12018-12-31 Max Filippov <jcmvbkbc@gmail.com> 2 3 PR target/86814 4 * config/xtensa/xtensa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): 5 Define to speculation_safe_value_not_needed. 6 72018-12-31 Martin Liska <mliska@suse.cz> 8 9 * predict.c (set_even_probabilities): Include also 10 unlikely_count in calculation. 11 (combine_predictions_for_bb): Consider also HOT and 12 COLD labels predictions. 13 * predict.def (PRED_HOT_LABEL): Move it just after 14 __builtin_expect_with_probability predictor. 15 (PRED_COLD_LABEL): Likewise. 16 172018-12-30 Jan Hubicka <hubicka@ucw.cz> 18 19 * x86-tune.def: Enable inter_unit_moves_to_vec for generic. 20 212018-12-30 Martin Sebor <msebor@redhat.com> 22 Jeff Law <law@redhat.com> 23 24 * gimple-fold.c (get_range_strlen): Minor logic cleanup. Add comments 25 on code's intent. 26 27 * gimple-fold.c (strlen_range_kind): New enum. 28 (get_range_strlen): Update signature to use strlen_range_kind 29 instead of type+fuzzy. 30 (get_range_strlen_tree): Update signature to use 31 strlen_range_kind instead of type+fuzzy. Pass rkind down to 32 get_range_strlen. Check the rkind enum value instead of 33 type+fuzzy. 34 (get_range_strlen): Similarly. 35 (get_maxval_strlen): Update signature to make NONSTR argument have 36 a default value and make static. Add asserts to ensure sanity of 37 arguments. Update calls to get_range_strlen. 38 (gimple_fold_builtin_strcpy): Update calls to get_maxval_strlen. 39 (gimple_fold_builtin_strcat, gimple_fold_builtin_fputs): Similarly. 40 (gimple_fold_builtin_memory_chk): Similarly. 41 (gimple_fold_builtin_stxcpy_chk): Similarly. 42 (gimple_fold_builtin_snprintf_chk): Similarly. 43 (gimple_fold_builtin_sprintf, gimple_fold_builtin_snprintf): Similarly. 44 * gimple-fold.h (get_maxval_strlen): Delete prototype. 45 462018-12-29 Jozef Lawrynowicz <jozef.l@mittosystems.com> 47 48 * config/msp430/msp430.c (msp430_attr): Warn when the critical and 49 interrupt function attributes are used together. 50 * gcc/doc/extend.texi: Update documentation on the critical attribute. 51 522018-12-29 John David Anglin <danglin@gcc.gnu.org> 53 54 * config.gcc (hppa*64*-*-linux*): Add pa/t-pa to tmake_file. Define 55 d_target_objs. 56 (hppa*-*-openbsd*): Likewise. 57 (hppa[12]*-*-hpux10*): Likewise. 58 (hppa*64*-*-hpux11*): Likewise. 59 (hppa[12]*-*-hpux11*): Likewise. 60 612018-12-29 Jakub Jelinek <jakub@redhat.com> 62 63 PR ipa/88586 64 * omp-low.c (lower_omp_task_reductions): Set DECL_CONTEXT on field 65 and ifield. Update TYPE_ALIGN from alignment of field, ifield or 66 bfield. 67 682018-12-28 Uros Bizjak <ubizjak@gmail.com> 69 70 * config/i386/i386.h (ADDITIONAL_REGISTER_NAMES): Add 71 sil, dil, bpl and spl aliases. 72 732018-12-28 Martin Sebor <msebor@redhat.com> 74 75 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Avoid 76 undesirable warning. 77 782018-12-28 Thomas Schwinge <thomas@codesourcery.com> 79 80 * omp-expand.c (expand_omp_target): Restructure OpenACC vs. OpenMP 81 code paths. Update for libgomp OpenACC entry points change. 82 832018-12-28 Thomas Schwinge <thomas@codesourcery.com> 84 Julian Brown <julian@codesourcery.com> 85 86 * gimplify.c (gimplify_scan_omp_clauses): Fix known_eq typo/bug. 87 882018-12-27 Jan Hubicka <hubicka@ucw.cz> 89 90 * ipa-devirt.c (polymorphic_call_target_d): Add n_odr_types. 91 (polymorphic_call_target_hasher::hash): Hash it. 92 (polymorphic_call_target_hasher::equal): Compare it. 93 (possible_polymorphic_call_targets): Set it. 94 * tree.c (free_lang_data): Rebuild type inheritance graph even on 95 non-LTO path. 96 972018-12-27 Martin Liska <mliska@suse.cz> 98 99 PR gcov-profile/88225 100 * gcov.c(source_info::get_functions_at_location): 101 Use newly added line_to_function_map. 102 (source_info::add_function): New. 103 (output_json_intermediate_file): Use a pointer return 104 type for get_functions_at_location. 105 (process_all_functions): Use add_function instead 106 of direct push to a s->functions container. 107 (release_structures): Release ident_to_fn. 108 (read_graph_file): Register function into ident_to_fn. 109 (read_count_file): Use the map. 110 (output_lines): Handle pointer return type of 111 get_functions_at_location. 112 1132018-12-27 Martin Liska <mliska@suse.cz> 114 115 * builtins.c (expand_movstr): Compare with RETURN_BEGIN. 116 * expr.c (move_by_pieces): Likewise. 117 (store_by_pieces): Likewise. 118 (store_expr): Fix GNU coding style. 119 1202018-12-26 Andrew Pinski <apinski@marvell.com> 121 122 * config/aarch64/aarch64-cores.def (octeontx): New. 123 (octeontx81): Likewise. 124 (octeontx83): Likewise. 125 * config/aarch64/aarch64-tune.md: Regenerate. 126 * doc/invoke.texi (AArch64 Options) [mtune]: Add octeontx, octeontx81 127 and octeontx83. 128 1292018-12-24 Martin Sebor <msebor@redhat.com> 130 131 * gimple-fold.c (get_range_strlen_tree): Factored out of 132 get_range_strlen. Minor comment updates/additions. Assert 133 argument is not a simple SSA_NAME. 134 (get_range_strlen): Call get_range_strlen_tree as appropriate. 135 Minor comment updates/additions. 136 1372018-12-24 Jan Hubicka <hubicka@ucw.cz> 138 139 * ipa-devirt.c (dump_targets): Cap number of targets printed. 140 (dump_possible_polymorphic_call_targets): Add verbose parameter. 141 (ipa_devirt): Use it. 142 * ipa-utils.h (dump_possible_polymorphic_call_targets): Add verbose 143 parametrs. 144 1452018-12-24 Jan Hubicka <hubicka@ucw.cz> 146 147 PR lto/88140 148 * tree.c (fld_simplified_type): Temporarily disable array 149 simplification. 150 1512018-12-24 Jan Hubicka <hubicka@ucw.cz> 152 153 * lto-symtab.c (lto_symtab_merge_decls_2): Do not report ODR violations 154 for method whose basetype was already reported. 155 * ipa-devirt.c (odr_type_violation_reported_p): New. 156 * ipa-utils.h (odr_type_violation_reported_p): Declare. 157 1582018-12-24 Iain Sandoe <iain@sandoe.co.uk> 159 160 * configure.ac (dwarf2_debug_line): Check for the debug_line 161 section using otool when there’s no objdump available. 162 * configure: Regenerate. 163 1642018-12-24 Jan Hubicka <hubicka@ucw.cz> 165 166 * ipa-utils.c (ipa_merge_profiles): Fix updating of fnsummary; 167 also handle rescaling of mismatched profiles. 168 * ipa-fnsummary.c (analyze_function): Handle speculative edges. 169 1702018-12-23 Martin Sebor <msebor@redhat.com> 171 Jeff Law <law@redhat.com> 172 173 * builtins.h (c_strlen_data): Add new fields and comments. 174 * builtins.c (unterminated_array): Change field reference from 175 "len" to "minlen" in c_strlen_data instance. 176 * gimple-fold.c (get_range_strlen): Likewise. 177 * gimple-ssa-sprintf.c (get_string_length): Likewise. 178 179 * builtins.c (unterminated_array): Rename "data" to "lendata". Fix 180 a few comments. 181 (expand_builtin_strnlen, expand_builtin_stpcpy_1): Likewise. 182 (fold_builtin_strlen): Likewise. 183 * gimple-fold.c (get_range_strlen): Likewise. Also rename a couple 184 instances of "type" to "optype" and "arg" to "fld". 185 186 * builtins.c (unterminated_array): Use empty brace initialization 187 for c_strlen_data. 188 (c_strlen, expand_builtin_strnlen): Likewise. 189 (expand_builtin_stpcpy_1, fold_builtin_strlen): Likewise. 190 * gimple-fold.c (get_range_strlen): Likewise. 191 (gimple_fold_builtin_stpcpy): Likewise. 192 * gimple-ssa-sprintf.c (get_string_length): Likewise. 193 1942018-12-23 Alan Modra <amodra@gmail.com> 195 196 PR 88346 197 * config/rs6000/rs6000.h (ASM_CPU_SPEC): Correct %e message. Handle 198 -mcpu=rs64, not -mcpu=rs64a. Handle -mcpu=powerpc64 and -mcpu=titan. 199 * config/rs6000/driver-rs6000.c (asm_names): Similarly. 200 * config/rs6000/aix71.h (ASM_CPU_SPEC): Delete %e message. Handle 201 -mcpu=rs64, not -mcpu=rs64a. 202 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise. 203 2042018-12-23 Iain Sandoe <iain@sandoe.co.uk> 205 206 * lto-wrapper.c (tool_cleanup): Don’t report ficticious temp files. 207 (debug_objcopy): Improve temp filenames. Append the early 208 debug objects to the outfiles. Remove the phase prelinking the 209 early debug. Emit the early debug filenames for ld. 210 2112018-12-23 Iain Sandoe <iain@sandoe.co.uk> 212 213 * config/i386/darwin.h (TARGET_ASM_OUTPUT_IDENT): New. 214 2152018-12-23 Iain Sandoe <iain@sandoe.co.uk> 216 217 * collect2.c (maybe_run_lto_and_relink): Don’t say we have a temp file 218 unless we actually did some LTO. 219 (has_lto_section, is_lto_object_file): New. 220 (maybe_lto_object_file): Remove. 221 (scan_prog_file): Use is_lto_object_file() instead of scanning the 222 output of nm. 223 2242018-12-23 Iain Sandoe <iain@sandoe.co.uk> 225 226 * collect2.c (main): Parse the output file early so we can make nicer 227 temp names. Respond to “-save-temps” in the GCC OPTIONS. 228 (maybe_unlink): Don’t print “[Leaving…”] for files we never created 229 and don’t exist. 230 2312018-12-23 Iain Sandoe <iain@sandoe.co.uk> 232 233 * collect2.c (main): Combine flags from both the command line and 234 COLLECT_GCC_OPTIONS to determine the set in force 235 2362018-12-21 Jan Hubicka <hubicka@ucw.cz> 237 238 * ipa-utils.c (ipa_merge_profiles): Recompute summaries. 239 2402018-12-21 Jakub Jelinek <jakub@redhat.com> 241 242 PR middle-end/85594 243 PR middle-end/88553 244 * omp-expand.c (extract_omp_for_update_vars): Regimplify the condition 245 if needed. 246 (expand_omp_for_generic): Don't clobber t temporary for ordered loops. 247 2482018-12-15 Jan Hubicka <hubicka@ucw.cz> 249 250 PR ipa/88561 251 * ipa-polymorphic-call.c 252 (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Handle 253 arguments of thunks correctly. 254 (ipa_polymorphic_call_context::get_dynamic_context): Be ready for 255 NULL instance pinter. 256 * lto-cgraph.c (lto_output_node): Always stream thunk info. 257 2582018-12-21 Andreas Krebbel <krebbel@linux.ibm.com> 259 260 * config/s390/vector.md ("floatv2div2df2", "floatunsv2div2df2") 261 ("fix_truncv2dfv2di2", "fixuns_truncv2dfv2di2"): New pattern 262 definitions. 263 2642018-12-21 Eric Botcazou <ebotcazou@adacore.com> 265 266 PR rtl-optimization/87727 267 * combine.c (cant_combine_insn_p): On a LEAF_REGISTERS target, combine 268 again moves from leaf hard registers. 269 270 * final.c (final_scan_insn_1) <NOTE_INSN_INLINE_ENTRY>: Minor tweak. 271 2722018-12-21 Jakub Jelinek <jakub@redhat.com> 273 274 PR target/88522 275 * config/i386/sse.md (*avx512pf_gatherpf<mode>sf_mask, 276 *avx512pf_gatherpf<mode>df_mask, *avx512pf_scatterpf<mode>sf_mask, 277 *avx512pf_scatterpf<mode>df_mask): Use %X5 instead of %5 for 278 -masm=intel. 279 (gatherq_mode): Remove mode iterator. 280 (*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2): Use X instead 281 of <xtg_mode>. 282 (*avx512f_gatherdi<mode>): Use X instead of <gatherq_mode>. 283 (*avx512f_gatherdi<mode>_2, *avx512f_scattersi<mode>, 284 *avx512f_scatterdi<mode>): Use %X5 for -masm=intel. 285 286 PR rtl-optimization/88563 287 * expr.c (expand_expr_real_2) <case WIDEN_MULT_EXPR>: Swap innermode 288 and mode arguments to convert_modes. Likewise swap mode and word_mode 289 arguments. Handle both arguments with VOIDmode before convert_modes 290 of one of them. Formatting fixes. 291 2922018-12-21 Uros Bizjak <ubizjak@gmail.com> 293 294 PR target/88556 295 * internal-fn.def (COSH): New. 296 (SINH): Ditto. 297 (TANH): Ditto. 298 * optabs.def (cosh_optab): New. 299 (sinh_optab): Ditto. 300 (tanh_optab): Ditto. 301 * config/i386/i386-protos.h (ix86_emit_i387_sinh): New prototype. 302 (ix86_emit_i387_cosh): Ditto. 303 (ix86_emit_i387_tanh): Ditto. 304 * config/i386/i386.c (ix86_emit_i387_sinh): New function. 305 (ix86_emit_i387_cosh): Ditto. 306 (ix86_emit_i387_tanh): Ditto. 307 * config/i386/i386.md (sinhxf2): New expander. 308 (sinh<mode>2): Ditto. 309 (coshxf2): Ditto. 310 (cosh<mode>2): Ditto. 311 (tanhxf2): Ditto. 312 (tanh<mode>2): Ditto. 313 3142018-12-21 Jakub Jelinek <jakub@redhat.com> 315 316 PR target/88547 317 * config/i386/i386.c (ix86_expand_int_sse_cmp): Optimize 318 x > y ? 0 : -1 into min (x, y) == x ? -1 : 0. 319 3202018-12-21 Jan Beulich <jbeulich@suse.com> 321 322 * config/i386/sse.md (vaesdec_<mode>, vaesdeclast_<mode>): Allow 323 memory input. 324 3252018-12-21 Alexandre Oliva <oliva@adacore.com> 326 327 * config.gcc (tmake_file): Add name to tmake_file for 328 each @name in --with-multilib-list on arm-*-* targets. 329 * doc/install.texi (with-multilib-list): Document it. 330 3312018-12-21 Jakub Jelinek <jakub@redhat.com> 332 333 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Use 334 DR_TARGET_ALIGNMENT on dr_info rather than dr. Spelling fixes. 335 3362018-12-20 Joern Rennecke <joern.rennecke@riscy-ip.com> 337 338 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Don't do 339 versioning for data accesses with misaligned step. 340 3412018-12-20 H.J. Lu <hongjiu.lu@intel.com> 342 343 PR c/51628 344 * doc/invoke.texi: Document -Wno-address-of-packed-member. 345 3462018-12-20 Vladimir Makarov <vmakarov@redhat.com> 347 348 PR target/88457 349 * ira-color.c (fast_allocation): Choose the best cost hard reg. 350 3512018-12-20 Richard Sandiford <richard.sandiford@arm.com> 352 353 * config/aarch64/iterators.md (SVE_INT_UNARY, fp_int_op): Add abs. 354 (SVE_FP_UNARY): Sort. 355 3562018-12-20 Richard Sandiford <richard.sandiford@arm.com> 357 358 * config/aarch64/aarch64-sve.md (*cond_<optab><mode>_4): Use 359 sve_fmla_op rather than sve_fmad_op for the movprfx alternative. 360 3612018-12-20 Martin Jambor <mjambor@suse.cz> 362 363 PR ipa/88214 364 * tree-ssa-alias.c (ao_ref_init_from_ptr_and_size): Assert that 365 ptr is a pointer. 366 3672018-12-20 Richard Biener <rguenther@suse.de> 368 369 PR tree-optimization/84362 370 * tree-ssa-loop-im.c: Include alias.h, builtins.h and tree-dfa.h. 371 (struct im_mem_ref): add ref_canonical flag. 372 (struct mem_ref_hasher): Use ao_ref as compare_type. 373 (mem_ref_hasher::equal): Adjust and add variant comparing ao_ref 374 parts. 375 (mem_ref_alloc): Take ao_ref parameter, initialize ref_canonical 376 member. 377 (gather_mem_refs_stmt): Set up ao_ref early and do the lookup 378 using it. If we have non-equal refs canonicalize the one 379 in the hashtable used for insertion. 380 (tree_ssa_lim_initialize): Adjust. 381 3822018-12-20 Jakub Jelinek <jakub@redhat.com> 383 384 PR target/88547 385 * config/i386/i386.c (ix86_expand_sse_movcc): For maskcmp, try to 386 emit vpmovm2? instruction perhaps after knot?. Reorganize code 387 so that it doesn't have to test !maskcmp in almost every conditional. 388 3892018-12-20 Alan Modra <amodra@gmail.com> 390 391 * config/rs6000/sysv4.h (GNU_USER_DYNAMIC_LINKER): Define. 392 3932018-12-19 Andreas Tobler <andreast@gcc.gnu.org> 394 395 * config.gcc: Enable TARGET_DEFAULT_ASYNC_UNWIND_TABLES on 396 aarch64*-*-freebsd* 397 3982018-12-19 Uros Bizjak <ubizjak@gmail.com> 399 400 * config/i386/i386.md (SWI1248_AVX512BWDQ_64): Rename from 401 SWI1248_AVX512BWDQ2_64. Unconditionally enable HImode. 402 (*cmp<mode>_ccz_1): Emit kortest instead of ktest insn. 403 Use SWI1248_AVX512BWDQ_64 mode iterator and enable only for 404 TARGET_AVX512F. 405 4062018-12-19 Thomas Preud'homme <thomas.preudhomme@linaro.org> 407 408 * config/arm/elf.h: Update comment about condition that need to 409 match with libgcc/config/arm/lib1funcs.S to also include 410 libgcc/config/arm/t-arm. 411 * doc/sourcebuild.texi (output-exists, output-exists-not): Rename 412 subsubsection these directives are in to "Check for output files". 413 Move scan-symbol to that section and add to it new scan-symbol-not 414 directive. 415 4162018-12-19 Tom de Vries <tdevries@suse.de> 417 418 * config/nvptx/nvptx.c (PTX_CTA_SIZE): Define. 419 4202018-12-19 Tom de Vries <tdevries@suse.de> 421 422 * omp-offload.c (oacc_fn_attrib_level): Remove static. 423 * omp-offload.h (oacc_fn_attrib_level): Declare. 424 4252018-12-19 Tom de Vries <tdevries@suse.de> 426 427 * omp-offload.c (oacc_get_default_dim): New function. 428 * omp-offload.h (oacc_get_default_dim): Declare. 429 4302018-12-19 David Malcolm <dmalcolm@redhat.com> 431 432 PR c++/87504 433 * gcc-rich-location.c 434 (maybe_range_label_for_tree_type_mismatch::get_text): Move here from 435 c/c-typeck.c. 436 (binary_op_rich_location::binary_op_rich_location): New ctor. 437 (binary_op_rich_location::use_operator_loc_p): New function. 438 * gcc-rich-location.h 439 (class maybe_range_label_for_tree_type_mismatch)): Move here from 440 c/c-typeck.c. 441 (struct op_location_t): New forward decl. 442 (class binary_op_rich_location): New class. 443 * tree.h (struct op_location_t): New struct. 444 4452018-12-19 David Malcolm <dmalcolm@redhat.com> 446 447 PR c++/43064 448 PR c++/43486 449 * convert.c: Include "selftest.h". 450 (preserve_any_location_wrapper): New function. 451 (convert_to_pointer_maybe_fold): Update to handle location 452 wrappers. 453 (convert_to_real_maybe_fold): Likewise. 454 (convert_to_integer_1): Strip expr when using TREE_OVERFLOW. 455 Handle location wrappers when checking for INTEGER_CST. 456 (convert_to_integer_maybe_fold): Update to handle location 457 wrappers. 458 (convert_to_complex_maybe_fold): Likewise. 459 (selftest::test_convert_to_integer_maybe_fold): New functions. 460 (selftest::convert_c_tests): New function. 461 * convert.h (preserve_any_location_wrapper): New decl. 462 * fold-const.c (size_binop_loc): Strip location wrappers when 463 using TREE_OVERFLOW. 464 (operand_equal_p): Strip any location wrappers. 465 (integer_valued_real_p): Strip any location wrapper. 466 * selftest-run-tests.c (selftest::run_tests): Call 467 selftest::convert_c_tests. 468 * selftest.h (selftest::convert_c_tests): New decl. 469 * tree.c (build_complex): Assert that REAL and IMAG are constants. 470 (integer_zerop): Look through location wrappers. 471 (integer_onep): Likewise. 472 (integer_each_onep): Likewise. 473 (integer_all_onesp): Likewise. 474 (integer_minus_onep): Likewise. 475 (integer_pow2p): Likewise. 476 (integer_nonzerop): Likewise. 477 (integer_truep): Likewise. 478 (fixed_zerop): Likewise. 479 (real_zerop): Likewise. 480 (real_onep): Likewise. 481 (real_minus_onep): Likewise. 482 (tree_int_cst_equal): Likewise. 483 (simple_cst_equal): Treat location wrappers with non-equal source 484 locations as being unequal. 485 (uniform_integer_cst_p): Look through location wrappers. 486 (maybe_wrap_with_location): Don't create wrappers if any 487 auto_suppress_location_wrappers are active. 488 (suppress_location_wrappers): New variable. 489 (selftest::test_predicates): New test. 490 (selftest::tree_c_tests): Call it. 491 * tree.h (CONSTANT_CLASS_OR_WRAPPER_P): New macro. 492 (suppress_location_wrappers): New decl. 493 (class auto_suppress_location_wrappers): New class. 494 4952018-12-19 Paul A. Clarke <pc@us.ibm.com> 496 497 * config/rs6000/tmmintrin.h (_mm_hadds_epi16): Vector lanes swapped. 498 (_mm_hsub_epi32): Likewise. 499 (_mm_shuffle_epi8): Fix reversed interpretation of parameters. 500 (_mm_shuffle_pi8): Likewise. 501 (_mm_addubs_pi16): Likewise. 502 5032018-12-19 Thomas Preud'homme <thomas.preudhomme@linaro.org> 504 505 * config/arm/arm.h (TARGET_HARD_FLOAT): Restrict to TARGET_32BIT 506 targets. 507 * config/arm/arm.c (output_return_instruction): Only check 508 TARGET_HARD_FLOAT to decide whether FP instructions are available. 