dwi): Add TImode variants.
5396         (addv<mode>4): Use SWIDWI iterator instead of SWI.  Use
5397         <general_hilo_operand> instead of <general_operand>.  Use
5398         CONST_SCALAR_INT_P instead of CONST_INT_P.
5399         (*addv<mode>4_1): Rename to ...
5400         (addv<mode>4_1): ... this.
5401         (QWI): New mode attribute.
5402         (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
5403         define_insn_and_split patterns.
5404         (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
5405         patterns.
5406         (uaddv<mode>4): Use SWIDWI iterator instead of SWI.  Use
5407         <general_hilo_operand> instead of <general_operand>.
5408         (*addcarry<mode>_1): New define_insn.
5409         (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
5410
5411 2020-01-03  Konstantin Kharlamov  <Hi-Angel@yandex.ru>
5412
5413         * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
5414         Use "call" instead of "set".
5415
5416 2020-01-03  Martin Jambor  <mjambor@suse.cz>
5417
5418         PR ipa/92917
5419         * ipa-cp.c (print_all_lattices): Skip functions without info.
5420
5421 2020-01-03  Jakub Jelinek  <jakub@redhat.com>
5422
5423         PR target/93089
5424         * config/i386/i386-options.c (ix86_simd_clone_adjust): If
5425         TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
5426         simd clones.  If TARGET_PREFER_AVX256, use prefer-vector-width=512
5427         for 'e' simd clones.
5428
5429         PR target/93089
5430         * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
5431         entry.
5432         (mprefer-vector-width=): Add Save.
5433         * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
5434         -mprefer-vector-width= if non-zero.  Fix up -mfpmath= comment.
5435         (ix86_debug_options, ix86_function_specific_print): Adjust
5436         ix86_target_string callers.
5437         (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
5438         (ix86_valid_target_attribute_tree): Likewise.
5439         * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
5440         * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
5441         ix86_target_string caller.
5442
5443         PR target/93110
5444         * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
5445         emitting ASHIFTRT, XOR and MINUS by hand.  Use gen_int_mode with QImode
5446         instead of gen_int_shift_amount + convert_modes.
5447
5448         PR rtl-optimization/93088
5449         * loop-iv.c (find_single_def_src): Punt after looking through
5450         128 reg copies for regs with single definitions.  Move definitions
5451         to first uses.
5452
5453 2020-01-02  Dennis Zhang  <dennis.zhang@arm.com>
5454
5455         * config/arm/arm-c.c (arm_cpu_builtins): Define
5456         __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
5457         __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
5458         __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
5459         * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
5460         * config/arm/arm-tables.opt: Regenerated.
5461         * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
5462         arm_arch_i8mm and arm_arch_bf16 when enabled.
5463         * config/arm/arm.h (TARGET_I8MM): New macro.
5464         (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
5465         * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
5466         * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
5467         * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
5468         (v8_6_a_simd_variants): New.
5469         (v8_*_a_simd_variants): Add i8mm and bf16.
5470         * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
5471
5472 2020-01-02  Jakub Jelinek  <jakub@redhat.com>
5473
5474         PR ipa/93087
5475         * predict.c (compute_function_frequency): Don't call
5476         warn_function_cold on functions that already have cold attribute.
5477
5478 2020-01-01  John David Anglin  <danglin@gcc.gnu.org>
5479
5480         PR target/67834
5481         * config/pa/pa.c (pa_elf_select_rtx_section): New.  Put references to
5482         COMDAT group function labels in .data.rel.ro.local section.
5483         * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
5484
5485         PR target/93111
5486         * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
5487         comparison_operator in B and S integer comparisons.  Likewise, use
5488         ordered_comparison_operator instead of cmpib_comparison_operator in
5489         cmpib patterns.
5490         * config/pa/predicates.md (cmpib_comparison_operator): Remove.
5491
5492 2020-01-01  Jakub Jelinek  <jakub@redhat.com>
5493
5494         Update copyright years.
5495
5496         * gcc.c (process_command): Update copyright notice dates.
5497         * gcov-dump.c (print_version): Ditto.
5498         * gcov.c (print_version): Ditto.
5499         * gcov-tool.c (print_version): Ditto.
5500         * gengtype.c (create_file): Ditto.
5501         * doc/cpp.texi: Bump @copying's copyright year.
5502         * doc/cppinternals.texi: Ditto.
5503         * doc/gcc.texi: Ditto.
5504         * doc/gccint.texi: Ditto.
5505         * doc/gcov.texi: Ditto.
5506         * doc/install.texi: Ditto.
5507         * doc/invoke.texi: Ditto.
5508
5509 2020-01-01  Jan Hubicka  <hubicka@ucw.cz>
5510
5511         * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
5512         summary.
5513
5514 2020-01-01  Jakub Jelinek  <jakub@redhat.com>
5515
5516         PR tree-optimization/93098
5517         * match.pd (popcount): For shift amounts, use integer_onep
5518         or wi::to_widest () == cst instead of tree_to_uhwi () == cst
5519         tests.  Make sure that precision is power of two larger than or equal
5520         to 16.  Ensure shift is never negative.  Use HOST_WIDE_INT_UC macro
5521         instead of ULL suffixed constants.  Formatting fixes.
5522 \f
5523 Copyright (C) 2020 Free Software Foundation, Inc.
5524
5525 Copying and distribution of this file, with or without modification,
5526 are permitted in any medium without royalty provided the copyright
5527 notice and this notice are preserved.