gcc/ChangeLog

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