509 5102018-12-19 Tom de Vries <tom@codesourcery.com> 511 512 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization 513 dump files): Add offload-rtl. 514 5152018-12-19 Segher Boessenkool <segher@kernel.crashing.org> 516 517 PR target/88213 518 * config/rs6000/vsx.md (*vsx_extract_<P:mode>_<VSX_D:mode>_load): 519 Require TARGET_POWERPC64. 520 5212018-12-19 Richard Biener <rguenther@suse.de> 522 523 PR tree-optimization/88533 524 Revert 525 2018-04-30 Richard Biener <rguenther@suse.de> 526 527 PR tree-optimization/28364 528 PR tree-optimization/85275 529 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after 530 copying first exit test. 531 532 * tree-ssa-loop-ch.c: Include tree-phinodes.h and 533 ssa-iterators.h. 534 (should_duplicate_loop_header_p): Track whether stmt compute 535 loop invariants or values based on IVs. Apart from the 536 original loop header only duplicate blocks with exit tests 537 that are based on IVs or invariants. 538 5392018-12-19 Tom de Vries <tdevries@suse.de> 540 541 * config/nvptx/nvptx.c (nvptx_gen_shared_bcast, shared_prop_gen) 542 (nvptx_goacc_expand_accel_var): Use MAX and ROUND_UP. 543 (nvptx_assemble_value, nvptx_output_skip): Use MIN. 544 (nvptx_shared_propagate, nvptx_single, nvptx_expand_shared_addr): Use 545 MAX. 546 5472018-12-19 Tom de Vries <tdevries@suse.de> 548 549 * config/nvptx/nvptx.c (nvptx_gen_wcast): Rename as 550 nvptx_gen_warp_bcast. 551 (nvptx_gen_wcast): Rename to nvptx_gen_shared_bcast, add bool 552 vector argument, and update call to nvptx_gen_shared_bcast. 553 (propagator_fn): Add bool argument. 554 (nvptx_propagate): New bool argument, pass bool argument to fn. 555 (vprop_gen): Rename to warp_prop_gen, update call to 556 nvptx_gen_warp_bcast. 557 (nvptx_vpropagate): Rename to nvptx_warp_propagate, update call to 558 nvptx_propagate. 559 (wprop_gen): Rename to shared_prop_gen, update call to 560 nvptx_gen_shared_bcast. 561 (nvptx_wpropagate): Rename to nvptx_shared_propagate, update call 562 to nvptx_propagate. 563 (nvptx_wsync): Rename to nvptx_cta_sync. 564 (nvptx_single): Update calls to nvptx_gen_warp_bcast, 565 nvptx_gen_shared_bcast and nvptx_cta_sync. 566 (nvptx_process_pars): Likewise. 567 (write_worker_buffer): Rename as write_shared_buffer. 568 (nvptx_file_end): Update calls to write_shared_buffer. 569 (nvptx_expand_worker_addr): Rename as nvptx_expand_shared_addr. 570 (nvptx_expand_builtin): Update call to nvptx_expand_shared_addr. 571 (nvptx_get_worker_red_addr): Rename as nvptx_get_shared_red_addr. 572 (nvptx_goacc_reduction_setup): Update call to 573 nvptx_get_shared_red_addr. 574 (nvptx_goacc_reduction_fini): Likewise. 575 (nvptx_goacc_reduction_teardown): Likewise. 576 5772018-12-19 Tom de Vries <tdevries@suse.de> 578 579 * config/nvptx/nvptx.c (worker_bcast_size): Rename as 580 oacc_bcast_size. 581 (worker_bcast_align): Rename as oacc_bcast_align. 582 (worker_bcast_sym): Rename as oacc_bcast_sym. 583 (nvptx_option_override): Update usage of oacc_bcast_*. 584 (struct wcast_data_t): Rename as broadcast_data_t. 585 (nvptx_gen_wcast): Update type of data argument and usage of 586 oacc_bcast_align. 587 (wprop_gen): Update type of data_ and usage of oacc_bcast_align. 588 (nvptx_wpropagate): Update type of data and usage of 589 oacc_bcast_{sym,size}. 590 (nvptx_single): Update type of data and usage of oacc_bcast_size. 591 (nvptx_file_end): Update usage of oacc_bcast_{sym,align,size}. 592 5932018-12-19 Tom de Vries <tdevries@suse.de> 594 595 * config/nvptx/nvptx.md (nvptx_barsync): Add and handle operand. 596 * config/nvptx/nvptx.c (nvptx_wsync): Update call to gen_nvptx_barsync. 597 5982018-12-19 Tom de Vries <tdevries@suse.de> 599 600 * config/nvptx/nvptx.c (nvptx_single): Always pass false to 601 nvptx_wsync. 602 (nvptx_process_pars): Likewise. 603 6042018-12-19 Tom de Vries <tdevries@suse.de> 605 606 * config/nvptx/nvptx.c (nvptx_previous_fndecl): Declare. 607 (nvptx_set_current_function): New function. 608 (TARGET_SET_CURRENT_FUNCTION): Define. 609 6102018-12-19 Shaokun Zhang <zhangshaokun@hisilicon.com> 611 612 * config/aarch64/aarch64-cores.def (tsv110): Fix architecture. This 613 part is really Armv8.2 with some permitted Armv8.4 extensions. 614 6152018-12-19 Jakub Jelinek <jakub@redhat.com> 616 617 PR target/88541 618 * config/i386/vpclmulqdqintrin.h (_mm256_clmulepi64_epi128): Enable 619 for -mavx -mvpclmulqdq rather than just for -mavx512vl -mvpclmulqdq. 620 6212018-12-19 Alexandre Oliva <aoliva@redhat.com> 622 623 PR testsuite/86153 624 PR middle-end/83239 625 * vr-values.c 626 (vr_values::vrp_evaluate_conditional_warnv_with_ops): Extend 627 simplification of overflow tests to cover cases in which we 628 can determine the result of the comparison. 629 6302018-12-19 Bin Cheng <bin.cheng@linux.alibaba.com> 631 632 * auto-profile.c (afdo_indirect_call): Skip generating histogram 633 value if we can't find cgraph_node for then indirected callee. Save 634 profile_id of the cgraph_node in histogram value's first counter. 635 * value-prof.c (gimple_value_profile_transformations): Don't skip 636 for flag_auto_profile. 637 6382018-12-18 Vladimir Makarov <vmakarov@redhat.com> 639 640 PR rtl-optimization/87759 641 * lra-assigns.c (lra_split_hard_reg_for): Recalculate 642 non_reload_pseudos. 643 6442018-12-18 Jakub Jelinek <jakub@redhat.com> 645 646 PR target/88464 647 * config/i386/i386-builtin-types.def 648 (VOID_FTYPE_PDOUBLE_QI_V8SI_V4DF_INT, 649 VOID_FTYPE_PFLOAT_QI_V4DI_V8SF_INT, 650 VOID_FTYPE_PLONGLONG_QI_V8SI_V4DI_INT, 651 VOID_FTYPE_PINT_QI_V4DI_V8SI_INT, 652 VOID_FTYPE_PDOUBLE_QI_V4SI_V2DF_INT, 653 VOID_FTYPE_PFLOAT_QI_V2DI_V4SF_INT, 654 VOID_FTYPE_PLONGLONG_QI_V4SI_V2DI_INT, 655 VOID_FTYPE_PINT_QI_V2DI_V4SI_INT): New builtin types. 656 * config/i386/i386.c (enum ix86_builtins): Add 657 IX86_BUILTIN_SCATTERALTSIV4DF, IX86_BUILTIN_SCATTERALTDIV8SF, 658 IX86_BUILTIN_SCATTERALTSIV4DI, IX86_BUILTIN_SCATTERALTDIV8SI, 659 IX86_BUILTIN_SCATTERALTSIV2DF, IX86_BUILTIN_SCATTERALTDIV4SF, 660 IX86_BUILTIN_SCATTERALTSIV2DI and IX86_BUILTIN_SCATTERALTDIV4SI. 661 (ix86_init_mmx_sse_builtins): Fix up names of IX86_BUILTIN_GATHERALT*, 662 IX86_BUILTIN_GATHER3ALT* and IX86_BUILTIN_SCATTERALT* builtins to 663 match the IX86_BUILTIN codes. Build IX86_BUILTIN_SCATTERALTSIV4DF, 664 IX86_BUILTIN_SCATTERALTDIV8SF, IX86_BUILTIN_SCATTERALTSIV4DI, 665 IX86_BUILTIN_SCATTERALTDIV8SI, IX86_BUILTIN_SCATTERALTSIV2DF, 666 IX86_BUILTIN_SCATTERALTDIV4SF, IX86_BUILTIN_SCATTERALTSIV2DI and 667 IX86_BUILTIN_SCATTERALTDIV4SI decls. 668 (ix86_vectorize_builtin_scatter): Expand those new builtins. 669 6702018-12-18 Bill Schmidt <wschmidt@linux.ibm.com> 671 672 * doc/extend.texi (PowerPC Altivec/VSX Built-in Functions): 673 Describe when a typedef name can be used as the type specifier for 674 a vector type, and when it cannot. 675 6762018-12-18 Jozef Lawrynowicz <jozef.l@mittosystems.com> 677 678 * config/msp430/msp430.h: Define TARGET_VTABLE_ENTRY_ALIGN. 679 6802018-12-18 Jakub Jelinek <jakub@redhat.com> 681 682 PR target/88513 683 PR target/88514 684 * optabs.def (vec_pack_sbool_trunc_optab, vec_unpacks_sbool_hi_optab, 685 vec_unpacks_sbool_lo_optab): New optabs. 686 * optabs.c (expand_widen_pattern_expr): Use vec_unpacks_sbool_*_optab 687 and pass additional argument if both input and target have the same 688 scalar mode of VECTOR_BOOLEAN_TYPE_P vectors. 689 * expr.c (expand_expr_real_2) <case VEC_PACK_TRUNC_EXPR>: Handle 690 VECTOR_BOOLEAN_TYPE_P pack where result has the same scalar mode 691 as the operands using vec_pack_sbool_trunc_optab. 692 * tree-vect-stmts.c (supportable_widening_operation): Use 693 vec_unpacks_sbool_{lo,hi}_optab for VECTOR_BOOLEAN_TYPE_P conversions 694 where both wider_vectype and vectype have the same scalar mode. 695 (supportable_narrowing_operation): Similarly use 696 vec_pack_sbool_trunc_optab if narrow_vectype and vectype have the same 697 scalar mode. 698 * config/i386/i386.c (ix86_get_builtin) 699 <case IX86_BUILTIN_GATHER3ALTDIV8SF>: Check for VECTOR_MODE_P 700 rather than non-VOIDmode. 701 * config/i386/sse.md (vec_pack_trunc_qi, vec_pack_trunc_<mode>): 702 Remove useless ()s around "register_operand", formatting fixes. 703 (vec_pack_sbool_trunc_qi, vec_unpacks_sbool_lo_qi, 704 vec_unpacks_sbool_hi_qi): New expanders. 705 * doc/md.texi (vec_pack_sbool_trunc_M, vec_unpacks_sbool_hi_M, 706 vec_unpacks_sbool_lo_M): Document. 707 7082018-12-18 Jozef Lawrynowicz <jozef.l@mittosystems.com> 709 710 * combine.c (update_rsp_from_reg_equal): Only look for the nonzero bits 711 of src in nonzero_bits_mode if the mode of src is MODE_INT and 712 HWI_COMPUTABLE. 713 (reg_nonzero_bits_for_combine): Add clarification to comment. 714 7152018-12-18 Wei Xiao <wei3.xiao@intel.com> 716 717 * config/i386/driver-i386.c (host_detect_local_cpu): Detect cascadelake. 718 * config/i386/i386.c (fold_builtin_cpu): Handle cascadelake. 719 * doc/extend.texi: Add cascadelake. 7202018-12-17 Peter Bergner <bergner@linux.ibm.com> 721 722 PR target/87870 723 * config/rs6000/vsx.md (nW): New mode iterator. 724 (vsx_mov<mode>_64bit): Use it. Remove redundant GPR 0/-1 alternative. 725 Update length attribute for (<??r>, <nW>) alternative. 726 (vsx_mov<mode>_32bit): Likewise. 727 7282018-12-17 Tom de Vries <tdevries@suse.de> 729 730 * config/nvptx/nvptx.c (PTX_VECTOR_LENGTH, PTX_WORKER_LENGTH, 731 PTX_DEFAULT_RUNTIME_DIM): Move to the top of the file. 732 7332018-12-17 Tom de Vries <tdevries@suse.de> 734 735 * config/nvptx/nvptx.c (PTX_WARP_SIZE): Define. 736 (nvptx_simt_vf): Return PTX_WARP_SIZE instead of PTX_VECTOR_LENGTH. 737 7382018-12-17 Tom de Vries <tdevries@suse.de> 739 740 * config/nvptx/nvptx.c (nvptx_single): Fix whitespace. 741 (nvptx_neuter_pars): Likewise. 742 7432018-12-17 Tom de Vries <tdevries@suse.de> 744 745 * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Work around Fortran 746 bug PR72741 by overriding dims parameter for routines. 747 7482018-12-17 Tom de Vries <tdevries@suse.de> 749 750 * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Rewrite using 751 predicate vars. 752 7532018-12-17 Steve Ellcey <sellcey@cavium.com> 754 755 * config/aarch64/aarch64-protos.h (aarch64_use_simple_return_insn_p): 756 New prototype. 757 (aarch64_epilogue_uses): Ditto. 758 * config/aarch64/aarch64.c (aarch64_attribute_table): New array. 759 (aarch64_simd_decl_p): New function. 760 (aarch64_reg_save_mode): New function. 761 (aarch64_function_ok_for_sibcall): Check for simd calls. 762 (aarch64_layout_frame): Check for simd function. 763 (aarch64_gen_storewb_pair): Handle E_TFmode. 764 (aarch64_push_regs): Use aarch64_reg_save_mode to get mode. 765 (aarch64_gen_loadwb_pair): Handle E_TFmode. 766 (aarch64_pop_regs): Use aarch64_reg_save_mode to get mode. 767 (aarch64_gen_store_pair): Handle E_TFmode. 768 (aarch64_gen_load_pair): Ditto. 769 (aarch64_save_callee_saves): Handle different mode sizes. 770 (aarch64_restore_callee_saves): Ditto. 771 (aarch64_components_for_bb): Check for simd function. 772 (aarch64_epilogue_uses): New function. 773 (aarch64_process_components): Check for simd function. 774 (aarch64_expand_prologue): Ditto. 775 (aarch64_expand_epilogue): Ditto. 776 (aarch64_expand_call): Ditto. 777 (aarch64_use_simple_return_insn_p): New function. 778 (TARGET_ATTRIBUTE_TABLE): New define. 779 * config/aarch64/aarch64.h (EPILOGUE_USES): Redefine. 780 (FP_SIMD_SAVED_REGNUM_P): New macro. 781 * config/aarch64/aarch64.md (simple_return): New define_expand. 782 (load_pair_dw_tftf): New instruction. 783 (store_pair_dw_tftf): Ditto. 784 (loadwb_pair<TX:mode>_<P:mode>): Ditto. 785 (storewb_pair<TX:mode>_<P:mode>): Ditto. 786 7872018-12-17 Uros Bizjak <ubizjak@gmail.com> 788 789 PR target/88502 790 * internal-fn.def (ACOSH): New. 791 (ASINH): Ditto. 792 (ATANH): Ditto. 793 * optabs.def (acosh_optab): New. 794 (asinh_optab): Ditto. 795 (atanh_optab): Ditto. 796 * config/i386/i386-protos.h (ix86_emit_i387_asinh): New prototype. 797 (ix86_emit_i387_acosh): Ditto. 798 (ix86_emit_i387_atanh): Ditto. 799 * config/i386/i386.c (ix86_emit_i387_asinh): New function. 800 (ix86_emit_i387_acosh): Ditto. 801 (ix86_emit_i387_atanh): Ditto. 802 * config/i386/i386.md (asinhxf2): New expander. 803 (asinh<mode>2): Ditto. 804 (acoshxf2): Ditto. 805 (acosh<mode>2): Ditto. 806 (atanhxf2): Ditto. 807 (atanh<mode>2): Ditto. 808 8092018-12-17 David Edelsohn <dje.gcc@gmail.com> 810 811 * config.gcc (powerpc-ibm-aix6.*): Delete extra_headers. 812 (powerpc-ibm-aix7.1.*): Same. 813 (powerpc-ibm-aix[789].*): Same. 814 8152018-12-17 H.J. Lu <hongjiu.lu@intel.com> 816 817 PR debug/79342 818 * dwarf2out.c (find_AT_string_in_table): Add insert argument 819 defaulting to INSERT and replace INSERT. 820 (find_AT_string): Likewise. 821 (output_macinfo_op): Pass NO_INSERT to find_AT_string. 822 8232018-12-15 Jan Hubicka <hubicka@ucw.cz> 824 825 * coverage.c (struct conts_entry): Add n_counts. 826 (remap_counts_file): Record number of ocunts. 827 (get_coverage_counts): Verify that counts match. 828 * coverage.h (get_coverage_counts): Update prototype. 829 * profile.c (get_exec_counts. compute_value_histograms): Add 830 n_counts parametrs. 831 8322018-12-17 Senthil Kumar Selvaraj <senthilkumar.selvaraj@microchip.com> 833 834 PR rtl-optimization/88253 835 * combine.c (combine_simplify_rtx): Test for side-effects before 836 substituting by zero. 837 8382018-12-17 Richard Sandiford <richard.sandiford@arm.com> 839 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com> 840 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 841 842 * doc/invoke.texi (-fversion-loops-for-strides): Document 843 (loop-versioning-group-size, loop-versioning-max-inner-insns) 844 (loop-versioning-max-outer-insns): Document new --params. 845 * Makefile.in (OBJS): Add gimple-loop-versioning.o. 846 * common.opt (fversion-loops-for-strides): New option. 847 * opts.c (default_options_table): Enable fversion-loops-for-strides 848 at -O3. 849 * params.def (PARAM_LOOP_VERSIONING_GROUP_SIZE) 850 (PARAM_LOOP_VERSIONING_MAX_INNER_INSNS) 851 (PARAM_LOOP_VERSIONING_MAX_OUTER_INSNS): New parameters. 852 * passes.def: Add pass_loop_versioning. 853 * timevar.def (TV_LOOP_VERSIONING): New time variable. 854 * tree-ssa-propagate.h 855 (substitute_and_fold_engine::substitute_and_fold): Add an optional 856 block parameter. 857 * tree-ssa-propagate.c 858 (substitute_and_fold_engine::substitute_and_fold): Likewise. 859 When passed, only walk blocks dominated by that block. 860 * tree-vrp.h (range_includes_p): Declare. 861 (range_includes_zero_p): Turn into an inline wrapper around 862 range_includes_p. 863 * tree-vrp.c (range_includes_p): New function, generalizing... 864 (range_includes_zero_p): ...this. 865 * tree-pass.h (make_pass_loop_versioning): Declare. 866 * gimple-loop-versioning.cc: New file. 867 8682018-12-15 Jan Hubicka <hubicka@ucw.cz> 869 870 * ipa-fnsummary.c (remap_edge_change_prob): Do not ICE when changes 871 are not streamed in. 872 8732018-12-15 H.J. Lu <hongjiu.lu@intel.com> 874 875 PR target/88483 876 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Revert 877 reversion 267133. 878 8792018-12-15 Jan Hubicka <hubicka@ucw.cz> 880 881 * ipa-fnsummary.c (analyze_function_body): Do not loeak conds and 882 size_time_table. 883 (ipa_fn_summary_generate): Add prevails parameter; do not allocate 884 data when symbol is not prevailing. 885 (inline_read_section): Likewise. 886 8872018-12-15 Jan Hubicka <hubicka@ucw.cz> 888 889 * cgraph.h (cgraph_node): Add predicate prevailing_p. 890 (cgraph_edge): Add predicate possible_call_in_translation_unit_p. 891 * ipa-prop.c (ipa_write_jump_function): Optimize streaming of ADDR_EXPR. 892 (ipa_read_jump_function): Add prevails parameter; optimize streaming. 893 (ipa_read_edge_info): Break out from ... 894 (ipa_read_node_info): ... here; optimize streaming. 895 * cgraph.c (cgraph_edge::possibly_call_in_translation_unit_p): New 896 predicate. 897 8982018-12-15 Jan Hubicka <hubicka@ucw.cz> 899 900 * ipa-utils.c (ipa_merge_profiles): Do no merging when source function 901 has zero count. 902 9032018-12-15 Jakub Jelinek <jakub@redhat.com> 904 905 PR tree-optimization/88464 906 PR target/88498 907 * tree-vect-stmts.c (vect_build_gather_load_calls): For NARROWING 908 and mask with integral masktype, don't try to permute mask vectors, 909 instead emit VEC_UNPACK_{LO,HI}_EXPR. Fix up NOP_EXPR operand. 910 (vectorizable_store): Handle masked scatters with decl and integral 911 mask type. 912 (permute_vec_elements): Allow scalar_dest to be NULL. 913 * config/i386/i386.c (ix86_get_builtin) 914 <case IX86_BUILTIN_GATHER3ALTDIV16SF>: Use lowpart_subreg for masks. 915 <case IX86_BUILTIN_GATHER3ALTDIV8SF>: Don't assume mask and src have 916 to be the same. 917 9182018-12-15 Jan Hubicka <hubicka@ucw.cz> 919 920 * ipa.c (cgraph_build_static_cdtor_1): Add OPTIMIZATION and TARGET 921 parameters. 922 (cgraph_build_static_cdtor): Update. 923 (build_cdtor): Use OPTIMIZATION and TARGET of the first real cdtor 924 callsed. 925 9262018-12-15 Alan Modra <amodra@gmail.com> 927 928 * config/rs6000/rs6000.c (generate_set_vrsave, rs6000_emit_savres_rtx), 929 (rs6000_emit_prologue, rs6000_call_aix, rs6000_call_sysv), 930 (rs6000_call_darwin_1): Use gen_hard_reg_clobber. 931 9322018-12-15 Jakub Jelinek <jakub@redhat.com> 933 934 PR target/88489 935 * config/i386/sse.md (UNSPEC_SFIXUPIMM): New unspec enumerator. 936 (avx512f_sfixupimm<mode><mask_name><round_saeonly_name>): Use it 937 instead of UNSPEC_FIXUPIMM. 938 939 PR rtl-optimization/88478 940 * cselib.c (cselib_record_sets): Move sets[i].src_elt tests 941 after REG_P (dest) test. 942 9432018-12-14 Martin Sebor <msebor@redhat.com> 944 945 PR tree-optimization/88372 946 * calls.c (maybe_warn_alloc_args_overflow): Handle function pointers. 947 * tree-object-size.c (alloc_object_size): Same. Simplify. 948 * doc/extend.texi (Object Size Checking): Update. 949 (Other Builtins): Add __builtin_object_size. 950 (Common Type Attributes): Add alloc_size. 951 (Common Variable Attributes): Ditto. 952 9532018-12-14 Martin Sebor <msebor@redhat.com> 954 955 PR rtl-optimization/87096 956 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Avoid 957 folding calls whose bound may exceed INT_MAX. Diagnose bound ranges 958 that exceed the limit. 959 9602018-12-14 Martin Sebor <msebor@redhat.com> 961 962 PR web/79738 963 * doc/extend.texi (attribute const, pure): Clarify. 964 9652018-12-14 H.J. Lu <hongjiu.lu@intel.com> 966 967 * config/i386/i386.c (rest_of_insert_endbranch): Insert ENDBR 968 at the function entry only when -mmanual-endbr isn't used or 969 there is cf_check function attribute. 970 (ix86_attribute_table): Add cf_check. 971 * config/i386/i386.opt: Add -mmanual-endbr. 972 * doc/extend.texi: Document cf_check attribute. 973 * doc/invoke.texi: Document -mmanual-endbr. 974 9752018-12-14 Thomas Schwinge <thomas@codesourcery.com> 976 Cesar Philippidis <cesar@codesourcery.com> 977 978 * omp-offload.c (inform_oacc_loop): New function. 979 (execute_oacc_device_lower): Use it to display loop parallelism. 980 9812018-12-14 Jakub Jelinek <jakub@redhat.com> 982 983 PR c++/82294 984 PR c++/87436 985 * expr.h (categorize_ctor_elements): Add p_unique_nz_elts argument. 986 * expr.c (categorize_ctor_elements_1): Likewise. Compute it like 987 p_nz_elts, except don't multiply it by mult. Adjust recursive call. 988 Fix up COMPLEX_CST handling. 989 (categorize_ctor_elements): Add p_unique_nz_elts argument, initialize 990 it and pass it through to categorize_ctor_elements_1. 