gcc/ChangeLog

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