991 (mostly_zeros_p, all_zeros_p): Adjust categorize_ctor_elements callers. 992 * gimplify.c (gimplify_init_constructor): Likewise. Don't force 993 ctor into readonly data section if num_unique_nonzero_elements is 994 smaller or equal to 1/8 of num_nonzero_elements and size is >= 64 995 bytes. 996 9972018-12-14 Thomas Preud'homme <thomas.preudhomme@arm.com> 998 999 * config/arm/arm.md (arm_movdi): Split if -mslow-flash-data and 1000 source is a constant that would be loaded by literal pool. 1001 (movsf expander): Generate a no_literal_pool_sf_immediate insn if 1002 -mslow-flash-data is present, targeting hardfloat ABI and source is a 1003 float constant that cannot be loaded via vmov. 1004 (movdf expander): Likewise but generate a no_literal_pool_df_immediate 1005 insn. 1006 (arm_movsf_soft_insn): Split if -mslow-flash-data and source is a 1007 float constant that would be loaded by literal pool. 1008 (softfloat constant movsf splitter): Splitter for the above case. 1009 (movdf_soft_insn): Split if -mslow-flash-data and source is a float 1010 constant that would be loaded by literal pool. 1011 (softfloat constant movdf splitter): Splitter for the above case. 1012 * config/arm/constraints.md (Pz): Document existing constraint. 1013 (Ha): Define constraint. 1014 (Tu): Likewise. 1015 * config/arm/predicates.md (hard_sf_operand): New predicate. 1016 (hard_df_operand): Likewise. 1017 * config/arm/thumb2.md (thumb2_movsi_insn): Split if 1018 -mslow-flash-data and constant would be loaded by literal pool. 1019 * constant/arm/vfp.md (thumb2_movsi_vfp): Likewise and disable constant 1020 load in VFP register. 1021 (movdi_vfp): Likewise. 1022 (thumb2_movsf_vfp): Use hard_sf_operand as predicate for source to 1023 prevent match for a constant load if -mslow-flash-data and constant 1024 cannot be loaded via vmov. Adapt constraint accordingly by 1025 using Ha instead of E for generic floating-point constant load. 1026 (thumb2_movdf_vfp): Likewise using hard_df_operand predicate instead. 1027 (no_literal_pool_df_immediate): Add a clobber to use as the 1028 intermediate general purpose register and also enable it after reload 1029 but disable it constant is a valid FP constant. Add constraints and 1030 generate a DI intermediate load rather than 2 SI loads. 1031 (no_literal_pool_sf_immediate): Add a clobber to use as the 1032 intermediate general purpose register and also enable it after 1033 reload. 1034 10352018-12-14 Uros Bizjak <ubizjak@gmail.com> 1036 1037 PR target/88474 1038 * internal-fn.def (HYPOT): New. 1039 * optabs.def (hypot_optab): New. 1040 * config/i386/i386.md (hypot<mode>3): New expander. 1041 10422018-12-14 Jeff Law <law@redhat.com> 1043 1044 * target.def (post_cfi_startproc): Fix text. 1045 10462018-12-14 Sam Tebbs <sam.tebbs@arm.com> 1047 1048 * doc/tm.texi (TARGET_ASM_POST_CFI_STARTPROC): Define. 1049 * doc/tm.texi.in (TARGET_ASM_POST_CFI_STARTPROC): Define. 1050 * dwarf2out.c (dwarf2out_do_cfi_startproc): Trigger the hook. 1051 * hooks.c (hook_void_FILEptr_tree): Define. 1052 * hooks.h (hook_void_FILEptr_tree): Define. 1053 * target.def (post_cfi_startproc): Define. 1054 10552018-12-14 Tom de Vries <tdevries@suse.de> 1056 1057 * lto-cgraph.c (verify_node_partition): New function. 1058 (input_overwrite_node, input_varpool_node): Use verify_node_partition. 1059 10602018-12-14 H.J. Lu <hongjiu.lu@intel.com> 1061 1062 PR target/88483 1063 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Don't 1064 use get_frame_size (). 1065 10662018-12-14 Andrew Stubbs <ams@codesourcery.com> 1067 1068 * gcc/lra-int.h (lra_register_new_scratch_op): Add third parameter. 1069 * gcc/lra-remat.c (update_scratch_ops): Pass icode to 1070 lra_register_new_scratch_op. 1071 * gcc/lra.c (struct sloc): Add icode field. 1072 (lra_register_new_scratch_op): Add icode parameter. 1073 Use icode to skip insns that have changed beyond recognition. 1074 10752018-12-14 Eric Botcazou <ebotcazou@adacore.com> 1076 1077 * dwarf2out.c (analyze_discr_in_predicate): Simplify. 1078 (analyze_variants_discr): Deal with naked boolean discriminants. 1079 10802018-12-14 Claudiu Zissulescu <claziss@synopsys.com> 1081 1082 * config/arc/arc.c (arc_restore_callee_milli) Don't clobber off 1083 variable. 1084 10852018-12-14 Claudiu Zissulescu <claziss@synopsys.com> 1086 1087 * config/arc/arc.h (reg_class): Add SIBCALL_REGS. 1088 10892018-12-14 Richard Biener <rguenther@suse.de> 1090 1091 * gimple-ssa-split-paths.c (is_feasible_trace): Remove 1092 duplicated code block. 1093 10942018-12-14 Alan Modra <amodra@gmail.com> 1095 1096 PR rtl-optimization/88311 1097 * config/rs6000/rs6000.c (rs6000_call_sysv): Do not mask cookie. 1098 (rs6000_sibcall_sysv): Likewise. 1099 11002018-12-14 Segher Boessenkool <segher@kernel.crashing.org> 1101 1102 PR rtl-optimization/88001 1103 * function.c (match_asm_constraints_1): Don't invalidly share RTL. 1104 11052018-12-14 Bin Cheng <bin.cheng@linux.alibaba.com> 1106 1107 * auto-profile.c (afdo_annotate_cfg): Call update_max_bb_count even 1108 if autofdo counts are all zeros. 1109 11102018-12-14 Bin Cheng <bin.cheng@linux.alibaba.com> 1111 1112 * auto-profile.c (afdo_indirect_call): Skip generating histogram 1113 value for internal call. 1114 11152018-12-14 Bin Cheng <bin.cheng@linux.alibaba.com> 1116 1117 * auto-profile.c (AFDO_EINFO): New macro. 1118 (class edge_info): New class. 1119 (is_edge_annotated, set_edge_annotated): Delete. 1120 (afdo_propagate_edge, afdo_propagate_circuit, afdo_propagate): Remove 1121 parameter. Adjust edge count computation and annotation using class 1122 edge_info. 1123 (afdo_calculate_branch_prob, afdo_annotate_cfg): Likewise. 1124 11252018-12-13 Michael Ploujnikov <michael.ploujnikov@oracle.com> 1126 1127 * ipa-cp.c (print_all_lattices): Skip cp clones. 1128 1129 * ipa-cp.c: Fix various comment typos. 1130 11312018-12-13 Jakub Jelinek <jakub@redhat.com> 1132 1133 PR tree-optimization/88444 1134 * tree-inline.c (fold_marked_statements): Iterate up to 1135 last_basic_block_for_fn rather than n_basic_blocks_for_fn. 1136 1137 PR rtl-optimization/88470 1138 * cfgcleanup.c (outgoing_edges_match): If the function is 1139 shrink-wrapped and bb1 ends with a JUMP_INSN with a single fake 1140 edge to EXIT, return false. 1141 11422018-12-13 Martin Sebor <msebor@redhat.com> 1143 1144 * doc/extend.texi: Consistently use @code for const and volatile 1145 qualifiers, the true and false constants, and asm statements. 1146 11472018-12-13 Vladimir Makarov <vmakarov@redhat.com> 1148 1149 PR rtl-optimization/88414 1150 * lra-int.h (lra_asm_error_p): New. 1151 * lra-assigns.c (lra_assign): Check lra_asm_error_p for checking 1152 call crossed pseudo assignment correctness. 1153 (lra_split_hard_reg_for): Set up lra_asm_error_p. 1154 * lra-constraints.c (curr_insn_transform): Ditto. 1155 * lra.c (lra_asm_error_p): New. 1156 (lra): Reset lra_asm_error_p. 1157 11582018-12-13 Jakub Jelinek <jakub@redhat.com> 1159 1160 PR rtl-optimization/88416 1161 * valtrack.c (cleanup_auto_inc_dec): Handle pre/post-inc/dec/modify 1162 even if !AUTO_INC_DEC. 1163 11642018-12-13 David Edelsohn <dje.gcc@gmail.com> 1165 1166 * config/rs6000/rs6000.c (rs6000_function_arg): Ensure type is 1167 non-NULL. 1168 (rs6000_arg_partial_bytes): Same. 1169 11702018-12-13 Jason Merrill <jason@redhat.com> 1171 1172 * gdbinit.in (pp): New macro. 1173 (pbb): Remove. 1174 11752018-12-13 Jakub Jelinek <jakub@redhat.com> 1176 1177 PR tree-optimization/88464 1178 * tree-vect-stmts.c (vect_build_gather_load_calls): Handle INTEGER_TYPE 1179 masktype if mask is non-NULL. 1180 (vectorizable_load): Don't reject masked gather loads if masktype 1181 in the decl is INTEGER_TYPE. 1182 11832018-12-13 Andreas Krebbel <krebbel@linux.ibm.com> 1184 1185 * config/s390/s390-builtins.def (s390_vec_double_s64): Map to 1186 s390_vec_double_s64 instead of s390_vcdgb. 1187 (s390_vec_double_u64): Map to s390_vec_double_u64 instead of 1188 s390_vcdlgb. 1189 11902018-12-13 Andreas Krebbel <krebbel@linux.ibm.com> 1191 1192 * config/s390/vx-builtins.md ("vec_ctd_s64", "vec_ctd_u64") 1193 ("vec_ctsl", "vec_ctul"): Replace 0 with VEC_NOINEXACT. 1194 ("vec_double_s64", "vec_double_u64"): Replace 4 with VEC_INEXACT. 1195 11962018-12-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 1197 1198 * doc/invoke.texi (-msve-vector-bits): Clarify -msve-vector-bits=128 1199 behavior. 1200 12012018-12-13 Wei Xiao <wei3.xiao@intel.com> 1202 1203 * common/config/i386/i386-common.c (processor_names): Add cascadelake. 1204 (processor_alias_table): Add cascadelake. 1205 * config.gcc: Add -march=cascadelake. 1206 * config/i386/i386-c.c (ix86_target_macros_internal): Handle 1207 cascadelake. 1208 * config/i386/i386.c (Add m_CASCADELAKE): New. 1209 (processor_cost_table): Add cascadelake. 1210 (get_builtin_code_for_version): Handle cascadelake. 1211 * config/i386/i386.h (TARGET_CASCADELAKE, PROCESSOR_CASCADELAKE): New. 1212 (PTA_CASCADELAKE): Ditto. 1213 * doc/invoke.texi: Add -march=cascadelake. 1214 12152018-12-13 Jakub Jelinek <jakub@redhat.com> 1216 1217 PR target/88465 1218 * config/i386/i386.md (*movdi_internal, *movsi_internal, 1219 *movhi_internal, *movqi_internal): Add alternative(s) to load 1220 0 or -1 into k registers using kxor or kxnoq instructions. 1221 1222 PR target/88461 1223 * config/i386/sse.md (VI1248_AVX512VLBW, AVX512ZEXTMASK): New 1224 mode iterators. 1225 (<avx512>_testm<mode>3<mask_scalar_merge_name>, 1226 <avx512>_testnm<mode>3<mask_scalar_merge_name>): Merge patterns 1227 with VI12_AVX512VL and VI48_AVX512VL iterators into ones with 1228 VI1248_AVX512VLBW iterator. 1229 (*<avx512>_testm<VI1248_AVX512VLBW:mode>3_zext, 1230 *<avx512>_testm<VI1248_AVX512VLBW:mode>3_zext_mask, 1231 *<avx512>_testnm<VI1248_AVX512VLBW:mode>3_zext, 1232 *<avx512>_testnm<VI1248_AVX512VLBW:mode>3_zext_mask): New 1233 define_insns. 1234 1235 PR target/88461 1236 * config/i386/i386.md (*zero_extendsidi2, zero_extend<mode>di2, 1237 *zero_extend<mode>si2, *zero_extendqihi2): Add =*k, *km alternatives. 1238 12392018-12-12 Tom de Vries <tdevries@suse.de> 1240 Jakub Jelinek <jakub@redhat.com> 1241 1242 * omp-builtins.def (BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START, 1243 BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START, 1244 BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START, 1245 BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START, 1246 BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME, 1247 BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME): Fix up 1248 function types - remove one argument. 1249 12502018-12-12 Martin Sebor <msebor@redhat.com> 1251 1252 * doc/extend.texi (attribute copy): Fix typos. 1253 12542018-12-12 Martin Sebor <msebor@redhat.com> 1255 1256 * doc/extend.texi (Function Attributes): Clarify C++ aspects. 1257 (Variable Attributes): Same. 1258 (Type Attributes): Same. 1259 12602018-12-12 Olivier Hainque <hainque@adacore.com> 1261 1262 * config/aarch64/aarch64.c (aarch64_override_options): Once arch, 1263 cpu and tune were validated, insert SUBTARGET_OVERRIDE_OPTIONS if 1264 defined. 1265 12662018-12-12 Olivier Hainque <hainque@adacore.com> 1267 1268 * config/aarch64/aarch64.c (PROBE_STACK_FIRST_REG) : Redefine as 1269 R9_REGNUM instead of 9. 1270 (PROBE_STACK_SECOND_REG): Redefine as R10_REGNUM instead of 10. 1271 12722018-12-12 Eric Botcazou <ebotcazou@adacore.com> 1273 1274 PR target/86806 1275 * config/sparc/sparc.md (unspecv): Add UNSPECV_SPECULATION_BARRIER. 1276 (speculation_barrier): New instruction for V9. 1277 12782018-12-12 Segher Boessenkool <segher@kernel.crashing.org> 1279 Iain Sandoe <iain@sandoe.co.uk> 1280 1281 PR target/88343 1282 * config/rs6000/rs6000.c (save_reg_p): Do not save the picbase reg 1283 unless it has been used. 1284 (first_reg_to_save): Remove dead code. 1285 12862018-12-12 Iain Sandoe <iain@sandoe.co.uk> 1287 1288 * config/rs6000/rs6000-protos.h (rs6000_call_darwin): New. 1289 (rs6000_sibcall_darwin): New. (macho_call_template): Remove. 1290 * config/rs6000/rs6000.c (get_prev_label): Forward declaration. 1291 (rs6000_call_template_1): Handle Darwin. 1292 (macho_call_template): Remove. 1293 (rs6000_call_sysv): Remove handling for Darwin. 1294 (rs6000_call_darwin_1, rs6000_call_darwin): New 1295 (rs6000_sibcall_darwin): New. 1296 * config/rs6000/rs6000.md (define_expand “call”): Handle Darwin 1297 with its own expander. (define_expand “call_value”): Likewise. 1298 (define_expand “sibcall”): Likewise. 1299 (define_expand “sibcall_value”): Likewise. 1300 (call_nonlocal_sysv): Remove Darwin special-casing. 1301 (call_value_nonlocal_sysv): Likewise. 1302 13032018-12-12 Iain Sandoe <iain@sandoe.co.uk> 1304 1305 * config/rs6000/darwin.md (call_indirect_nonlocal_darwin64): Remove. 1306 (call_nonlocal_darwin64): Remove. 1307 (call_value_indirect_nonlocal_darwin64): Remove. 1308 (call_value_nonlocal_darwin64): Remove. 1309 * config/rs6000/rs6000.c (rs6000_call_template_1): Handle Darwin with 1310 the same asm output as AIX/ELFv2. (rs6000_call_sysv): Preserve the 1311 CALL_LONG flag when needed for Darwin. 1312 * config/rs6000/rs6000.md (define expand “call”): Expand Darwin as 1313 per sysv. 1314 (define_expand “call_value”): Likewise. 1315 (define_expand “sibcall”): Likewise. 1316 (define_expand “sibcall_value”): Likewise. 1317 (call_indirect_nonlocal_sysv): Mark the clobber mode P. 1318 (call_nonlocal_sysv): Likewise. 1319 (call_nonlocal_sysv_secure): Likewise. 1320 (call_value_indirect_nonlocal_sysv): Likewise. 1321 (call_value_nonlocal_sysv): Likewise. 1322 (call_value_nonlocal_sysv_secure): Likewise. 1323 (call_local64): Mark the clobber DI. 1324 (call_value_local64): Likewise. 1325 13262018-12-11 Jakub Jelinek <jakub@redhat.com> 1327 1328 PR tree-optimization/88444 1329 * tree-vrp.c (register_edge_assert_for_2): Only register assertions 1330 for conversions if rhs1 is a SSA_NAME. 1331 13322018-12-11 Dimitar Dimitrov <dimitar@dinux.eu> 1333 1334 * cfgexpand.c (asm_clobber_reg_is_valid): Also produce 1335 error when stack pointer is clobbered. 1336 (expand_asm_stmt): Refactor clobber check in separate function. 1337 13382018-12-11 Eric Botcazou <ebotcazou@adacore.com> 1339 1340 * config/rs6000/vxworks.h (RS6000_STARTING_FRAME_OFFSET): Define, 1341 accounting for STACK_BOUNDARY 128. 1342 (STACK_DYNAMIC_OFFSET): Likewise. 1343 13442018-12-11 Jakub Jelinek <jakub@redhat.com> 1345 1346 PR target/88425 1347 * config/i386/i386.md (*x86_mov<SWI48:mode>cc_0_m1_neg_leu<SWI:mode>): 1348 New define_insn_and_split. 1349 13502018-12-11 Richard Biener <rguenther@suse.de> 1351 1352 PR middle-end/88448 1353 PR middle-end/88415 1354 * gimple.c (gimple_assign_set_rhs_with_ops): Revert previous 1355 change. 1356 * tree-complex.c (update_complex_assignment): Properly transfer 1357 or clean EH info around gimple_assign_set_rhs_with_ops. 1358 13592018-12-11 Richard Earnshaw <rearnsha@arm.com> 1360 1361 PR target/87369 1362 * config/aarch64/iterators.md (sizem1): Add sizes for SFmode and DFmode. 1363 (Vbtype): Add SFmode mapping. 1364 * config/aarch64/aarch64.md (copysigndf3, copysignsf3): Delete. 1365 (copysign<GPF:mode>3): New expand pattern. 1366 (copysign<GPF:mode>3_insn): New insn pattern. 1367 13682018-12-11 Richard Biener <rguenther@suse.de> 1369 1370 * ccmp.c (ccmp_candidate_p): Use GIMPLE API properly. 1371 13722018-12-11 Jakub Jelinek <jakub@redhat.com> 1373 1374 PR lto/86004 1375 * doc/sourcebuild.texi (lto_incremental): Document new effective 1376 target. 1377 13782018-12-11 Segher Boessenkool <segher@kernel.crashing.org> 1379 1380 PR target/88145 1381 * config/rs6000/rs6000.c (rs6000_expand_zeroop_builtin): Use 1382 rs6000_isa_flags instead of rs6000_isa_flags_explicit to decide 1383 whether soft float is enabled. 1384 (rs6000_expand_mtfsb_builtin): Ditto. 1385 (rs6000_expand_set_fpscr_rn_builtin): Ditto. 1386 (rs6000_expand_set_fpscr_drn_builtin): Ditto. 1387 13882018-12-10 Jeff Law <law@redhat.com> 1389 1390 PR tree-optimization/80520 1391 * gimple-ssa-split-paths.c (is_feasible_trace): Recognize half 1392 diamonds that are likely if convertable. 1393 13942018-12-10 Martin Sebor <msebor@redhat.com> 1395 1396 PR tree-optimization/86196 1397 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use 1398 base size only of arrays. 1399 14002018-12-10 Segher Boessenkool <segher@kernel.crashing.org> 1401 1402 * config.gcc (Obsolete configurations): Delete powerpc*-*-*spe*. 1403 (Unsupported targets): Add powerpc*-*-*spe*. 1404 (powerpc*-*-*spe*): Delete. 1405 (powerpc-*-eabispe*): Delete. 1406 (powerpc-*-rtems*spe*): Delete. 1407 (powerpc*-*-linux*spe*): Delete. 1408 (powerpc*-*-linux*): Do not handle the linux*spe* targets. 1409 (powerpc-wrs-vxworks*spe): Delete. 1410 (with_cpu setting code): Delete powerpc*-*-*spe* handling. 1411 * config.host (target powerpc*-*-*spe*): Delete. 1412 * doc/invoke.texi (PowerPC SPE Options): Delete. 1413 (PowerPC SPE Options): Delete. 1414 * config/powerpcspe: Delete. 1415 14162018-12-10 Uros Bizjak <ubizjak@gmail.com> 1417 1418 PR target/88418 1419 * config/i386/i386.c (ix86_expand_sse_cmp): For vector modes, 1420 check operand 1 with vector_operand predicate. 1421 (ix86_expand_sse_movcc): For vector modes, check op_true with 1422 vector_operand, not nonimmediate_operand. 1423 14242018-12-10 Richard Biener <rguenther@suse.de> 1425 1426 * tree-affine.c (tree_to_aff_combination): Remove unreachable 1427 MEM_REF case. 1428 (aff_combination_expand): Cache on SSA names, not possibly 1429 on conversion trees. Avoid expanding cases we do not handle. 1430 14312018-12-10 Richard Biener <rguenther@suse.de> 1432 1433 PR tree-optimization/88427 1434 * vr-values.c (vr_values::extract_range_from_phi_node): 1435 Handle symbolic ranges conservatively when trying to drop 1436 to Inf +- 1. 1437 14382018-12-10 Martin Jambor <mjambor@suse.cz> 1439 1440 PR ipa/88214 1441 * ipa-prop.c (determine_locally_known_aggregate_parts): Make sure 1442 we check pointers against pointers. 1443 14442018-12-10 Richard Biener <rguenther@suse.de> 1445 1446 PR middle-end/88415 1447 * gimple.c (gimple_assign_set_rhs_with_ops): Transfer EH 1448 info to a newly allocated stmt. 1449 14502018-12-10 Jerome Lambourg <lambourg@adacore.com> 1451 1452 * config/vxworksae.h (TARGET_VXWORKS_HAVE_CTORS_DTORS): Define. 1453 14542018-12-10 Fredrik Nyström <fredrik@lysator.liu.se> 1455 1456 PR bootstrap/65725 1457 * config/sol2.h: Only use libgcc-unwind.map if 1458 ENABLE_SHARED_LIBGCC. 1459 14602018-12-10 Bin Cheng <bin.cheng@linux.alibaba.com> 1461 1462 * auto-profile.c (afdo_calculate_branch_prob): Convert profile_count 1463 to afdo precision quality. 1464 (afdo_callsite_hot_enough_for_early_inline): Likewise. 1465 (afdo_annotate_cfg): Set profile_count for exit basic block. 1466 14672018-12-10 Bin Cheng <bin.cheng@linux.alibaba.com> 1468 1469 * auto-profile.c (string_table::get_index_by_decl) 1470 (function_instance::get_function_instance_by_decl): Avoid infinite 1471 recursion by using DECL_FROM_INLINE. 1472 14732018-12-09 John David Anglin <danglin@gcc.gnu.org> 1474 1475 * config.gcc (hppa*-*-linux*): Add pa/t-pa to tmake_file. Define 1476 d_target_objs. 1477 * config/pa/pa-protos.h (pa_d_target_versions): New prototype. 1478 * config/pa/pa.h (TARGET_D_CPU_VERSIONS): Define. 1479 * config/pa/pa-d.c: New file. 1480 * config/pa/t-pa: New file. 1481 14822018-12-08 Bin Cheng <bin.cheng@linux.alibaba.com> 1483 1484 * ipa-cp.c (update_profiling_info): Call adjust_for_ipa_scaling for 1485 zero profile count. 1486 14872018-12-08 Jakub Jelinek <jakub@redhat.com> 1488 1489 PR fortran/88304 1490 * tree-nested.c (convert_local_reference_stmt): Handle clobbers where 1491 lhs is not a decl normally, don't call use_pointer_in_frame on that 1492 lhs. 1493 14942018-12-08 Eric Botcazou <ebotcazou@adacore.com> 1495 1496 PR rtl-optimization/88390 1497 * dwarf2cfi.c (struct dw_cfi_row): Add window_save field. 1498 (cfi_row_equal_p): Compare it. 1499 (dwarf2out_frame_debug_cfa_window_save): Add FAKE parameter. 1500 If FAKE is false, set window_save of the current row. 1501 (dwarf2out_frame_debug) <REG_CFA_TOGGLE_RA_MANGLE>: Call above 1502 function with FAKE parameter set to true. 1503 <REG_CFA_WINDOW_SAVE>: Likewise but with FAKE parameter set to false. 1504 (change_cfi_row): Emit a DW_CFA_GNU_window_save if necessary. 1505 15062018-12-07 Peter Bergner <bergner@linux.ibm.com> 1507 1508 PR target/87496 1509 * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow 1510 -mabi=ieeelongdouble and -mabi=ibmlongdouble without -mlong-double-128. 1511 Do not error for -mabi=ibmlongdouble and no ISA 2.06 support. 1512 * doc/invoke.texi: Document -mabi=ibmlongdouble and -mabi=ieeelongdouble 1513 require -mlong-double-128. 1514 15152018-12-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 1516 1517 * config/aarch64/aarch64.c (aarch64_tuning_override_functions): Add 1518 sve_width entry. 1519 (aarch64_parse_sve_width_string): Define. 1520 15212018-12-07 Jeff Law <law@redhat.com> 1522 1523 PR middle-end/87813 1524 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Add 1525 m_update_global_ranges member. Add corresponding argument to ctor. 1526 * gimple-ssa-evrp-analyze.c 1527 (evrp_range_analyzer::evrp_range_analyzer): Add new argument and 1528 initialize m_update_global_ranges. 1529 (evrp_range_analyzer::set_ssa_range_info): Assert that we are 1530 updating global ranges. 1531 (evrp_range_analyzer::record_ranges_from_incoming_edge): Only 1532 update global ranges if explicitly requested. 1533 (evrp_range_analyzer::record_ranges_from_phis): Similarly. 1534 (evrp_range_analyzer::record_ranges_from_stmt): Similarly. 1535 * gimple-ssa-evrp.c (evrp_dom_walker): Pass new argument to 1536 evrp_range_analyzer ctor. 1537 * gimple-ssa-sprintf.c (sprintf_dom_walker): Similarly. 1538 * tree-ssa-dom.c (dom_opt_dom_walker): Similarly. 1539 15402018-12-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 1541 1542 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): 1543 Add SVE_NOT_IMPLEMENTED value. 1544 * config/aarch64/aarch64-protos.h (struct tune_params): Add sve_width 1545 field. 1546 * config/aarch64/aarch64.c (generic_tunings,cortexa35_tunings, 1547 cortexa53_tunings, cortexa57_tunings, cortexa72_tunings, 1548 cortexa73_tunings, exynosm1_tunings, thunderx_tunings, 1549 thunderx_tunings, tsv110_tunings, xgene1_tunings, qdf24xx_tunings, 1550 saphira_tunings, thunderx2t99_tunings, emag_tunings): 1551 Specify sve_width. 1552 (aarch64_estimated_poly_value): Define. 1553 (TARGET_ESTIMATED_POLY_VALUE): Define. 1554 15552018-12-07 Paul A. Clarke <pc@us.ibm.com> 1556 1557 PR target/88408 1558 * config/rs6000/mmintrin.h (_mm_packs_pu16): Correctly use "__vector". 1559 15602018-12-07 Vladimir Makarov <vmakarov@redhat.com> 1561 1562 PR rtl-optimization/88349 1563 * ira-costs.c (record_operand_costs): Check bigger reg class on 1564 NO_REGS. 1565 15662018-12-07 Richard Sandiford <richard.sandiford@arm.com> 1567 1568 * config/aarch64/aarch64-sve.md (*mul<mode>3, *v<optab><mode>3): 1569 Split the patterns after reload if we don't need the predicate 1570 operand. 1571 (*post_ra_mul<mode>3, *post_ra_v<optab><mode>3): New patterns. 1572 15732018-12-07 Richard Sandiford <richard.sandiford@arm.com> 1574 1575 * config/aarch64/iterators.md (SVE_UNPRED_FP_BINARY): New code 1576 iterator. 1577 (sve_fp_op): Handle minus and mult. 1578 * config/aarch64/aarch64-sve.md (*add<mode>3, *sub<mode>3) 1579 (*mul<mode>3): Split the patterns after reload if we don't 1580 need the predicate operand. 1581 (*post_ra_<sve_fp_op><mode>3): New pattern. 1582 15832018-12-07 Bin Cheng <bin.cheng@linux.alibaba.com> 1584 1585 * profile-count.h (profile_count::oeprator>=): Fix typo by inverting 1586 return condition when *this is precise zero. 1587 15882018-12-07 Jakub Jelinek <jakub@redhat.com> 1589 1590 PR target/85593 1591 * final.c (rest_of_handle_final): Don't call collect_fn_hard_reg_usage 1592 for functions with naked attribute. 1593 1594 PR c/88367 1595 * tree-vrp.c (extract_range_from_binary_expr): For POINTER_PLUS_EXPR 1596 with -fno-delete-null-pointer-checks, set_nonnull only if the pointer 1597 is non-NULL and offset is known to have most significant bit clear. 1598 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): For ADDR_EXPR 1599 of MEM_EXPR, return true if the MEM_EXPR has non-zero offset with 1600 most significant bit clear. If offset does have most significant bit 1601 set and -fno-delete-null-pointer-checks, don't return true even if 1602 the base pointer is non-NULL. 1603 16042018-12-06 Alexandre Oliva <aoliva@redhat.com> 1605 1606 * cselib.c (cselib_record_sets): Skip strict low part sets 1607 with NULL src_elt. 1608 16092018-12-06 Paul A. Clarke <pc@us.ibm.com> 1610 1611 PR target/88316 1612 * config/rs6000/smmintrin.h: New file. 1613 * config.gcc: Add smmintrin.h to extra_headers for powerpc*-*-*. 1614 16152018-12-06 Paul A. Clarke <pc@us.ibm.com> 1616 1617 PR target/88316 1618 * config/rs6000/mmintrin.h (_mm_unpackhi_pi8): Fix for big-endian. 1619 (_mm_unpacklo_pi8): Likewise. 1620 (_mm_mulhi_pi16): Likewise. 1621 (_mm_packs_pi16): Fix for big-endian. Use preferred API. 1622 (_mm_packs_pi32): Likewise. 1623 (_mm_packs_pu16): Likewise. 1624 * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix for big-endian. 1625 (_mm_cvtss_si64): Likewise. 1626 (_mm_cvtpi32x2_ps): Likewise. 1627 (_mm_shuffle_ps): Likewise. 1628 (_mm_movemask_pi8): Likewise. 1629 (_mm_mulhi_pu16): Likewise. 1630 (_mm_sad_pu8): Likewise. 1631 (_mm_sad_pu8): Likewise. 1632 (_mm_cvtpu16_ps): Fix for big-endian. Use preferred API. 1633 (_mm_cvtpu8_ps): Likewise. 1634 (_mm_movemask_ps): Better #else case for big-endian (no functional 1635 change). 1636 (_mm_shuffle_pi16): Likewise. 1637 * config/rs6000/emmintrin.h (_mm_movemask_pd): Fix for big-endian. 1638 Better #else case for big-endian (no functional change). 1639 (_mm_movemask_epi8): Likewise. 1640 (_mm_shufflehi_epi16): Likewise. 1641 (_mm_shufflelo_epi16): Likewise. 1642 (_mm_shuffle_epi32): Likewise. 1643 (_mm_mul_epu32): Fix for big-endian. 1644 (_mm_bsrli_si128): Likewise. 1645 (_mm_cvtps_pd): Better #else case for big endian. 1646 (_mm_mulhi_epi16): Likewise. 1647 (_mm_mul_epu32): Likewise. 1648 (_mm_slli_si128): Likewise. 1649 (_mm_sll_epi16): Likewise. 1650 (_mm_sll_epi32): Likewise. 1651 (_mm_sra_epi16): Likewise. 1652 (_mm_sra_epi32): Likewise. 1653 (_mm_srl_epi16): Likewise. 1654 (_mm_srl_epi32): Likewise. 1655 (_mm_mulhi_epu16): Likewise. 1656 (_mm_sad_epu8): Likewise. 1657 * config/rs6000/pmmintrin.h (_mm_hadd_ps): Fix for big-endian. 1658 (_mm_sub_ps): Likewise. 1659 * config/rs6000/mmintrin.h (_mm_cmpeq_pi8): Fix for 32-bit mode. 1660 * gcc/config/rs6000/tmmintrin.h (_mm_alignr_epi8): Use ENDIAN 1661 macros consistently (no functional changes). 1662 (_mm_alignr_pi8): Likewise. 1663 16642018-12-06 Iain Sandoe <iain@sandoe.co.uk> 1665 1666 PR c++/87380 1667 * config/darwin.h (TARGET_WEAK_NOT_IN_ARCHIVE_TOC) Remove, use the 1668 default. 1669 * config/rs6000/darwin7.h (TARGET_WEAK_NOT_IN_ARCHIVE_TOC): New. 1670 16712018-12-06 Vladimir Makarov <vmakarov@redhat.com> 1672 1673 PR target/88282 1674 * ira.c (ira_init_register_move_cost): Use info from 1675 hard_regno_mode_ok instead of contains_reg_of_mode. 1676 * ira-costs.c (contains_reg_of_mode): Don't use cost from bigger 1677 hard register class for some fixed hard registers. 1678 16792018-12-06 Segher Boessenkool <segher@kernel.crashing.org> 1680 1681 * doc/extend.texi (Using Assembly Language with C): Document asm inline. 1682 (Size of an asm): Fix typo. Document asm inline. 1683 * gimple-pretty-print.c (dump_gimple_asm): Handle asm inline. 1684 * gimple.h (enum gf_mask): Add GF_ASM_INLINE. 1685 (gimple_asm_set_volatile): Fix typo. 1686 (gimple_asm_inline_p): New. 1687 (gimple_asm_set_inline): New. 1688 * gimplify.c (gimplify_asm_expr): Propagate the asm inline flag from 1689 tree to gimple. 1690 * ipa-icf-gimple.c (func_checker::compare_gimple_asm): Compare the 1691 gimple_asm_inline_p flag, too. 1692 * tree-core.h (tree_base): Document that protected_flag is ASM_INLINE_P 1693 in an ASM_EXPR. 1694 * tree-inline.c (estimate_num_insns): If gimple_asm_inline_p return 1695 a minimum size for an asm. 1696 * tree.h (ASM_INLINE_P): New. 1697 16982018-12-06 Segher Boessenkool <segher@kernel.crashing.org> 1699 1700 PR inline-asm/55681 1701 * doc/extend.texi (Basic Asm): Update grammar. 1702 (Extended Asm): Update grammar. 1703 17042018-12-06 Iain Sandoe <iain@sandoe.co.uk> 1705 1706 PR target/78444 1707 * config/i386/darwin.h (STACK_BOUNDARY): Remove macro. 1708 * config/i386/i386.c (ix86_compute_frame_layout): Ensure at least 128b 1709 stack alignment in non-leaf functions. 1710 17112018-12-06 Jakub Jelinek <jakub@redhat.com> 1712 1713 PR target/87598 1714 * config/aarch64/aarch64.c (aarch64_print_address_internal): Don't 1715 call output_operand_lossage on VOIDmode CONST_INTs. After 1716 output_operand_lossage do return false. 1717 17182018-12-06 Richard Biener <rguenther@suse.de> 1719 1720 * df-problems.c (df_rd_local_compute): Use bitmap_release. 1721 (df_live_free): Likewise. 1722 (df_md_local_compute): Likewise. 1723 (df_md_free): Release df_md_scratch bitmap. 1724 * loop-invariant.c (calculate_loop_reg_pressure): Use 1725 bitmap_release. 1726 * sched-deps.c (true_dependency_cache, output_dependency_cache, 1727 anti_dependency_cache, control_dependency_cache, 1728 spec_dependency_cache): Use bitmap instead of bitmap_head *. 1729 * sched-ebb.c (schedule_ebbs_init): Initialize non-GTY 1730 dont_calc_deps as bitmap allocated from obstack not GC. 1731 (schedule_ebbs_finish): Use bitmap_release. 1732 * sched-rgn.c (schedule_insns): Initialize non-GTY 1733 not_in_df as bitmap allocated from obstack not GC. 1734 Use bitmap_release. 1735 * sel-sched.c (_forced_ebb_heads): Remove premature optimization. 1736 (sel_region_init): Allocate forced_ebb_heads. 1737 (sel_region_finish): Free forced_ebb_heads. 1738 17392018-12-06 Richard Biener <rguenther@suse.de> 1740 1741 * bitmap.c (bitmap_head::crashme): Define. 1742 * bitmap.h (bitmap_head): Add constexpr default constructor 1743 poisoning the obstack member. 1744 (bitmap_head::crashme): Declare. 1745 (bitmap_release): New function clearing a bitmap and poisoning 1746 the obstack member. 1747 * gengtype.c (main): Make it recognize CONSTEXPR. 1748 * lra-constraints.c (lra_inheritance): Use bitmap_release 1749 instead of bitmap_clear. 1750 * ira.c (ira): Work around class-memaccess warning. 1751 * regrename.c (create_new_chain): Likewise. 1752 17532018-12-06 Jakub Jelinek <jakub@redhat.com> 1754 1755 PR tree-optimization/85726 1756 * generic-match-head.c (optimize_successive_divisions_p): New function. 1757 * gimple-match-head.c (optimize_successive_divisions_p): Likewise. 1758 * match.pd: Don't combine successive divisions if they aren't exact 1759 and optimize_successive_divisions_p is false. 1760 1761 PR fortran/88304 1762 * tree-nested.c (convert_nonlocal_reference_stmt): Remove clobbers 1763 for non-local automatic decls. 1764 17652018-12-05 David Edelsohn <dje.gcc@gmail.com> 1766 1767 * config/rs6000/aix72.h (ASM_DEFAULT_SPEC): Match Power7 processor 1768 default. 1769 17702018-12-05 Uros Bizjak <ubizjak@gmail.com> 1771 1772 * dwarf2out.c (field_byte_offset): Remove 1773 #ifdef PCC_BITFIELD_TYPE_MATTERS check. 1774 17752018-12-05 Richard Biener <rguenther@suse.de> 1776 Martin Sebor <msebor@redhat.com> 1777 1778 PR c/87028 1779 * calls.c (get_attr_nonstring_decl): Avoid setting *REF to 1780 SSA_NAME_VAR. 1781 * gcc/gimple-low.c (lower_stmt): Fold builtin calls here. 1782 * gimplify (maybe_fold_stmt): Avoid folding builtin calls. 1783 17842018-12-05 Iain Sandoe <iain@sandoe.co.uk> 1785 1786 * configure.ac (gcc_cv_otool): Set. 1787 * configure: Regenerate. 1788 17892018-12-05 Uros Bizjak <ubizjak@gmail.com> 1790 1791 * config/i386/cygming.h (PCC_BITFIELD_TYPE_MATTERS): Remove. 1792 (GROUP_BITFIELDS_BY_ALIGN): Ditto. 1793 17942018-12-05 Richard Sandiford <richard.sandiford@arm.com> 1795 1796 PR tree-optimization/88064 1797 * tree-vect-patterns.c (vect_convert_input): Convert the result of 1798 an existing cast if it has the right width but the wrong sign. 1799 Do not test the signedness of the required result when 1800 considering whether to split an existing cast; instead split to 1801 a type with the same signedness as the source of the cast, then 1802 convert it to the opposite signedness where necessary. 1803 (vect_recog_over_widening_pattern): Handle sign changes between 1804 the final PLUS_EXPR and the RSHIFT_EXPR. 1805 (vect_recog_average_pattern): Use an unsigned operation when 1806 truncating an addition, subtraction or multiplication. Cast the 1807 result back to the "real" signedness before promoting. 1808 18092018-12-05 Richard Biener <rguenther@suse.de> 1810 1811 PR tree-optimization/86637 1812 * tree-vectorizer.c (pass_slp_vectorize::execute): Reset 1813 vect_location at the end. 1814 18152018-12-05 Jakub Jelinek <jakub@redhat.com> 1816 1817 PR tree-optimization/87360 1818 * gimple-loop-jam.c (tree_loop_unroll_and_jam): On failure to analyze 1819 data dependencies, don't return false, just continue. Formatting 1820 fixes. 1821 (merge_loop_tree, bb_prevents_fusion_p, unroll_jam_possible_p, 1822 fuse_loops): Formatting fixes. 1823 1824 PR sanitizer/88333 1825 * cfgexpand.c (expand_stack_vars): If asan_vec is empty, start with 1826 aligning frame offset to ASAN_RED_ZONE_SIZE bytes. 1827 18282018-12-05 Xianmiao Qu <xianmiao_qu@c-sky.com> 1829 1830 * config.gcc (csky-*-linux-gnu*): Force .init_array support. 1831 18322018-12-05 Segher Boessenkool <segher@kernel.crashing.org> 1833 1834 * config/rs6000/rs6000.c (rs6000_function_arg): Only do the special 1835 aggregate handling on actual AIX, not on somewhat similar systems. 1836 (rs6000_arg_partial_bytes): Ditto. 1837 18382018-12-04 Vladimir Makarov <vmakarov@redhat.com> 1839 1840 PR rtl-optimization/88317 1841 * lra-constraints.c (split_reg): Don't set up check_only_regs if 1842 we are outside the inheritance pass. 1843 18442018-12-04 Jan Hubicka <hubicka@ucw.cz> 1845 1846 * ipa-prop.c (jump_function_useful_p): New. 1847 (ipa_read_node_info): Do not allocated useless jump functions. 1848 18492018-12-04 David Malcolm <dmalcolm@redhat.com> 1850 1851 PR c/82967 1852 * spellcheck.c (selftest::test_suggestions): Add another assertion. 1853 18542018-12-04 Martin Sebor <msebor@redhat.com> 1855 1856 * doc/extend.texi (__builtin_has_attribute): Fix typo. 1857 18582018-12-04 Jakub Jelinek <jakub@redhat.com> 1859 1860 PR target/88188 1861 * config/rs6000/rs6000.c (ccr_bit): Return -1 instead of assertion 1862 failures. 1863 (print_operand): Use REG_P instead of GET_CODE == REG. 1864 <case 'D'>: Also check CR_REGNO_P (REGNO (x)). 1865 <case 't'>: Likewise. Remove GET_MODE check. 1866 * config/rs6000/rs6000.md (scc patterns): Assert ccr_bit didn't 1867 return -1. 1868 18692018-12-04 Martin Sebor <msebor@redhat.com> 1870 1871 * doc/extend.texi (attribute aligned): Expand. 1872 18732018-12-04 David Edelsohn <dje.gcc@gmail.com> 1874 1875 PR target/61976 1876 * config/rs6000/rs6000.c (rs6000_function_arg): Don't pass aggregates 1877 in FPRs on AIX. 1878 (rs6000_arg_partial_bytes): Same. 1879 18802018-12-04 Vladimir Makarov <vmakarov@redhat.com> 1881 1882 PR target/88282 1883 * ira-costs.c (exec): Try bigger class to use smaller register 1884 move cost. 1885 18862018-12-04 Michael Ploujnikov <michael.ploujnikov@oracle.com> 1887 1888 PR ipa/88297 1889 * ipa-cp.c (create_specialized_node): Track clone counters by 1890 node assembler names. 1891 (ipcp_driver): Change type of clone_num_suffixes key to const char*. 1892 18932018-12-04 Richard Biener <rguenther@suse.de> 1894 1895 * tree-vect-loop.c (vect_transform_loop): Fix epilogue iterations 1896 for epilogue vectorization. 1897 18982018-12-04 Richard Biener <rguenther@suse.de> 1899 1900 PR tree-optimization/88301 1901 * tree-vrp.c (register_edge_assert_for_2): Fix sign-conversion 1902 issues in last commit. 1903 19042018-12-04 Richard Biener <rguenther@suse.de> 1905 1906 PR tree-optimization/88315 1907 * tree-vect-loop.c (get_initial_defs_for_reduction): Simplify 1908 and fix initialization vector for SAD and DOT_PROD SLP reductions. 1909 19102018-12-03 Sandra Loosemore <sandra@codesourcery.com> 1911 1912 PR c/59039 1913 * doc/extend.texi (Nonlocal gotos): New section. 1914 19152018-12-03 Uros Bizjak <ubizjak@gmail.com> 1916 1917 * config/i386/sse.md (vec_concatv2df): Change (v,xm,C) 1918 alternative to (v,vm,c). 1919 19202018-12-03 Segher Boessenkool <segher@kernel.crashing.org> 1921 1922 * gcc/config.gcc (Unsupported targets): Add powerpc*-*-linux*paired*. 1923 (powerpc*-*-linux*): Don't handle it here. 1924 19252018-12-03 Jeff Law <law@redhat.com> 1926 1927 * config/h8300/h8300.md (<code>qi3_1): Add gcc_unreachable to keep 1928 warnings quiet. 1929 1930 * config/h8300/h8300.md (low byte masking peepholes): Only accept 1931 nonimmediate_operand for the to-be-masked operand. 1932 19332018-12-03 Richard Biener <rguenther@suse.de> 1934 1935 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only 1936 allow expected function-pointer cast re-instantiation. 1937 19382018-12-03 Ilya Leoshkevich <iii@linux.ibm.com> 1939 1940 * common/config/s390/s390-common.c (s390_option_init_struct): 1941 Use section anchors by default. 1942 * config/s390/s390.c (s390_check_symref_alignment): Handle 1943 anchors. 1944 (TARGET_MAX_ANCHOR_OFFSET): Use short displacement. 1945 * output.h (assemble_align): Pass `align' as unsigned int, so 1946 that the value 0x80000000, which corresponds to `aligned(1 << 1947 28)', would pass the `align > BITS_PER_UNIT' check. 1948 * varasm.c (assemble_align): Likewise. 1949 19502018-12-03 Julian Brown <julian@codesourcery.com> 1951 1952 * tree-pretty-print.c (dump_omp_clause): Make default case 1953 gcc_unreachable. 1954 19552018-12-03 Richard Biener <rguenther@suse.de> 1956 1957 PR tree-optimization/88301 1958 * tree-vrp.c (register_edge_assert_for_2): Handle conversions 1959 that do not change the value by registering the same assert 1960 for the operand. 1961 19622018-12-03 Richard Biener <rguenther@suse.de> 1963 1964 * tree-vect-loop.c (vect_transform_loop): Properly compute 1965 upper bound for the epilogue when doing epilogue vectorization. 1966 19672018-12-03 Ilya Leoshkevich <iii@linux.ibm.com> 1968 1969 PR target/80080 1970 * cfgcleanup.c (class pass_postreload_jump): New pass. 1971 (pass_postreload_jump::execute): Likewise. 1972 (make_pass_postreload_jump): Likewise. 1973 * passes.def: Add pass_postreload_jump before 1974 pass_postreload_cse. 1975 * tree-pass.h (make_pass_postreload_jump): New pass. 1976 19772018-12-02 Uros Bizjak <ubizjak@gmail.com> 1978 1979 * config/i386/i386.c (inline_memory_move_cost): Correctly 1980 compute number of TFmode GPR moves. 1981 19822018-12-02 Jakub Jelinek <jakub@redhat.com> 1983 1984 PR target/88278 1985 * config/i386/sse.md (*vec_concatv4sf_0, *vec_concatv4si_0): New insns. 1986 19872018-12-02 Jeff Law <law@redhat.com> 1988 1989 * config/h8300/h8300.md (call, call_value): Drop mode from 1990 operand containing number of bytes pushed. 1991 (call_insn, call_value_insn): Use mode iterator for number 1992 of bytes pushed operand. Make pattern name vary appropriately. 1993 19942018-12-02 Jan Hubicka <jh@suse.cz> 1995 1996 * bb-reorer.c (find_rarely_executed_basic_blocks_and_crossin_edges): 1997 trust precise profiles. 1998 19992018-12-02 Jakub Jelinek <jakub@redhat.com> 2000 2001 * tree-nested.c (convert_nonlocal_omp_clauses, 2002 convert_local_omp_clauses): Handle OMP_CLAUSE_IN_REDUCTION, 2003 OMP_CLAUSE_TASK_REDUCTION and OMP_CLAUSE__SIMT_ clauses. 2004 (convert_nonlocal_reference_stmt, convert_local_reference_stmt): 2005 Convert clauses for GIMPLE_OMP_TASKGROUP. 2006 2007 * omp-low.c (check_omp_nesting_restrictions): Allow cancel or 2008 cancellation point with taskgroup clause inside of taskloop. Consider 2009 a taskloop construct without nogroup clause as implicit taskgroup for 2010 diagnostics if cancel/cancellation point with taskgroup clause is 2011 closely nested inside of taskgroup region. 2012 2013 PR sanitizer/88291 2014 * asan.c (asan_clear_shadow): Move assert that len is multiple of 4 2015 to the start of the function. 2016 (asan_emit_stack_protection): When emitting clearing sequence for 2017 epilogue, align offset down to ASAN_RED_ZONE_SIZE granularity, 2018 add last_size_aligned which is last_size padded to multiples of 2019 ASAN_RED_ZONE_SIZE and emit asan_clear_shadow always on 4 byte 2020 boundaries. 2021 20222018-12-01 Jakub Jelinek <jakub@redhat.com> 2023 2024 PR sanitizer/88289 2025 * asan.c (asan_redzone_buffer::flush_redzone_payload): Fix up 2026 an off-by-one for BYTES_BIG_ENDIAN. 2027 2028 PR target/54589 2029 * combine.c (find_split_point): For invalid memory address 2030 nonobj + obj + const, if reg + obj + const is valid addressing 2031 mode, split at nonobj. Use if rather than else if for the 2032 fallback. Comment fixes. 2033 20342018-11-30 Indu Bhagat <indu.bhagat@oracle.com> 2035 2036 * coverage.c (get_coverage_counts): Use from_function_decl for precise 2037 function location. 2038 * profile-count.c (profile_count::dump): Add handling for precise 2039 profile quality. 2040 * profile.c (compute_branch_probabilities): Rely on exec_counts instead 2041 of profile_info to set profile_status of function. 2042 (branch_prob): Do not set profile_status of function based on 2043 profile_info. Done above based on exec_counts. 2044 20452018-11-30 Jakub Jelinek <jakub@redhat.com> 2046 2047 PR tree-optimization/88274 2048 * tree-ssa-reassoc.c (optimize_range_tests_xor, 2049 optimize_range_tests_diff): If type has smaller precision than 2050 corresponding mode or if it has non-standard min/max, compute 2051 everything in a standard type for the precision. 2052 2053 PR testsuite/85368 2054 * params.def (PARAM_LOGICAL_OP_NON_SHORT_CIRCUIT): New param. 2055 * tree-ssa-ifcombine.c (ifcombine_ifandif): If 2056 --param logical-op-non-short-circuit is present, override 2057 LOGICAL_OP_NON_SHORT_CIRCUIT value from the param. 2058 * fold-const.c (fold_range_test, fold_truth_andor): Likewise. 2059 20602018-11-30 Jeff Law <law@redhat.com> 2061 2062 * optabs.c (expand_binop): Use "machine_mode" rather than 2063 "enum machine mode" in most recent change. 2064 20652018-11-30 Wilco Dijkstra <wdijkstr@arm.com> 2066 2067 PR middle-end/64242 2068 * builtins.c (expand_builtin_longjmp): Use a temporary when restoring 2069 the frame pointer. 2070 (expand_builtin_nonlocal_goto): Likewise. 2071 20722018-11-30 David Malcolm <dmalcolm@redhat.com> 2073 2074 * diagnostic-core.h (emit_diagnostic): New decl. 2075 * diagnostic.c (emit_diagnostic): New overload, taking a 2076 rich_location *. 2077 20782018-11-30 David Malcolm <dmalcolm@redhat.com> 2079 2080 * pretty-print.c (class selftest::test_pretty_printer): New 2081 subclass of pretty_printer. 2082 (selftest::test_prefixes_and_wrapping): New test. 2083 (selftest::pretty_print_c_tests): Call it. 2084 20852018-11-30 Michael Ploujnikov <michael.ploujnikov@oracle.com> 2086 2087 Minimize clone counter memory usage in create_virtual_clone. 2088 * cgraph.h (clone_function_name): Add a variant that takes a 2089 tree decl. 2090 * cgraph.h (cgraph_node::create_virtual_clone): Add a new 2091 argument: num_suffix. 2092 * cgraphclones.c (cgraph_node::create_virtual_clone): Pass 2093 num_suffix to clone_function_name. 2094 (clone_function_name): Add a variant that takes a tree decl. 2095 * ipa-cp.c (create_specialized_node): Keep track of clone 2096 counters in clone_num_suffixes hash map. 2097 (ipcp_driver): Free the counter hash map. 2098 * ipa-hsa.c (process_hsa_functions): Creates at most one hsa 2099 clone per function. 2100 21012018-11-30 Michael Ploujnikov <michael.ploujnikov@oracle.com> 2102 Make function assembly more independent. 2103 2104 This is achieved by having clone_function_name assign unique clone 2105 numbers for each function independently. 2106 2107 * cgraphclones.c: Replaced clone_fn_id_num with clone_fn_ids; 2108 hash map. 2109 (clone_function_name_numbered): Use clone_fn_ids. 2110 21112018-11-30 Vladimir Makarov <vmakarov@redhat.com> 2112 2113 PR rtl-optimization/88179 2114 * lra-constraints.c (address_eliminator::address_eliminator): 2115 Don't eleminate regs in illegitimate address. 2116 21172018-11-30 David Malcolm <dmalcolm@redhat.com> 2118 2119 PR preprocessor/88257 2120 * input.c (get_substring_ranges_for_loc): Fix indentation. 2121 Bulletproof against getting a different files back from 2122 linemap_resolve_location and expand_location_to_spelling_point. 2123 21242018-11-30 Alexander Monakov <amonakov@ispras.ru> 2125 2126 PR gcov-profile/88279 2127 * tree-profile.c (init_ic_make_global_vars): Call 2128 decl_default_tls_model for variable and not it's type. 2129 21302018-11-30 Sam Tebbs <sam.tebbs@arm.com> 2131 2132 * config/arm/arm.h (TARGET_DOTPROD): Add TARGET_VFP5 constraint. 2133 21342018-11-30 Martin Liska <mliska@suse.cz> 2135 2136 PR sanitizer/81715 2137 * asan.c (asan_shadow_cst): Remove, partially transform 2138 into flush_redzone_payload. 2139 (RZ_BUFFER_SIZE): New. 2140 (struct asan_redzone_buffer): New. 2141 (asan_redzone_buffer::emit_redzone_byte): Likewise. 2142 (asan_redzone_buffer::flush_redzone_payload): Likewise. 2143 (asan_redzone_buffer::flush_if_full): Likewise. 2144 (asan_emit_stack_protection): Use asan_redzone_buffer class 2145 that is responsible for proper aligned stores and flushing 2146 of shadow memory payload. 2147 * asan.h (ASAN_MIN_RED_ZONE_SIZE): New. 2148 (asan_var_and_redzone_size): Likewise. 2149 * cfgexpand.c (expand_stack_vars): Use smaller alignment 2150 (ASAN_MIN_RED_ZONE_SIZE) in order to make shadow memory 2151 for automatic variables more compact. 2152 21532018-11-30 Alan Modra <amodra@gmail.com> 2154 2155 * config/rs6000/predicates.md (easy_fp_constant): Avoid long 2156 dependent insn sequences. 2157 * config/rs6000/rs6000.c (num_insns_constant): Support long 2158 double constants. 2159 * config/rs6000/rs6000.md (mov<mode>_softfloat <FMOVE128>): Adjust 2160 length attribute. 2161 21622018-11-30 Alan Modra <amodra@gmail.com> 2163 2164 * config/rs6000/rs6000.c (num_insns_constant_gpr): Renamed from 2165 num_insns_constant_wide. Make static. Revise comment. 2166 (num_insns_constant_multi): New function. 2167 (num_insns_constant): Formatting. Correct CONST_WIDE_INT 2168 calculation. Simplify and extract code common to both 2169 CONST_INT and CONST_DOUBLE. Add gcc_unreachable for unhandled 2170 const_double modes. 2171 * config/rs6000/rs6000-protos.h (num_insns_const_wide): Delete. 2172 21732018-11-30 Richard Biener <rguenther@suse.de> 2174 2175 PR tree-optimization/88274 2176 * tree-vrp.c (ranges_from_anti_range): Fix handling of 2177 TYPE_MIN/MAX_VALUE. 2178 21792018-11-29 Andi Kleen <ak@linux.intel.com> 2180 2181 * config/i386/i386.c (current_fentry_section): Add. 2182 (x86_function_profiler): Handle fentry section. 2183 (ix86_attribute_table): Add fentry section. 2184 * config/i386/i386.opt: Add -mfentry-section. 2185 * doc/extend.texi: Document fentry_section attribute. 2186 * doc/invoke.texi: Document -mfentry-section. 2187 21882018-11-29 Andi Kleen <ak@linux.intel.com> 2189 2190 * config/i386/i386.c (x86_print_call_or_nop): Handle nop name. 2191 (current_fentry_name): Add. 2192 (x86_function_profiler): Handle fentry_name attribute. 2193 (ix86_handle_fentry_name): Add. 2194 (ix86_attribute_table): Add fentry_name. 2195 * config/i386/i386.opt: Add -mfentry-name 2196 * doc/extend.texi: Document fentry_name. 2197 * doc/invoke.texi: Document minstrument-return. 2198 21992018-11-29 Andi Kleen <ak@linux.intel.com> 2200 2201 * config/i386/i386-opts.h (enum instrument_return): Add. 2202 * config/i386/i386.c (output_return_instrumentation): Add. 2203 (ix86_output_function_return): Call output_return_instrumentation. 2204 (ix86_output_call_insn): Call output_return_instrumentation. 2205 * config/i386/i386.opt: Add -minstrument-return=. 2206 * doc/invoke.texi (-minstrument-return): Document. 2207 22082018-11-29 Eric Botcazou <ebotcazou@adacore.com> 2209 2210 PR target/87807 2211 * config/sparc/sparc-modes.def: Minor tweak. 2212 * config/sparc/sparc.c: Minor reordering. 2213 (sparc_pass_by_reference): Move around. 2214 (traverse_record_type): Change offset from HOST_WIDE_INT to int. 2215 (classify_registers): Likewise for bitpos. 2216 (function_arg_slotno): Remove dead test and tweak comments. 2217 <MODE_RANDOM>: Remove useless assertion and test whether the 2218 parameter is named in order to pass it in FP registers. Return 2219 the regno for floating-point vector types. 2220 (compute_int_layout): Change bitpos from HOST_WIDE_INT to int. 2221 (compute_fp_layout): Likewise. 2222 (count_registers): Likewise. 2223 (assign_int_registers): Likewise. 2224 (assign_fp_registers): Likewise. 2225 (assign_registers): Likewise. 2226 (function_arg_record_value): Change size from HOST_WIDE_INT to int 2227 and use CEIL_NWORDS to compute the number of registers. 2228 (function_arg_union_value): Minor tweaks. 2229 (function_arg_vector_value): Add slotno and named parameters, use 2230 CEIL_NWORDS to compute the number of registers. 2231 (sparc_function_arg_1): Rework handling of vector types. Change 2232 size from HOST_WIDE_INT to int. 2233 (sparc_arg_partial_bytes): Rework handling of 32-bit ABI and deal 2234 with vector types for the 64-bt ABI. 2235 (sparc_function_arg_advance): Likewise. 2236 (sparc_return_in_memory): Add reference to -fpcc-struct-return. 2237 (sparc_struct_value_rtx): Return NULL_RTX instead of 0. 2238 (sparc_function_value_1): Rework handling of vector types. Change 2239 size from HOST_WIDE_INT to int. 2240 22412018-11-29 Jakub Jelinek <jakub@redhat.com> 2242 2243 PR target/88152 2244 * config/i386/sse.md (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_lt, 2245 *<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt, 2246 *<sse>_movmsk<ssemodesuffix><avxsizesuffix>_shift, 2247 *<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift, 2248 *<sse2_avx2>_pmovmskb_lt, *<sse2_avx2>_pmovmskb_zext_lt): New 2249 define_insn_and_split patterns. 2250 2251 PR target/54700 2252 * config/i386/sse.md 2253 (*<sse4_1>_blendv<ssemodesuffix><avxsizesuffix>_lt, 2254 *<sse4_1>_blendv<ssefltmodesuffix><avxsizesuffix>_ltint, 2255 *<sse4_1_avx2>_pblendvb_lt): Change define_insn into 2256 define_insn_and_split. 2257 22582018-11-29 Peter Bergner <bergner@linux.ibm.com> 2259 2260 PR target/87496 2261 * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow 2262 -mabi=ieeelongdouble without both -mpopcntd and -mvsx. 2263 22642018-11-29 Uros Bizjak <ubizjak@gmail.com> 2265 2266 * config/i386/i386.c (inline_memory_move_cost): 2267 Check "in" for 2 in MMX_CLASS_P case. 2268 * config/i386/mmx.md (*mov<mode>_internal): Correct 2269 TARGET_INTER_UNIT_MOVES_FROM_VEC and TARGET_INTER_UNIT_MOVES_TO_VEC 2270 alternatives in preferred_for_speed attribute calculation. 2271 22722018-11-29 Martin Sebor <msebor@redhat.com> 2273 2274 PR c/88172 2275 PR testsuite/88208 2276 * doc/extend.texi (attribute constructor): Clarify. 2277 22782018-11-29 Martin Liska <mliska@suse.cz> 2279 2280 PR middle-end/88246 2281 * builtins.c (expand_movstr): Fix thinko introduced 2282 when switching to the new enum. 2283 22842018-11-29 qing zhao <qing.zhao@oracle.com> 2285 2286 * cif-code.def (EXTERN_LIVE_ONLY_STATIC): New CIF code. 2287 * common.opt: Add -flive-patching flag. 2288 * doc/invoke.texi: Document -flive-patching. 2289 * flag-types.h (enum live_patching_level): New enum. 2290 * ipa-inline.c (can_inline_edge_p): Disable external functions from 2291 inlining when flag_live_patching is LIVE_PATCHING_INLINE_ONLY_STATIC. 2292 * opts.c (control_options_for_live_patching): New function. 2293 (finish_options): Make flag_live_patching incompatible with flag_lto. 2294 Control IPA optimizations based on different levels of 2295 flag_live_patching. 2296 22972018-11-29 Giuliano Belinassi <giuliano.belinassi@usp.br> 2298 2299 * match.pd (sinh (atanh (x))): New simplification rules. 2300 (cosh (atanh (x))): Likewise. 2301 23022018-11-29 Jakub Jelinek <jakub@redhat.com> 2303 2304 PR target/54700 2305 * config/i386/sse.md (ssebytemode): Add V16SI, V8SI and V4SI entries. 2306 (ssefltmodesuffix, ssefltvecmode): New define_mode_attrs. 2307 (*<sse4_1>_blendv<ssemodesuffix><avxsizesuffix>_lt, 2308 *<sse4_1>_blendv<ssefltmodesuffix><avxsizesuffix>_ltint, 2309 *<sse4_1_avx2>_pblendvb_lt): New define_insns. 2310 2311 PR target/88152 2312 * tree.h (build_uniform_cst, uniform_integer_cst_p): Declare. 2313 * tree.c (build_uniform_cst, uniform_integer_cst_p): New functions. 2314 * match.pd (define_predicates): Add uniform_integer_cst_p. 2315 (cmp @0 INTEGER_CST@1, cmp (convert?@2 @0) INTEGER_CST@1): Adjust 2316 so that it works also for vector comparisons with uniform constants 2317 with INTEGER_CST element. 2318 2319 PR target/88234 2320 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): For 2321 vec_add and vec_sub builtins, perform PLUS_EXPR or MINUS_EXPR 2322 in unsigned_type_for instead of vector integral type where overflow 2323 doesn't wrap. 2324 23252018-11-29 Michael Ploujnikov <michael.ploujnikov@oracle.com> 2326 2327 There can be at most one .resolver clone per function 2328 * config/rs6000/rs6000.c (make_resolver_func): Generate 2329 resolver symbol with clone_function_name instead of 2330 clone_function_name_numbered. 2331 23322018-11-29 Richard Biener <rguenther@suse.de> 2333 2334 PR tree-optimization/88243 2335 * tree-vect-patterns.c (vect_mark_pattern_stmts): Set the def 2336 type of all pattern-sequence stmts to vect_internal_def. 2337 23382018-11-29 Andre Vieira <andre.simoesdiasvieira@arm.com> 2339 2340 PR target/88224 2341 * config/arm/arm-cpus.in (armv7-r): Add FP16conv configurations. 2342 (cortex-r7, cortex-r8): Update default and add new configuration. 2343 * doc/invoke.texi (armv7-r): Add two new vfp options. 2344 (nofp.dp): Add cortex-r7 and cortex-r8 to the list of targets that 2345 support this option. 2346 23472018-11-29 Alan Modra <amodra@gmail.com> 2348 2349 * config/rs6000/rs6000.c (rs6000_emit_move): Disable long 2350 double split for targets other than Darwin. 2351 23522018-11-29 Alan Modra <amodra@gmail.com> 2353 2354 * config/rs6000/rs6000.md (movdi_internal64): Correct lengths. 2355 23562018-11-29 Alan Modra <amodra@gmail.com> 2357 2358 * config/rs6000/predicates.md (easy_fp_constant): Remove code 2359 dealing with integers in const_double. Assert on unexpected 2360 modes. Delete superfluous ABI_V4 test. 2361 * config/rs6000/rs6000.c (rs6000_emit_move): Comment fixes. 2362 Don't call easy_fp_constant for Pmode. 2363 (rs6000_hash_constant): Remove code dealing with integers in 2364 const_double. 2365 (rs6000_legitimate_constant_p): Likewise. 2366 (output_toc): Formatting, use CONST_DOUBLE_P. 2367 23682018-11-29 Alan Modra <amodra@gmail.com> 2369 2370 * config/rs6000/constraints.md (G, H): Comment on purpose of 2371 constraint. Correct mode comments and "G" mode test. 2372 * config/rs6000/rs6000.md (movdi_internal32): Remove "GH" from 2373 alternative handling "F". Add length attr. Formatting. 2374 (mov<mode>_softfloat <FMOVE128>): Delete "GH" from 2375 alternative, and move "F" to separate alternative. Correct 2376 insn lengths. 2377 23782018-11-29 Alan Modra <amodra@gmail.com> 2379 2380 * config/rs6000/predicates.md (easy_vector_constant): Don't call 2381 easy_fp_constant. 2382 23832018-11-29 Alan Modra <amodra@gmail.com> 2384 2385 * config.in (HAVE_AS_PLTSEQ): Add. 2386 * config/rs6000/predicates.md (indirect_call_operand): New. 2387 * config/rs6000/rs6000-protos.h (rs6000_pltseq_template), 2388 (rs6000_sibcall_sysv): Declare. 2389 * config/rs6000/rs6000.c (init_cumulative_args): Set cookie 2390 CALL_LONG for -fno-plt. 2391 (print_operand <T, z, 0>): Handle UNSPEC_PLTSEQ. 2392 (rs6000_indirect_call_template_1): Emit .reloc directives for 2393 UNSPEC_PLTSEQ calls. 2394 (rs6000_pltseq_template): New function. 2395 (rs6000_longcall_ref): Add arg parameter. Use PLT16 insns if 2396 relocs supported by assembler. Move SYMBOL_REF test to callers. 2397 (rs6000_call_aix): Adjust rs6000_longcall_ref call. Package 2398 insns in UNSPEC_PLTSEQ, preserving original func_desc. 2399 (rs6000_call_sysv): Likewise. 2400 (rs6000_sibcall_sysv): New function. 2401 * config/rs6000/rs6000.h (HAVE_AS_PLTSEQ): Provide default. 2402 * config/rs6000/rs6000.md (UNSPEC_PLTSEQ, UNSPEC_PLT16_HA, 2403 UNSPEC_PLT16_LO): New. 2404 (pltseq_tocsave, pltseq_plt16_ha, pltseq_plt16_lo, pltseq_mtctr): New. 2405 (call_indirect_nonlocal_sysv): Don't differentiate zero from non-zero 2406 cookie in constraints. Test explicitly for flags in length attr. 2407 Handle unspec operand 1. 2408 (call_value_indirect_nonlocal_sysv): Likewise. 2409 (call_indirect_aix, call_value_indirect_aix): Handle unspec operand 1. 2410 (call_indirect_elfv2, call_value_indirect_elfv2): Likewise. 2411 (sibcall, sibcall_value): Use rs6000_sibcall_sysv. 2412 (sibcall_indirect_nonlocal_sysv): New pattern. 2413 (sibcall_value_indirect_nonlocal_sysv): Likewise. 2414 (sibcall_nonlocal_sysv, sibcall_value_nonlocal_sysv): Remove indirect 2415 call alternatives. 2416 * configure.ac: Check for gas plt sequence marker support. 2417 * configure: Regenerate. 2418 24192018-11-29 Alan Modra <amodra@gmail.com> 2420 2421 * config/rs6000/predicates.md (unspec_tls): New. 2422 * config/rs6000/rs6000-protos.h (rs6000_call_template), 2423 (rs6000_sibcall_template): Update prototype. 2424 (rs6000_longcall_ref): Delete. 2425 (rs6000_call_sysv): Declare. 2426 * config/rs6000/rs6000.c (edit_tls_call_insn): New function. 2427 (global_tlsarg): New variable. 2428 (rs6000_legitimize_tls_address): Rewrite __tls_get_addr call 2429 handling. 2430 (print_operand): Extract UNSPEC_TLSGD address operand. 2431 (rs6000_call_template, rs6000_sibcall_template): Remove arg 2432 parameter, extract from second call operand instead. 2433 (rs6000_longcall_ref): Make static, localize vars. 2434 (rs6000_call_aix): Rename parameter to reflect new usage. Take 2435 tlsarg from global_tlsarg. Don't create unused rtl or nop insns. 2436 (rs6000_sibcall_aix): Rename parameter to reflect new usage. Take 2437 tlsarg from global_tlsarg. 2438 (rs6000_call_sysv): New function. 2439 * config/rs6000/rs6000.md: Adjust rs6000_call_template and 2440 rs6000_sibcall_template throughout. 2441 (tls_gd_aix, tls_gd_sysv, tls_gd_call_aix, tls_gd_call_sysv): Delete. 2442 (tls_ld_aix, tls_ld_sysv, tls_ld_call_aix, tls_ld_call_sysv): Delete. 2443 (tls_gdld_nomark): New insn. 2444 (tls_gd): Swap operand order. Simplify mode selection. 2445 (tls_gd_high, tls_gd_low): Swap operand order. 2446 (tls_ld): Remove const_int 0 vector element from UNSPEC_TLSLD. 2447 Simplify mode selection. 2448 (tls_ld_high, tls_ld_low): Similarly adjust UNSPEC_TLSLD. 2449 (call, call_value): Don't assert for second call operand. 2450 Use rs6000_call_sysv. 2451 24522018-11-29 Alan Modra <amodra@gmail.com> 2453 2454 * config/rs6000/darwin.md (call_indirect_nonlocal_darwin64), 2455 (call_nonlocal_darwin64, call_value_indirect_nonlocal_darwin64), 2456 (call_value_nonlocal_darwin64): Remove constraints from second call 2457 arg, the rounded_stack_size_rtx arg. 2458 * config/rs6000/rs6000.md (tls_gd_aix, tls_gd_sysv, tls_gd_call_aix), 2459 (tls_gd_call_sysv, tls_ld_aix, tls_ld_sysv, tls_ld_call_aix), 2460 (tls_ld_call_sysv, call_local32, call_local64, call_value_local32), 2461 (call_value_local64, call_indirect_nonlocal_sysv), 2462 (call_nonlocal_sysv, call_nonlocal_sysv_secure), 2463 (call_value_indirect_nonlocal_sysv, call_value_nonlocal_sysv), 2464 (call_value_nonlocal_sysv_secure, call_local_aix), 2465 (call_value_local_aix, call_nonlocal_aix, call_value_nonlocal_aix), 2466 (call_indirect_aix, call_value_indirect_aix, call_indirect_elfv2), 2467 (call_value_indirect_elfv2, sibcall_local32, sibcall_local64), 2468 (sibcall_value_local32, sibcall_value_local64, sibcall_aix), 2469 (sibcall_value_aix): Likewise. 2470 24712018-11-29 Alan Modra <amodra@gmail.com> 2472 2473 * config/rs6000/rs6000.md (TLSmode): Delete mode iterator. Replace 2474 with P throughout except for call mems which should use SI. 2475 (tls_abi_suffix, tls_sysv_suffix, tls_insn_suffix): Delete mode 2476 attributes. Replace with bits, mode and ptrload respectively. 2477 24782018-11-29 Alan Modra <amodra@gmail.com> 2479 2480 * config/rs6000/rs6000-protos.h (rs6000_indirect_call_template), 2481 (rs6000_indirect_sibcall_template): Declare. 2482 * config/rs6000/rs6000.c (rs6000_indirect_call_template_1), 2483 (rs6000_indirect_call_template, rs6000_indirect_sibcall_template): 2484 New functions. 2485 * config/rs6000/rs6000.md (call_indirect_nonlocal_sysv), 2486 (call_value_indirect_nonlocal_sysv, sibcall_nonlocal_sysv), 2487 (call_indirect_aix, call_value_indirect_aix): Use 2488 rs6000_indirect_call_template and rs6000_indirect_sibcall_template. 2489 call_indirect_elfv2, call_value_indirect_elfv2): Likewise, and 2490 handle both speculation and non-speculation cases. 2491 (call_indirect_aix_nospec, call_value_indirect_aix_nospec): Delete. 2492 (call_indirect_elfv2_nospec, call_value_indirect_elfv2_nospec): Delete. 2493 24942018-11-29 Alan Modra <amodra@gmail.com> 2495 2496 * config/rs6000/rs6000-protos.h (rs6000_call_template): Declare. 2497 (rs6000_sibcall_template): Declare. 2498 (macho_call_template): Rename from output_call. 2499 * config/rs6000/rs6000.c (rs6000_call_template_1): New function. 2500 (rs6000_call_template, rs6000_sibcall_template): Likewise. 2501 (macho_call_template): Rename from output_call. 2502 * config/rs6000/rs6000.md (tls_gd_aix, tls_gd_sysv), 2503 (tls_gd_call_aix, tls_gd_call_sysv, tls_ld_aix, tls_ld_sysv), 2504 (tls_ld_call_aix, tls_ld_call_sysv, call_nonlocal_sysv), 2505 (call_nonlocal_sysv_secure, call_value_nonlocal_sysv), 2506 (call_value_nonlocal_sysv_secure, call_nonlocal_aix), 2507 (call_value_nonlocal_aix): Use rs6000_call_template and update 2508 occurrences of output_call to macho_call_template. 2509 (sibcall_nonlocal_sysv, sibcall_value_nonlocal_sysv, sibcall_aix), 2510 (sibcall_value_aix): Use rs6000_sibcall_template. 2511 25122018-11-28 Aaron Sawdey <acsawdey@linux.ibm.com> 2513 2514 * config/rs6000/rs6000-string.c (expand_block_clear): Change how 2515 we determine if unaligned vsx is ok. 2516 25172018-11-28 Jan Hubicka <jh@suse.cz> 2518 2519 * predict.c (determine_unlikely_bbs): Forward declare; also determine 2520 edges that are always known to be taken because it is only likely 2521 edge out of given BB. 2522 (tree_estimate_probability): Call before profile guessing to get bit 2523 of extra precision. 2524 25252018-11-28 Jan Hubicka <jh@suse.cz> 2526 2527 * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Handle 2528 profile_probability::always better. 2529 25302018-11-28 Jan Hubicka <jh@suse.cz> 2531 2532 * profile-count.h (profile_count::split): Give better result when 2533 splitting profile_probability::always. 2534 25352018-11-28 Vladimir Makarov <vmakarov@redhat.com> 2536 2537 PR target/88207 2538 * ira-costs.c (scan_one_insn): Process subregs when updating costs 2539 for pseudos and allocnos from insn. 2540 25412018-11-28 David Edelsohn <dje.gcc@gmail.com> 2542 2543 * config/rs6000/aix72.h: Update to match aix71.h changes. 2544 25452018-11-28 Jeff Law <law@redhat.com> 2546 2547 * config/h8300/constraints.md: Add "C" constraint for call insns. 2548 * config/h8300/h8300.md (call, call_value): Turn into a define_expand 2549 and define_insn pair. Move invalid call targets into a register in 2550 the expander and fix constraints in the matching pattern. 2551 * config/h8300/predicates.md (call_expander_operand): Renamed from 2552 call_insn_operand. Reject things we shouldn't be trying to handle. 2553 (call_insn_operand): New predicate for use by the call/call_value 2554 insns. 2555 (small_call_insn_operand): Update appropriately. 2556 25572018-11-28 Sam Tebbs <sam.tebbs@arm.com> 2558 2559 * config/aarch64/aarch64.c (aarch64_process_target_attr): Replace 2560 calls to strtok with strtok_r. 2561 25622018-11-28 Richard Biener <rguenther@suse.de> 2563 2564 PR tree-optimization/88223 2565 * tree-ssa-sccvn.c (vn_reference_lookup_3): When skipping 2566 over a stored-same value may-alias store make sure to consider 2567 partial overlaps which are valid when TBAA reasonings do not 2568 apply and byte-granular overlaps are possible at all. 2569 25702018-11-28 Richard Biener <rguenther@suse.de> 2571 2572 PR tree-optimization/88217 2573 * vr-values.c (vr_values::extract_range_from_phi_node): Make 2574 sure to handle results > +INF and < -INF correctly when 2575 trying to drop down to +INF - 1 or -INF + 1. 2576 25772018-11-28 Alan Modra <amodra@gmail.com> 2578 2579 * xcoffout.c (do_block): Signed/unsigned warning fix. 2580 25812018-11-28 Richard Biener <rguenther@suse.de> 2582 2583 PR tree-optimization/79351 2584 * tree-ssa-sccvn.c (vn_reference_lookup_3): For assignments from 2585 empty CONSTRUCTORs ensure the store is at a constant position. 2586 25872018-11-28 Richard Biener <rguenther@suse.de> 2588 2589 PR tree-optimization/88229 2590 * tree-ssa.c (non_rewritable_mem_ref_base): Check DECL_SIZE_UNIT 2591 is an INTEGER_CST before accessing it so. 2592 25932018-11-28 Sam Tebbs <sam.tebbs@arm.com> 2594 2595 * doc/invoke.texi (-mtune=): Remove obsolete CPU names. 2596 25972018-11-28 Jakub Jelinek <jakub@redhat.com> 2598 2599 PR target/88189 2600 * config/i386/i386.c (ix86_expand_sse_movcc): Handle DFmode and 2601 SFmode using sse4_1_blendvs[sd] with TARGET_SSE4_1. Formatting fixes. 2602 * config/i386/sse.md (sse4_1_blendv<ssemodesuffix>): New pattern. 2603 26042018-11-27 Martin Liska <mliska@suse.cz> 2605 2606 * gcov.c (generate_results): Append current_working_directory 2607 only when exists. 2608 26092018-11-27 Mike Gulick <mgulick@mathworks.com> 2610 2611 PR preprocessor/83173 2612 * input.c (dump_location_info): Dump reason and included_from 2613 fields from line_map_ordinary struct. Fix indentation when 2614 location > 5 digits. 2615 * diagnostic-show-locus.c (num_digits, num_digits): Move to 2616 diagnostic.c to allow it to be utilized by input.c. 2617 * diagnostic.c (num_digits, selftest::test_num_digits): Moved 2618 here. 2619 (selftest::diagnostic_c_tests): Run selftest::test_num_digits. 2620 * diagnostic.h (num_digits): Add extern definition. 2621 26222018-11-27 Fredrik Noring <noring@nocrew.org> 2623 2624 * config/mips/mips.c (mips_reorg_process_insns) 2625 (mips_option_override): Handle `-mfix-r5900'. 2626 * config/mips/mips.h (ASM_SPEC): Add `mfix-r5900' and 2627 `mno-fix-r5900'. 2628 * config/mips/mips.opt (mfix-r5900): New option. 2629 * doc/invoke.texi: Document the `r5900' processor name, and 2630 `-mfix-r5900' and `-mno-fix-r5900' options. 2631 26322018-11-27 Jakub Jelinek <jakub@redhat.com> 2633 2634 PR target/88188 2635 * config/rs6000/rs6000.c (print_operand) <case 'D'>: Use 2636 output_operand_lossage instead of gcc_assert. 2637 <case 't'>: Likewise. 2638 <case 'z'>: Likewise. 2639 <case 'V'>: Use output_operand_lossage instead of gcc_unreachable. 2640 26412018-11-27 Jeff Law <law@redhat.com> 2642 2643 * config/riscv/riscv (riscv_block_mvoe_straight): Use RETURN_BEGIN 2644 in call to move_by_pieces. 2645 2646 * config/sh/sh-mem.c (expand_block_move): Use RETURN_BEGIN in call 2647 to move_by_pieces. 2648 2649 * config/lm32/lm32.c (lm32_block_move_inline): Use RETURN_BEGIN in 2650 call to move_by_pieces. 2651 2652 * config/mips/mips.c (mips_block_move_straight): Use RETURN_BEGIN 2653 in call to move_by_pieces. 2654 2655 * config/microblaze/microblaze.c (microblaze_block_move_straight): Use 2656 RETURN_BEGIN in call to move_by_pieces. 2657 (microblaze_expand_block_move): Likewise. 2658 26592018-11-27 Tamar Christina <tamar.christina@arm.com> 2660 2661 * doc/md.texi (xorsign): Document it. 2662 26632018-11-27 Martin Liska <mliska@suse.cz> 2664 2665 * config/gnu-user.h (TARGET_F951_OPTIONS): New. 2666 * gcc.c (find_fortran_preinclude_file): New function 2667 to handle Fortran pre-include. 2668 26692018-11-27 Martin Liska <mliska@suse.cz> 2670 2671 * asan.c (asan_emit_stack_protection): Use new enum values 2672 instead of int constants. 2673 * builtins.c (expand_builtin_memory_copy_args): Replace int 2674 type with memop_ret enum type. 2675 (expand_builtin_mempcpy_args): Likewise. 2676 (expand_builtin_memcpy): Use new enum values 2677 instead of int constants. Likewise. 2678 (expand_builtin_mempcpy): Likewise. 2679 (expand_movstr): Likewise. 2680 (expand_builtin_strcpy_args): Likewise. 2681 (expand_builtin_stpcpy_1): Likewise. 2682 (expand_builtin_strncpy): Likewise. 2683 (expand_builtin_memset_args): Likewise. 2684 * expr.c (move_by_pieces_d::finish_endp): Rename to ... 2685 (move_by_pieces_d::finish_retmode): ... this. 2686 (move_by_pieces): Change last argument type to memop_ret. 2687 (store_by_pieces): Use new enum values 2688 instead of int constants. 2689 (emit_block_move_hints): Likewise. 2690 (emit_push_insn): Likewise. 2691 (store_expr): Likewise. 2692 * expr.h (store_by_pieces): Change int to newly added enum 2693 type. 2694 * rtl.h (enum memop_ret): Define. 2695 (move_by_pieces): Use the enum type. 2696 26972018-11-27 Alan Modra <amodra@gmail.com> 2698 2699 * config/rs6000/aix71.h (ASM_SPEC): Don't select default -maix64 2700 cpu here. 2701 (ASM_CPU_SPEC): Do so here. Rewrite using if .. else if .. specs 2702 form. Error on missing -mcpu case. 2703 * config/rs6000/driver-rs6000.c (asm_names <_AIX>): Update NULL case. 2704 (asm_names <!_AIX>): Add missing cpus. Update NULL case. Apply 2705 PR63177 fix for -mcpu=power8 and -mcpu=powerpc64le. 2706 * config/rs6000/rs6000.h (ASM_CPU_SPEC): Rewrite using if .. 2707 else if .. specs form. Error on missing -mcpu case. Don't output 2708 duplicate -maltivec. Apply PR63177 fix for -mcpu=powerpc64le. 2709 27102018-11-26 David Malcolm <dmalcolm@redhat.com> 2711 2712 * dump-context.h (dump_context::dump_loc): Convert 1st param from 2713 dump_flags_t to const dump_metadata_t &. Convert 2nd param from 2714 const dump_location_t & to const dump_user_location_t &. 2715 (dump_context::dump_loc_immediate): Convert 2nd param from 2716 const dump_location_t & to const dump_user_location_t &. 2717 (dump_context::dump_gimple_stmt): Convert 1st param from 2718 dump_flags_t to const dump_metadata_t &. 2719 (dump_context::void dump_gimple_stmt_loc): Likewise; convert 2720 2nd param from const dump_location_t & to 2721 const dump_user_location_t &. 2722 (dump_context::dump_gimple_expr): Convert 1st param from 2723 dump_flags_t to const dump_metadata_t &. 2724 (dump_context::dump_gimple_expr_loc): Likewise; convert 2725 2nd param from const dump_location_t & to 2726 const dump_user_location_t &. 2727 (dump_context::dump_generic_expr): Convert 1st param from 2728 dump_flags_t to const dump_metadata_t &. 2729 (dump_context::dump_generic_expr_loc): Likewise; convert 2730 2nd param from const dump_location_t & to 2731 const dump_user_location_t &. 2732 (dump_context::dump_printf_va): Convert 1st param from 2733 dump_flags_t to const dump_metadata_t &. 2734 (dump_context::dump_printf_loc_va): Likewise; convert 2735 2nd param from const dump_location_t & to 2736 const dump_user_location_t &. 2737 (dump_context::dump_dec): Convert 1st param from 2738 dump_flags_t to const dump_metadata_t &. 2739 (dump_context::dump_symtab_node): Likewise. 2740 (dump_context::begin_scope): Split out 2nd param into 2741 user and impl locations. 2742 (dump_context::ensure_pending_optinfo): Add metadata param. 2743 (dump_context::begin_next_optinfo): Replace dump_location_t param 2744 with metadata and user location. 2745 * dumpfile.c (dump_context::dump_loc): Convert 1st param from 2746 dump_flags_t to const dump_metadata_t &. Convert 2nd param from 2747 const dump_location_t & to const dump_user_location_t &. 2748 (dump_context::dump_loc_immediate): Convert 2nd param from 2749 const dump_location_t & to const dump_user_location_t &. 2750 (dump_context::dump_gimple_stmt): Convert 1st param from 2751 dump_flags_t to const dump_metadata_t &. 2752 (dump_context::void dump_gimple_stmt_loc): Likewise; convert 2753 2nd param from const dump_location_t & to 2754 const dump_user_location_t &. 2755 (dump_context::dump_gimple_expr): Convert 1st param from 2756 dump_flags_t to const dump_metadata_t &. 2757 (dump_context::dump_gimple_expr_loc): Likewise; convert 2758 2nd param from const dump_location_t & to 2759 const dump_user_location_t &. 2760 (dump_context::dump_generic_expr): Convert 1st param from 2761 dump_flags_t to const dump_metadata_t &. 2762 (dump_context::dump_generic_expr_loc): Likewise; convert 2763 2nd param from const dump_location_t & to 2764 const dump_user_location_t &. 2765 (dump_context::dump_printf_va): Convert 1st param from 2766 dump_flags_t to const dump_metadata_t &. 2767 (dump_context::dump_printf_loc_va): Likewise; convert 2768 2nd param from const dump_location_t & to 2769 const dump_user_location_t &. 2770 (dump_context::dump_dec): Convert 1st param from 2771 dump_flags_t to const dump_metadata_t &. 2772 (dump_context::dump_symtab_node): Likewise. 2773 (dump_context::begin_scope): Split out 2nd param into 2774 user and impl locations. 2775 (dump_context::ensure_pending_optinfo): Add metadata param. 2776 (dump_context::begin_next_optinfo): Replace dump_location_t param 2777 with metadata and user location. 2778 (dump_gimple_stmt): Convert 1st param from dump_flags_t to 2779 const dump_metadata_t &. 2780 (dump_gimple_stmt_loc): Likewise; convert 2nd param from 2781 const dump_location_t & to const dump_user_location_t &. 2782 (dump_gimple_expr): Convert 1st param from dump_flags_t to 2783 const dump_metadata_t &. 2784 (dump_gimple_expr_loc): Likewise; convert 2nd param from 2785 const dump_location_t & to const dump_user_location_t &. 2786 (dump_generic_expr): Convert 1st param from dump_flags_t to 2787 const dump_metadata_t &. 2788 (dump_generic_expr_loc): Likewise; convert 2nd param from 2789 const dump_location_t & to const dump_user_location_t &. 2790 (dump_printf): Convert 1st param from dump_flags_t to 2791 const dump_metadata_t &. 2792 (dump_printf_loc): Likewise; convert 2nd param from 2793 const dump_location_t & to const dump_user_location_t &. 2794 (dump_dec): Convert 1st param from dump_flags_t to 2795 const dump_metadata_t &. 2796 (dump_symtab_node): Likewise. 2797 (dump_begin_scope): Split out 2nd param into user and impl 2798 locations. 2799 (selftest::assert_impl_location_eq): New function. 2800 (ASSERT_IMPL_LOCATION_EQ): New macro. 2801 (selftest::test_impl_location): Update to use 2802 ASSERT_IMPL_LOCATION_EQ. 2803 (selftest::test_capture_of_dump_calls): Convert "loc" to 2804 dump_user_location_t. Add ASSERT_IMPL_LOCATION_EQ throughout, 2805 verifying line numbers of dump emissions. 2806 * dumpfile.h (class dump_metadata_t): New class. 2807 (dump_printf): Convert 1st param from dump_flags_t to 2808 const dump_metadata_t &. 2809 (dump_printf_loc): Likewise; convert 2nd param from 2810 const dump_location_t & to const dump_user_location_t &. 2811 (dump_generic_expr_loc): Likewise. 2812 (dump_generic_expr): Convert 1st param from dump_flags_t to 2813 const dump_metadata_t &. 2814 (dump_gimple_stmt_loc): Likewise; convert 2nd param from 2815 const dump_location_t & to const dump_user_location_t &. 2816 (dump_gimple_stmt): Convert 1st param from dump_flags_t to 2817 const dump_metadata_t &. 2818 (dump_gimple_expr_loc): Likewise; convert 2nd param from 2819 const dump_location_t & to const dump_user_location_t &. 2820 (dump_gimple_expr): Convert 1st param from dump_flags_t to 2821 const dump_metadata_t &. 2822 (dump_symtab_node): Likewise. 2823 (dump_dec): Likewise. 2824 (dump_begin_scope): Split out 2nd param into user and impl 2825 locations. 2826 (auto_dump_scope::auto_dump_scope): Split "loc" param into a user 2827 location and impl_location, and capture the impl_location. 2828 (AUTO_DUMP_SCOPE): Rename param from LOC to USER_LOC. 2829 * loop-unroll.c (report_unroll): Update for changes to 2830 dump_printf_loc and dump_printf. 2831 * opt-problem.cc (opt_problem::opt_problem): Update for change to 2832 dump_loc. 2833 * optinfo-emit-json.cc 2834 (selftest::test_building_json_from_dump_calls): Convert "loc" from 2835 dump_location_t to dump_user_location_t. 2836 * optinfo.cc (optinfo::emit_for_opt_problem): Update for change in 2837 dump_loc_immediate. 2838 * profile.c (compute_branch_probabilities): Update for change to 2839 dump_printf_loc. 2840 * selftest.h (ASSERT_STR_CONTAINS_AT): New macro. 2841 * tree-vect-slp.c (vect_print_slp_tree): Update for change to 2842 dump_printf_loc. 2843 28442018-11-27 Jozef Lawrynowicz <jozef.l@mittosystems.com> 2845 2846 * doc/sourcebuild.texi: Document check_effective_target_int_eq_float 2847 and check_effective_target_ptr_eq_long. 2848 28492018-11-27 Alan Modra <amodra@gmail.com> 2850 2851 * config.gcc (powerpc*-*-freebsd*, powerpc-*-netbsd*), 2852 (powerpc-*-eabisimaltivec*, powerpc-*-eabisim*, powerpc-*-elf*), 2853 (powerpc-*-eabialtivec*, powerpc-*-eabi*, powerpc-*-rtems*), 2854 (powerpc-wrs-vxworks*, powerpc-*-lynxos*, powerpcle-*-elf*), 2855 (powerpcle-*-eabisim*, powerpcle-*-eabi*): Add gnu-user.h to tm_file. 2856 * config/rs6000/freebsd.h (CPLUSPLUS_CPP_SPEC), 2857 (LINK_GCC_C_SEQUENCE_SPEC): Undef. 2858 (ASM_APP_ON, ASM_APP_OFF): Don't define. 2859 * config/rs6000/freebsd64.h (ASM_APP_ON, ASM_APP_OFF): Don't define. 2860 * config/rs6000/lynx.h (ASM_APP_ON, ASM_APP_OFF): Don't define. 2861 * config/rs6000/linux64.h (LINK_GCC_C_SEQUENCE_SPEC): Define. 2862 * config/rs6000/netbsd.h (CPLUSPLUS_CPP_SPEC), 2863 (LINK_GCC_C_SEQUENCE_SPEC): Undef. 2864 * config/rs6000/rtems.h (LINK_GCC_C_SEQUENCE_SPEC): Define. 2865 * config/rs6000/sysv4.h (GNU_USER_TARGET_CRTI): Redefine. 2866 (GNU_USER_TARGET_CRTN): Redefine. 2867 (CC1_SPEC): Use GNU_USER_TARGET_CC1_SPEC. 2868 (LIB_LINUX_SPEC): Use GNU_USER_TARGET_LIB_SPEC. 2869 (CRTOFFLOADBEGIN, CRTOFFLOADEND): Don't define. 2870 (STARTFILE_LINUX_SPEC): Define as GNU_USER_TARGET_STARTFILE_SPEC. 2871 (ENDFILE_LINUX_SPEC): Define as GNU_USER_TARGET_ENDFILE_SPEC. 2872 (UCLIBC_DYNAMIC_LINKER, CHOOSE_DYNAMIC_LINKER): Don't define. 2873 (LINK_EH_SPEC): Don't define. 2874 28752018-11-26 Martin Sebor <msebor@redhat.com> 2876 2877 PR 87756 2878 * expr.c (string_constant): Handle top-level decls of all character 2879 types and subobjects of narrow character type. 2880 28812018-11-27 Alan Modra <amodra@gmail.com> 2882 2883 * config.gcc (powerpc*-*-linux*): Add linux.h to tm_file. 2884 * config/rs6000/linux.h (TARGET_OS_CPP_BUILTINS): Use 2885 GNU_USER_TARGET_OS_CPP_BUILTINS. 2886 (RS6000_ABI_NAME): Define. 2887 * config/rs6000/linux64.h (TARGET_OS_CPP_BUILTINS): Use 2888 GNU_USER_TARGET_OS_CPP_BUILTINS. 2889 (MUSL_DYNAMIC_LINKER32): Undef before defining. 2890 (UCLIBC_DYNAMIC_LINKER32, UCLIBC_DYNAMIC_LINKER64): Don't define. 2891 (CHOOSE_DYNAMIC_LINKER): Don't define. 2892 (GNU_USER_DYNAMIC_LINKER32, GNU_USER_DYNAMIC_LINKER64): Don't define. 2893 * config/rs6000/sysv4.h (MUSL_DYNAMIC_LINKER): Undef before defining. 2894 (CHOOSE_DYNAMIC_LINKER, GNU_USER_DYNAMIC_LINKER): Only define when 2895 not already defined. 2896 (CPP_OS_LINUX_SPEC): Remove defines and asserts handled by 2897 TARGET_OS_CPP_BUILTINS. 2898 28992018-11-26 Jozef Lawrynowicz <jozef.l@mittosystems.com> 2900 2901 * doc/sourcebuild.texi: Document check_effective_target_newlib_nano_io. 2902 29032018-11-26 Jakub Jelinek <jakub@redhat.com> 2904 2905 PR target/88195 2906 * config/i386/i386.c (def_builtin2): If tcode == VOID_FTYPE_UINT64 2907 and !TARGET_64BIT, return NULL_TREE. 2908 2909 PR c++/86900 2910 * dwarf2out.c (secname_for_decl): For functions with 2911 DECL_SECTION_NAME if in_cold_section_p, try to return 2912 current_function_section's name if it is a named section. 2913 29142018-11-26 Maya Rashish <coypu@sdf.org> 2915 2916 PR target/58397 2917 * config.host (*-*-netbsd*): Use x-netbsd and host-netbsd.o. 2918 * config/x-netbsd: New file. 2919 * config/host-netbsd: New file. 2920 29212018-11-26 Jozef Lawrynowicz <jozef.l@mittosystems.com> 2922 2923 * doc/sourcebuild.texi: Document check_effective_target_size20plus. 2924 Clarify documentation for check_effective_target_size32plus. 2925 29262018-11-26 Aaron Sawdey <acsawdey@linux.ibm.com> 2927 2928 * config/rs6000/rs6000-string.c (expand_cmp_vec_sequence): Rename 2929 and modify expand_strncmp_vec_sequence. 2930 (emit_final_compare_vec): Rename and modify emit_final_str_compare_vec. 2931 (generate_6432_conversion): New function. 2932 (expand_block_compare): Add support for vsx. 2933 (expand_block_compare_gpr): New function. 2934 * config/rs6000/rs6000.opt (rs6000_block_compare_inline_limit): Increase 2935 default limit to 63 because of more compact vsx code. 2936 29372018-11-26 Uros Bizjak <ubizjak@gmail.com> 2938 2939 PR target/88178 2940 * config/i386/i386.c (dbx_register_map): Use IGNORED_DWARF_REGNUM 2941 for registers for which no debug information can be generated. 2942 Use INVALID_REGNUM for non-existent registers. 2943 (dbx64_register_map): Ditto. 2944 (svr4_dbx_register_map): Ditto. 2945 29462018-11-26 Sandra Loosemore <sandra@codesourcery.com> 2947 2948 PR c/57182 2949 * doc/invoke.texi (Optimize Options): Document that 2950 -freorder-functions also uses "hot" and "cold" attributes. 2951 29522018-11-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 2953 2954 * config/aarch64/aarch64.c (aarch64_mangle_type): Fix typo in comment. 2955 29562018-11-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 2957 2958 * config/arm/types.md (mul64): Rename to... 2959 (widen_mul64): ... This. 2960 * config/arm/arm-generic.md: Rename mul64 to widen_mul64. 2961 * config/arm/cortex-a15.md: Likewise. 2962 * config/arm/cortex-a5.md: Likewise. 2963 * config/arm/cortex-a53.md: Likewise. 2964 * config/arm/cortex-a57.md: Likewise. 2965 * config/arm/cortex-a7.md: Likewise. 2966 * config/arm/cortex-m4.md: Likewise. 2967 * config/arm/exynos-m1.md: Likewise. 2968 * config/arm/marvell-pj4.md: Likewise. 2969 * config/arm/xgene1.md: Likewise. 2970 29712018-11-26 Richard Biener <rguenther@suse.de> 2972 2973 PR tree-optimization/88182 2974 * tree-vect-loop.c (vectorizable_reduction): Pick up single 2975 correct reduc_def_info. 2976 * tree-vect-slp.c (vect_analyze_slp_instance): Set 2977 STMT_VINFO_REDUC_DEF of the first stmt. 2978 29792018-11-26 Andreas Krebbel <krebbel@linux.ibm.com> 2980 2981 * doc/invoke.texi: Document z14/arch12 -march option. 2982 29832018-11-25 Sandra Loosemore <sandra@codesourcery.com> 2984 2985 PR c/57166 2986 * doc/invoke.texi (Option Summary): Add -Wmissing-noreturn. 2987 (Warning Options): Likewise. 2988 29892018-11-25 Sandra Loosemore <sandra@codesourcery.com> 2990 2991 PR web/79738 2992 * doc/extend.texi (Common Function Attributes): Clarify that 2993 functions with "const" attribute can read const global variables. 2994 29952018-11-26 Alan Modra <amodra@gmail.com> 2996 2997 * config/rs6000/rs6000.h (OBJECT_PEF, TARGET_MACOS): Delete. 2998 29992018-11-25 Sandra Loosemore <sandra@codesourcery.com> 3000 3001 PR other/54265 3002 * doc/extend.texi (Common Variable Attributes): Use preferred 3003 placement of type attributes in examples, plus whitespace fixes. 3004 (Type Attributes): Clarify why placement of attributes 3005 immediately after struct/union/enum keyword is preferred. 3006 (Common Type Attributes): Use preferred placement of type 3007 attributes in examples, plus more whitespace fixes. 3008 30092018-11-25 Paul Koning <ni1d@arrl.net> 3010 3011 * config/pdp11/pdp11.h (TARGET_HAS_NO_HW_DIVIDE): Define. 3012 30132018-11-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE> 3014 3015 * doc/sourcebuild.texi (Effective-Target Keywords, Environment 3016 attributes): Document d_runtime. 3017 30182018-11-25 Vladimir Makarov <vmakarov@redhat.com> 3019 3020 PR bootstrap/88157 3021 * ira-costs.c (record_operand_costs): Initiate register move cost 3022 for mode. 3023 30242018-11-23 Jeff Law <law@redhat.com> 3025 3026 PR rtl-optimization/87468 3027 * tree-ssa-threadupdate.c (create_block_for_threading): Clear 3028 EDGE_IGNORE on all outgoing edges of the duplicate block. 3029 30302018-11-23 Vladimir Makarov <vmakarov@redhat.com> 3031 3032 PR bootstrap/88157 3033 * ira-costs.c (record_operand_costs): Use bigger hard reg class if 3034 its mode does not fit to the original class. 3035 30362018-11-23 Martin Sebor <msebor@redhat.com> 3037 3038 PR tree-optimization/87756 3039 * expr.c (string_constant): Synthesize a string literal from 3040 the address of a constant character. 3041 * tree.c (build_string_literal): Add an argument. 3042 * tree.h (build_string_literal): Same. 3043 30442018-11-23 Christoph Muellner <christoph.muellner@theobroma-systems.com> 3045 3046 * config/aarch64/aarch64-cores.def: Define emag. 3047 * config/aarch64/aarch64-tune.md: Regenerated with emag. 3048 * config/aarch64/aarch64.c (emag_tunings): New struct. 3049 * doc/invoke.texi: Document mtune value. 3050 30512018-11-23 Christoph Muellner <christoph.muellner@theobroma-systems.com> 3052 Philipp Tomsich <philipp.tomsich@theobroma-systems.com> 3053 3054 * config/aarch64/aarch64.c (xgene1_tunings): Optimize Xgene1 tunings 3055 for GCC 9. 3056 * config/aarch64/aarch64.c (xgene1_tunings): Add Xgene1 specific 3057 prefetch tunings. 3058 * config/aarch64/aarch64.c (xgene1_addrcost_table): Correct the post-modify 3059 costs. 3060 * config/arm/aarch-cost-tables.h (xgene1_extra_costs): Update the cost table 3061 for XGene1. 3062 30632018-11-23 Richard Biener <rguenther@suse.de> 3064 3065 PR tree-optimization/88149 3066 * tree-vect-slp.c (vect_slp_analyze_node_operations): Detect 3067 the case where there are two different def types for the 3068 same operand at different operand position in the same stmt. 3069 30702018-23-11 Mihail Ionescu <mihail.ionescu@arm.com> 3071 3072 * config/arm/arm.c (arm_expand_compare_and_swap): Simplify and call 3073 gen_atomic_compare_swap_1. 3074 (arm_evpc_neon_vuzp): Likewise gen_neon_vuzp_internal. 3075 (arm_evpc_neon_vtrn): Likewise gen_neon_vtrn_internal. 3076 (arm_evpc_neon_vext): Likewise gen_neon_vext_internal. 3077 (arm_evpc_neon_vzip): Likewise gen_neon_vzip_internal. 3078 (arm_evpc_neon_vrev): Replace the function pointer and simplify the mode 3079 checks. 3080 * config/arm/arm.md (neon_vext<mode>), 3081 (neon_vrev64<mode>, neon_vrev32<mode>), 3082 (neon_vrev16<mode>, neon_vtrn<mode>_internal), 3083 (neon_vzip<mode>_internal, neon_vuzp<mode>_internal): Add an '@'character 3084 before the pattern name. 3085 * config/arm/sync.md: 3086 (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1), 3087 (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise. 3088 30892018-11-23 Jakub Jelinek <jakub@redhat.com> 3090 3091 PR tree-optimization/86614 3092 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return early 3093 if TREE_NO_WARNING is set on ref.ref. 3094 30952018-11-22 Sandra Loosemore <sandra@codesourcery.com> 3096 Alan Coopersmith <alan.coopersmith@oracle.com> 3097 3098 PR c/53608 3099 * doc/extend.texi (Designated Inits): Clarify handling of multiple 3100 initializers for unions. 3101 31022018-11-22 Jan Hubicka <jh@suse.cz> 3103 3104 PR lto/88142 3105 * ipa-devirt.c (type_variants_equivalent_p): Drop warn and warned 3106 parameters; do not warn here. 3107 (odr_subtypes_equivalent_p): Likewise. 3108 (warn_odr): Fix typo. 3109 (warn_types_mismatch): Do not output confused warnings on integer types. 3110 (odr_types_equivalent_p): Update. 3111 31122018-11-22 Jakub Jelinek <jakub@redhat.com> 3113 3114 * config/i386/i386.c (ix86_option_override_internal, 3115 ix86_can_inline_p, classify_argument, construct_container, 3116 ix86_expand_prologue, ix86_expand_split_stack_prologue, 3117 ix86_expand_carry_flag_compare, expand_set_or_movmem_via_loop, 3118 expand_setmem_epilogue_via_loop, promote_duplicated_reg, 3119 ix86_expand_set_or_movmem, ix86_init_builtins_va_builtins_abi): 3120 Formatting fixes. 3121 3122 * config/i386/i386.c (ix86_option_override_internal): For 3123 stack_protector_guard related options, use opts_set->x_ instead 3124 of global_options_set. and prefix options with opts->x_ . Move 3125 defaults for offset and reg into else block. 3126 31272018-11-22 Uros Bizjak <ubizjak@gmail.com> 3128 3129 * config/i386/i386.c (ix86_check_avx_upper_register): 3130 Return true for all SSE registers with mode bitsize > 128. 3131 31322018-11-22 Vladimir Makarov <vmakarov@redhat.com> 3133 3134 PR rtl-optimization/87718 3135 * ira-costs.c: Remove trailing white-spaces. 3136 (record_operand_costs): Add a special treatment for moves 3137 involving a hard register. 3138 31392018-11-22 Uros Bizjak <ubizjak@gmail.com> 3140 3141 * config/i386/i386.c (ix86_avx_emit_vzeroupper): Remove. 3142 (ix86_emit_mode_set) <case AVX_U128>: Emit vzeroupper here. 3143 31442018-11-22 Martin Liska <mliska@suse.cz> 3145 3146 * common/config/i386/i386-common.c (processor_names): Add 3147 static assert and add missing "znver2". 3148 (ix86_get_valid_option_values): Add checking assert for null 3149 values and add "native" value if feasible. 3150 * config/i386/i386.h: Do not declare size of processor_names. 3151 * common/config/i386/i386-common.c: 3152 * config/i386/i386.c: Add static assert for size 3153 of processor_cost_table. 3154 31552018-11-22 Thomas Preud'homme <thomas.preudhomme@linaro.org> 3156 3157 * target-insns.def (stack_protect_combined_set): Define new standard 3158 pattern name. 3159 (stack_protect_combined_test): Likewise. 3160 * cfgexpand.c (stack_protect_prologue): Try new 3161 stack_protect_combined_set pattern first. 3162 * function.c (stack_protect_epilogue): Try new 3163 stack_protect_combined_test pattern first. 3164 * config/arm/arm.c (require_pic_register): Add pic_reg and compute_now 3165 parameters to control which register to use as PIC register and force 3166 reloading PIC register respectively. Insert in the stream of insns if 3167 possible. 3168 (legitimize_pic_address): Expose above new parameters in prototype and 3169 adapt recursive calls accordingly. Use pic_reg if non null instead of 3170 cached one. 3171 (arm_load_pic_register): Add pic_reg parameter and use it if non null. 3172 (arm_legitimize_address): Adapt to new legitimize_pic_address 3173 prototype. 3174 (thumb_legitimize_address): Likewise. 3175 (arm_emit_call_insn): Adapt to require_pic_register prototype change. 3176 (arm_expand_prologue): Adapt to arm_load_pic_register prototype change. 3177 (thumb1_expand_prologue): Likewise. 3178 * config/arm/arm-protos.h (legitimize_pic_address): Adapt to prototype 3179 change. 3180 (arm_load_pic_register): Likewise. 3181 * config/arm/predicated.md (guard_addr_operand): New predicate. 3182 (guard_operand): New predicate. 3183 * config/arm/arm.md (movsi expander): Adapt to legitimize_pic_address 3184 prototype change. 3185 (builtin_setjmp_receiver expander): Adapt to thumb1_expand_prologue 3186 prototype change. 3187 (stack_protect_combined_set): New expander.. 3188 (stack_protect_combined_set_insn): New insn_and_split pattern. 3189 (stack_protect_set_insn): New insn pattern. 3190 (stack_protect_combined_test): New expander. 3191 (stack_protect_combined_test_insn): New insn_and_split pattern. 3192 (arm_stack_protect_test_insn): New insn pattern. 3193 * config/arm/thumb1.md (thumb1_stack_protect_test_insn): New insn pattern. 3194 * config/arm/unspecs.md (UNSPEC_SP_SET): New unspec. 3195 (UNSPEC_SP_TEST): Likewise. 3196 * doc/md.texi (stack_protect_combined_set): Document new standard 3197 pattern name. 3198 (stack_protect_set): Clarify that the operand for guard's address is 3199 legal. 3200 (stack_protect_combined_test): Document new standard pattern name. 3201 (stack_protect_test): Clarify that the operand for guard's address is 3202 legal. 3203 32042018-11-22 Richard Biener <rguenther@suse.de> 3205 3206 PR tree-optimization/88148 3207 * tree-ssa-loop-niter.c (simplify_replace_tree): Get optional 3208 valueization callback parameter and handle it. 3209 * tree-ssa-loop-niter.h (simplify_replace_tree): Export. 3210 * tree-ssa-sccvn.c (process_bb): Eliminate in loop niter trees. 3211 32122018-11-22 Richard Biener <rguenther@suse.de> 3213 3214 PR lto/87229 3215 PR lto/88112 3216 * lto-streamer-out.c (lto_is_streamable): Allow CALL_EXPRs 3217 which can appear in size expressions. 3218 * tree-streamer-in.c (unpack_ts_base_value_fields): Stream 3219 CALL_EXPR_BY_DESCRIPTOR. 3220 (streamer_read_tree_bitfields): Stream CALL_EXPR_IFN. 3221 * tree-streamer-out.c (pack_ts_base_value_fields): Stream 3222 CALL_EXPR_BY_DESCRIPTOR. 3223 (streamer_write_tree_bitfields): Stream CALL_EXPR_IFN. 3224 3225 Revert 3226 PR lto/87229 3227 * tree.c (free_lang_data_in_one_sizepos): Free non-gimple-val 3228 sizepos values. 3229 32302018-11-22 Richard Biener <rguenther@suse.de> 3231 3232 PR tree-optimization/88069 3233 * tree-ssa-sccvn.c (visit_phi): Tweak previous fix to not 3234 apply to default defs. 3235 32362018-11-22 Jakub Jelinek <jakub@redhat.com> 3237 3238 PR target/85644 3239 PR target/86832 3240 * config/i386/i386.c (ix86_option_override_internal): Default 3241 ix86_stack_protector_guard to SSP_TLS only if TARGET_THREAD_SSP_OFFSET 3242 is defined. 3243 * config/i386/i386.md (stack_protect_set, stack_protect_set_<mode>, 3244 stack_protect_test, stack_protect_test_<mode>): Use empty condition 3245 instead of TARGET_SSP_TLS_GUARD. 3246 32472018-11-22 Martin Liska <mliska@suse.cz> 3248 3249 PR sanitizer/88017 3250 * config/i386/i386.c (ix86_option_override_internal): Error 3251 about usage -mabi=ms and -fsanitize=thread. 3252 32532018-11-21 Uros Bizjak <ubizjak@gmail.com> 3254 3255 Revert the revert: 3256 2013-10-26 Vladimir Makarov <vmakarov@redhat.com> 3257 3258 Revert: 3259 2013-10-25 Vladimir Makarov <vmakarov@redhat.com> 3260 3261 * lra-spills.c (lra_final_code_change): Remove useless move insns. 3262 32632018-11-21 Uros Bizjak <ubizjak@gmail.com> 3264 3265 PR middle-end/88129 3266 * function.c (expand_function_end): Do not emit extra blockage insn. 3267 32682018-11-21 Lokesh Janghel <lokeshjanghel91@gmail.com> 3269 3270 PR target/85667 3271 * config/i386/i386.c (function_value_ms_64): Return AX_REG instead 3272 of FIRST_SSE_REG for 4 or 8 byte modes. 3273 32742018-11-21 Jan Hubicka <jh@suse.cz> 3275 3276 PR lto/87957 3277 * tree.c (fld_decl_context): Break out from ... 3278 (free_lang_data_in_decl): ... here; free TREE_PUBLIC, TREE_PRIVATE 3279 DECL_ARTIFICIAL of TYPE_DECL; do not free TREE_TYPE of TYPE_DECL. 3280 (fld_incomplete_type_of): Build copy of TYP_DECL. 3281 * ipa-devirt.c (free_enum_values): Rename to ... 3282 (free_odr_warning_data): ... this one; free also duplicated TYPE_DECLs 3283 and TREE_TYPEs of TYPE_DECLs. 3284 (get_odr_type): Initialize odr_vtable_hash if needed. 3285 32862018-11-21 Alexandre Oliva <oliva@adacore.com> 3287 3288 * final.c (compute_discriminator): Declare. Renamed from... 3289 (maybe_set_discriminator): ... this. Set and return a local. 3290 (override_discriminator): New. 3291 (final_scan_insn_1): Set it. 3292 (notice_source_line): Adjust. Always set discriminator. 3293 32942018-11-21 Jakub Jelinek <jakub@redhat.com> 3295 3296 PR target/87839 3297 * config/aarch64/atomics.md (@aarch64_compare_and_swap<mode>): Use 3298 rIJ constraint for aarch64_plus_operand rather than rn. 3299 33002018-11-21 Renlin Li <renlin.li@arm.com> 3301 3302 PR middle-end/84877 3303 * explow.h (get_dynamic_stack_size): Declare it as external. 3304 * explow.c (record_new_stack_level): Remove function static attribute. 3305 * function.c (assign_stack_local_1): Dynamically align the stack slot 3306 addr for parameter copy on the stack. 3307 33082018-11-21 Richard Biener <rguenther@suse.de> 3309 3310 PR bootstrap/88133 3311 * bitmap.c (bitmap_last_set_bit): Refactor to avoid warning. 3312 * Makefile.in (bitmap.o-warn): Remove again. 3313 33142018-11-20 Jeff Law <law@redhat.com> 3315 3316 PR tree-optimization/88069 3317 * tree-ssa-dom.c (record_equivalences_from_phis): Propagate away 3318 degenerate virtual PHIs. 3319 33202018-11-21 H.J. Lu <hongjiu.lu@intel.com> 3321 3322 PR target/87317 3323 * config/i386/sse.md (sse4_1_<code>v8qiv8hi2<mask_name>): Replace 3324 nonimmediate_operand with register_operand. 3325 (avx2_<code>v8qiv8si2<mask_name>): Likewise. 3326 (sse4_1_<code>v4qiv4si2<mask_name>): Likewise. 3327 (sse4_1_<code>v4hiv4si2<mask_name>): Likewise. 3328 (sse4_1_<code>v2qiv2di2<mask_name>): Likewise. 3329 (avx512f_<code>v8qiv8di2<mask_name>): Likewise. 3330 (avx2_<code>v4qiv4di2<mask_name>): Likewise. 3331 (avx2_<code>v4hiv4di2<mask_name>): Likewise. 3332 (sse4_1_<code>v2hiv2di2<mask_name>): Likewise. 3333 (sse4_1_<code>v2siv2di2<mask_name>): Likewise. 3334 (*sse4_1_<code>v8qiv8hi2<mask_name>_1): New pattern. 3335 (*sse4_1_<code>v8qiv8hi2<mask_name>_2): Likewise. 3336 (*avx2_<code>v8qiv8si2<mask_name>_1): Likewise. 3337 (*avx2_<code>v8qiv8si2<mask_name>_2): Likewise. 3338 (*sse4_1_<code>v4qiv4si2<mask_name>_1): Likewise. 3339 (*sse4_1_<code>v4qiv4si2<mask_name>_2): Likewise. 3340 (*sse4_1_<code>v4hiv4si2<mask_name>_1): Likewise. 3341 (*sse4_1_<code>v4hiv4si2<mask_name>_2): Likewise. 3342 (*avx512f_<code>v8qiv8di2<mask_name>_1): Likewise. 3343 (*avx512f_<code>v8qiv8di2<mask_name>_2): Likewise. 3344 (*avx2_<code>v4qiv4di2<mask_name>_1): Likewise. 3345 (*avx2_<code>v4qiv4di2<mask_name>_2): Likewise. 3346 (*avx2_<code>v4hiv4di2<mask_name>_1): Likewise. 3347 (*avx2_<code>v4hiv4di2<mask_name>_2): Likewise. 3348 (*sse4_1_<code>v2hiv2di2<mask_name>_1): Likewise. 3349 (*sse4_1_<code>v2hiv2di2<mask_name>_2): Likewise. 3350 (*sse4_1_<code>v2siv2di2<mask_name>_1): Likewise. 3351 (*sse4_1_<code>v2siv2di2<mask_name>_2): Likewise. 3352 33532018-11-21 H.J. Lu <hongjiu.lu@intel.com> 3354 3355 * read-rtl.c (apply_subst_iterator): Handle define_split and 3356 define_insn_and_split. 3357 33582018-11-21 Jakub Jelinek <jakub@redhat.com> 3359 3360 PR rtl-optimization/87817 3361 * config/i386/i386.md (bmi2_bzhi_<mode>3, *bmi2_bzhi_<mode>3, 3362 *bmi2_bzhi_<mode>3_1, *bmi2_bzhi_<mode>3_1_ccz): Use IF_THEN_ELSE 3363 in the pattern to avoid triggering UB when operands[2] is zero. 3364 (tbm_bextri_<mode>): New expander. Renamed the old define_insn to ... 3365 (*tbm_bextri_<mode>): ... this. 3366 33672018-11-21 Tom de Vries <tdevries@suse.de> 3368 3369 PR driver/79855 3370 * params.def (HOT_BB_COUNT_FRACTION): Terminate help message with 3371 period. 3372 33732018-11-21 Andreas Krebbel <krebbel@linux.ibm.com> 3374 3375 * configure.ac: Add check for Binutils to determine whether vector 3376 load/store alignments hints are being supported. 3377 * config.in: Regenerate. 3378 * configure: Regenerate. 3379 * config/s390/s390.c (print_operand): Support new output 3380 modifier A. 3381 * config/s390/s390.md ("movti"): Append alignment hint output 3382 using the new output modifier 'A'. 3383 * config/s390/vector.md ("mov<mode>", "*vec_tf_to_v1tf") 3384 ("*vec_ti_to_v1ti"): Likewise. 3385 33862018-11-20 Martin Sebor <msebor@redhat.com> 3387 3388 * doc/extend.texi (Other Builtins): Add __builtin_has_attribute. 3389 33902018-11-20 Jan Hubicka <hubicka@ucw.cz> 3391 3392 PR lto/84044 3393 * ipa-devirt.c (odr_types_equivalent_p): Use operand_equal_p to 3394 compare ENUM values. 3395 33962018-11-20 Jakub Jelinek <jakub@redhat.com> 3397 3398 PR tree-optimization/87895 3399 * omp-simd-clone.c (ipa_simd_modify_function_body): When removing 3400 or replacing GIMPLE_RETURN, set EDGE_FALLTHRU on the edge to EXIT. 3401 (simd_clone_adjust): Don't set EDGE_FALLTHRU here. In a loop that 3402 redirects edges to EXIT to edges to incr_bb, iterate while EXIT 3403 has any preds and always use EDGE_PRED (, 0). 3404 34052018-11-20 Uros Bizjak <ubizjak@gmail.com> 3406 3407 PR target/88070 3408 * mode-switching.c (create_pre_exit): After reload, always split the 3409 fallthrough edge to the exit block. 3410 34112018-11-20 Jan Hubicka <hubicka@ucw.cz> 3412 3413 * ipa-devirt.c (add_type_duplicate): Do not ICE on incomplete enums. 3414 * tree.c (build_array_type_1): Forward declare. 3415 (fld_type_variant_equal_p): Add INNER_TYPE parameter. 3416 (fld_type_variant): Likewise. 3417 (fld_simplified_types): New hash. 3418 (fld_process_array_type): New function. 3419 (fld_incomplete_type_of): Handle array and enumeration types. 3420 (fld_simplified_type): Handle simplification of arrays. 3421 (free_lang_data): Allocate and free simplified types hash. 3422 34232018-11-20 Jan Hubicka <hubicka@ucw.cz> 3424 3425 PR lto/87957 3426 * ipa-devirt.c (odr_subtypes_equivalent_p): Report ODR violation 3427 when sybtype already violates ODR. 3428 (get_odr_type): Do not ICE when insert is false and type duplicate 3429 is not registered yet. 3430 (register_odr_type): Be sure to register subtypes first. 3431 34322018-11-20 Andreas Krebbel <krebbel@linux.ibm.com> 3433 3434 * config/s390/s390.md ("clztidi2"): Swap the RTX's written to the 3435 DImode parts of the target operand. 3436 34372018-11-20 Nathan Sidwell <nathan@acm.org> 3438 3439 PR 87926 3440 * Makefile.in (bitmap.o-warn): Use -Wno-error=array-bounds. 3441 34422018-11-20 Jan Hubicka <hubicka@ucw.cz> 3443 3444 PR lto/87997 3445 * tree.c (free_lang_data_in_cgraph): Add argument fld; break out 3446 type checking to... 3447 (free_lang_data) ... here; update call of free_lang_data_in_cgraph. 3448 34492018-11-20 Jan Hubicka <hubicka@ucw.cz> 3450 3451 PR ipa/87706 3452 * ipa-fnsummary.c (pass_ipa_fnsummary): Do not remove functions 3453 * ipa.c (possible_inline_candidate_p): Break out from .. 3454 (process_references): ... here ; drop before_inlining_p; 3455 cleanup handling of alises. 3456 (walk_polymorphic_call_targets): Likewise. 3457 (symbol_table::remove_unreachable_nodes): Likewise. 3458 * passes.c (pass_data_ipa_remove_symbols): New structure. 3459 (pass_ipa_remove_symbols): New pass. 3460 (make_pass_ipa_remove_symbols): New function. 3461 * tree-pass.h (make_pass_ipa_remove_symbols): Declare. 3462 * passes.def (pass_ipa_remove_symbols): Schedule after early passes. 3463 34642018-11-20 Richard Biener <rguenther@suse.de> 3465 3466 * tree-vect-stmts.c (vectorizable_condition): Do not get 3467 at else_clause vect def for EXTRACT_LAST_REDUCTION. Remove 3468 pointless vect_is_simple_use calls. 3469 34702018-11-20 Richard Biener <rguenther@suse.de> 3471 3472 PR tree-optimization/88074 3473 * tree-ssa-dom.c (pass_dominator::execute): Do not walk 3474 backedges. 3475 34762018-11-20 Richard Biener <rguenther@suse.de> 3477 3478 PR tree-optimization/88069 3479 * tree-ssa-sccvn.c (visit_phi): Do not value-number to unvisited 3480 virtual PHI arguments. 3481 34822018-11-20 Ilya Leoshkevich <iii@linux.ibm.com> 3483 3484 PR target/88083 3485 * config/s390/s390.md: Skip LT(G) peephole when literal pool is 3486 involved. 3487 * rtl.h (contains_constant_pool_address_p): New function. 3488 * rtlanal.c (contains_constant_pool_address_p): Likewise. 3489 34902018-11-20 Richard Biener <rguenther@suse.de> 3491 3492 PR middle-end/83215 3493 * alias.c (component_uses_parent_alias_set_from): Remove 3494 alias-set zero and TYPE_TYPELESS_STORAGE case both already 3495 handled in other ways. 3496 34972018-11-20 Richard Biener <rguenther@suse.de> 3498 3499 PR tree-optimization/88087 3500 * tree-ssa-pre.c (create_expression_by_pieces): Re-materialize 3501 call fntype. 3502 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Remember 3503 call fntype. 3504 35052018-11-20 Richard Biener <rguenther@suse.de> 3506 3507 PR middle-end/88089 3508 * tree-data-ref.c (lambda_matrix_right_hermite): Use abs_hwi. 3509 35102018-11-19 Jan Hubicka <hubicka@ucw.cz> 3511 3512 PR lto/87957 3513 * ipa-devirt.c (free_enum_values): Do not ICE on ODR vilations. 3514 35152018-11-19 Sandra Loosemore <sandra@codesourcery.com> 3516 3517 PR driver/50250 3518 * doc/invoke.texi (Link Options): Mention shared libraries 3519 in documentation for the -l option. Simplify discussion and 3520 point to the system linker documentation for details. 3521 35222018-11-19 Peter Bergner <bergner@linux.ibm.com> 3523 3524 PR rtl-optimization/88033 3525 * ira-lives.c (non_conflicting_reg_copy_p): Skip copies from a register 3526 to itself. Use HARD_REGISTER_NUM_P. 3527 35282018-11-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 3529 3530 * tree-vect-loop.c (vect_transform_loop): Disable further unrolling 3531 of the loop if vf is non-constant. 3532 35332018-11-19 David Malcolm <dmalcolm@redhat.com> 3534 3535 PR tree-optimization/87025 3536 * dumpfile.c (dump_context::begin_scope): Call end_any_optinfo 3537 immediately after creating the scope optinfo. 3538 (selftest::test_pr87025): New function. 3539 (selftest::dumpfile_c_tests): Call it. 3540 * optinfo-emit-json.cc (optrecord_json_writer::pop_scope): Assert 3541 that we're not popping the top-level records array. 3542 * optinfo.cc (optinfo::handle_dump_file_kind): Assert that we're 3543 not changing the kind of a "scope" optinfo. 3544 35452018-11-19 David Malcolm <dmalcolm@redhat.com> 3546 3547 PR tree-optimization/87025 3548 * dump-context.h: Include "optinfo.h". 3549 (class optrecord_json_writer): New forward decl. 3550 (dump_context::forcibly_enable_optinfo_p): Delete. 3551 (dump_context::optinfo_enabled_p): New member function. 3552 (dump_context::optimization_records_enabled_p): New member 3553 function. 3554 (dump_context::set_json_writer): New member function. 3555 (dump_context::emit_optinfo): New member function. 3556 (dump_context::m_forcibly_enable_optinfo): Delete. 3557 (dump_context::m_json_writer): New member data. 3558 * dumpfile.c (dump_context::set_json_writer): New member function. 3559 (dump_context::finish_any_json_writer): New member function. 3560 (dump_context::end_scope): Replace call to 3561 optimization_records_maybe_pop_dump_scope with call to 3562 m_json_writer->pop_scope. 3563 (dump_context::optinfo_enabled_p): New member function. 3564 (dump_context::end_any_optinfo): Replace call to optinfo::emit with call 3565 to dump_context::emit_optinfo. 3566 (dump_context::emit_optinfo): New member function. 3567 (temp_dump_context::temp_dump_context): Replace 3568 m_forcibly_enable_optinfo with call to set_json_writer. 3569 (temp_dump_context::~temp_dump_context): Clean up any json writer. 3570 * optinfo-emit-json.cc (class optrecord_json_writer): Move to 3571 optinfo-emit-json.h 3572 (the_json_writer): Delete. 3573 (optimization_records_start): Delete. 3574 (optimization_records_finish): Delete. 3575 (optimization_records_enabled_p): Delete, in favor of 3576 dump_context::optimization_records_enabled_p. 3577 (optimization_records_maybe_record_optinfo): Delete. 3578 (optimization_records_maybe_pop_dump_scope): Delete. 3579 * optinfo-emit-json.h: Include "json.h". Delete forward 3580 decl of opt_pass. 3581 (optimization_records_start): Delete. 3582 (optimization_records_finish): Delete. 3583 (optimization_records_enabled_p): Delete. 3584 (optimization_records_maybe_record_optinfo): Delete. 3585 (optimization_records_maybe_pop_dump_scope): Delete. 3586 (class optrecord_json_writer): Move here from 3587 optinfo-emit-json.cc. 3588 * optinfo.cc (optinfo::emit_for_opt_problem): Replace call 3589 to optinfo::emit with call to dump_context::emit_optinfo. 3590 (optinfo::emit): Delete, in favor of dump_context::emit_optinfo. 3591 (optinfo_enabled_p): Delete, in favor of 3592 dump_context::optinfo_enabled_p. 3593 (optinfo_wants_inlining_info_p): Update for conversion o 3594 optimization_records_enabled_p to a member function of 3595 dump_context. 3596 * optinfo.h (optinfo_enabled_p): Delete, in favor of 3597 dump_context::optinfo_enabled_p. 3598 (optinfo::emit): Delete, in favor of dump_context::emit_optinfo. 3599 * toplev.c: Include "dump-context.h". 3600 (compile_file): Replace call to optimization_records_finish with 3601 dump_context::finish_any_json_writer. 3602 (do_compile): Replace call to optimization_records_start with 3603 conditionally creating a optrecord_json_writer for the 3604 dump_context. 3605 36062018-11-19 Jakub Jelinek <jakub@redhat.com> 3607 3608 PR tree-optimization/88071 3609 * tree-vect-loop.c (vectorize_fold_left_reduction): Pass true instead 3610 of false as last argument to gsi_remove. 3611 * tree-vect-stmts.c (vect_finish_replace_stmt): Pass true instead of 3612 false as last argument to gsi_replace. 3613 3614 PR debug/87039 3615 * omp-expand.c: Don't include debug.h. 3616 (adjust_context_and_scope): Add REGION argument. Find DECL_CONTEXT 3617 from innermost outer parallel, task, teams or target that has a 3618 child_fn set, or, if there is no such outer region, use 3619 current_function_decl. Do the DECL_CONTEXT adjustment regardless of 3620 whether a suitable BLOCK is found or not. 3621 (expand_parallel_call, expand_teams_call): Don't call 3622 adjust_context_and_scope here. 3623 (grid_expand_target_grid_body): Revert 2017-01-25 changes. 3624 (expand_omp_taskreg, expand_omp_target): Likewise. Call 3625 adjust_context_and_scope. 3626 * dwarf2out.c (dwarf2out_early_global_decl): For 3627 decl_function_context recurse instead of calling dwarf2out_decl. 3628 36292018-11-20 Eric Botcazou <ebotcazou@adacore.com> 3630 3631 PR rtl-optimization/85925 3632 * rtl.h (word_register_operation_p): New predicate. 3633 * combine.c (record_dead_and_set_regs_1): Only apply specific handling 3634 for WORD_REGISTER_OPERATIONS targets to word_register_operation_p RTX. 3635 * rtlanal.c (nonzero_bits1): Likewise. Adjust couple of comments. 3636 (num_sign_bit_copies1): Likewise. 3637 36382018-11-19 Richard Biener <rguenther@suse.de> 3639 3640 PR lto/87229 3641 * tree.c (free_lang_data_in_one_sizepos): Free non-gimple-val 3642 sizepos values. 3643 36442018-11-19 Eric Botcazou <ebotcazou@adacore.com> 3645 3646 * compare-elim.c (struct comparison): Add not_in_a field. 3647 (is_not): New static function. 3648 (strip_not): Likewise. 3649 (conforming_compare): Handle a NOT in the first operand. 3650 (can_eliminate_compare): Likewise. 3651 (find_comparison_dom_walker::before_dom_children): Likewise. 3652 (try_eliminate_compare): Likewise. 3653 * config/visium/visium.md (negsi2_insn_set_carry): Turn into... 3654 (neg<mode>2_insn_set_carry): ...this and add missing NEG operation. 3655 36562018-11-19 Jonathan Wakely <jwakely@redhat.com> 3657 3658 * doc/extend.texi (Common Type Attributes): Fix typos. 3659 36602018-11-19 Martin Liska <mliska@suse.cz> 3661 3662 PR gcov-profile/88045 3663 * coverage.c (coverage_begin_function): Add assert. 3664 36652018-11-18 Sandra Loosemore <sandra@codesourcery.com> 3666 3667 PR other/40498 3668 * doc/extend.texi (Common Function Attributes): Document that 3669 no_instrument_function applies to -p and -pg, too. 3670 * doc/invoke.texi (Instrumentation Options): Add cross-references 3671 to docs for -p, -pg, and -finstrument-functions. 3672 36732018-11-18 Alan Modra <amodra@gmail.com> 3674 3675 * config/gnu-user.h (GNU_USER_TARGET_CRTI): Define. 3676 (GNU_USER_TARGET_STARTFILE_SPEC): Use it here. 3677 (GNU_USER_TARGET_CRTN): Define. 3678 (GNU_USER_TARGET_ENDFILE_SPEC): Use it here. 3679 (GNU_USER_TARGET_LINK_GCC_C_SEQUENCE_SPEC): Define. 3680 36812018-11-18 Alan Modra <amodra@gmail.com> 3682 3683 * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Delete 3684 !HAVE_LD_PIE variant. 3685 (GNU_USER_TARGET_ENDFILE_SPEC): Likewise. 3686 36872018-11-17 Nick Clifton <nickc@redhat.com> 3688 Sandra Loosemore <sandra@codesourcery.com> 3689 3690 PR driver/31357 3691 * doc/invoke.texi (Overall Options): Document that --help and 3692 --help=<value> options cannot be combined. 3693 36942018-11-17 Jim Wilson <jimw@sifive.com> 3695 3696 * config/riscv/riscv.c (epilogue_cfa_sp_offset): New. 3697 (riscv_restore_reg): If restoring HARD_FRAME_POINTER_REGNUM, and 3698 epilogue_cfa_sp_offset set, then add REG_CFA_DEF_CFA regnote. 3699 (riscv_expand_epilogue): Initialize epilogue_cfa_sp_offset. Set it 3700 to step2 if frame_pointer_needed and step1 is 0. 3701 37022018-11-17 Sandra Loosemore <sandra@codesourcery.com> 3703 3704 PR c++/4225 3705 * doc/cpp.texi (System Headers): Add note about implicit 3706 extern "C" block on targets that define SYSTEM_IMPLICIT_EXTERN_C. 3707 37082018-11-17 Sandra Loosemore <sandra@codesourcery.com> 3709 3710 PR c++/4025 3711 * doc/invoke.texi (C++ Dialect Options): Clarify usage of 3712 -fno-implicit-templates. 3713 37142018-11-17 Jakub Jelinek <jakub@redhat.com> 3715 3716 PR tree-optimization/87546 3717 * tree-vect-patterns.c (vect_look_through_possible_promotion): Add 3718 min_precision variable, initially set it to orig_precision, only does 3719 something if op_type's precision is <= min_precision and update 3720 min_precision whenever calling set_op. 3721 37222018-11-16 Jan Hubicka <hubicka@ucw.cz> 3723 3724 PR ipa/87957 3725 * ipa-devirt.c (warn_odr): Look for main variant to get TYPE_DECL. 3726 37272018-11-16 Sandra Loosemore <sandra@codesourcery.com> 3728 3729 * doc/invoke.texi (Option Summary): Fix whitespace and line 3730 breaks in @gccoptlist environments. 3731 (Warning Options): Likewise. 3732 (Optimize Options): Likewise. 3733 (PowerPC SPE Options): Likewise. 3734 (RS/6000 and PowerPC Options): Likewise. 3735 37362018-11-16 Jeff Law <law@redhat.com> 3737 3738 * config/mn10300/mn10300.md (adddi3_degenerate): Remove bogus 3739 gcc_assert. 3740 37412018-11-16 Sandra Loosemore <sandra@codesourcery.com> 3742 3743 PR middle-end/23197 3744 * doc/invoke.texi (Optimize Options): Update options enabled by 3745 fprofile-generate, -fprofile-use, and -fauto-profile. 3746 37472018-11-16 Jozef Lawrynowicz <jozef.l@mittosystems.com> 3748 3749 PR target/87927 3750 * target-def.h: Initialize TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP. 3751 Add them to the TARGET_ASM_{,UN}ALIGNED_INT_OP structs. 3752 * target.def: Enumerate TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP in 3753 the byte_op hook. 3754 * target.h: Add psi, pdi, pti to struct asm_int_op definition. 3755 * targhooks.c (default_print_patchable_function_entry): Assert 3756 asm_int_op does not return a NULL string. 3757 * varasm.c (integer_asm_op): Return the op for a partial int type 3758 when the requested size does not correspond to an integer type. 3759 * config/msp430/msp430.c: Initialize TARGET_ASM_{,UN}ALIGNED_PSI_OP. 3760 * doc/tm.texi: Regenerate. 3761 37622018-11-12 Jason Merrill <jason@redhat.com> 3763 3764 * gimplify.c (gimplify_case_label_expr): Handle hot/cold attributes. 3765 37662018-11-16 Michael Meissner <meissner@linux.ibm.com> 3767 3768 * config/rs6000/constraints.md (wF constraint): Remove power9 3769 fusion documentation. Just document wF's use for power8 fusion. 3770 * config/rs6000/predicates.md (p9_fusion_reg_operand): Delete. 3771 (fusion_gpr_addis): Delete power9 fusion support. Change power8 3772 fusion support to require the upper 12 bits to be all 0's or all 3773 1's. 3774 (fusion_gpr_mem_load): Add comment. 3775 (fusion_addis_mem_combo_load): Remove power9 fusion support. 3776 (fusion_addis_mem_combo_store): Delete. 3777 (fusion_offsettable_mem_operand): Delete. 3778 * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Do not set 3779 power8 fusion here. 3780 (ISA_3_0_MASKS_SERVER): Delete power9 fusion. 3781 (POWERPC_MASKS): Delete power9 fusion. 3782 * config/rs6000/rs6000-protos.h (emit_fusion_load_store): Delete. 3783 (fusion_p9_p): Delete. 3784 (expand_fusion_p9_load): Delete. 3785 (expand_fusion_p9_store): Delete. 3786 (emit_fusion_p9_load): Delete. 3787 (emit_fusion_p9_store): Delete. 3788 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Delete power9 3789 fusion support. 3790 (rs6000_option_override_internal): Set power8 fusion based on 3791 whether we are tuning for power8. Delete power9 fusion support. 3792 (rs6000_opt_masks): Delete -mpower9-fusion switch. 3793 (emit_fusion_load): Rename emit_fusion_load_store to 3794 emit_fusion_load, and drop fusion store support. Update callers. 3795 (emit_fusion_load_store): Likewise. 3796 (emit_fusion_gpr_load): Likewise. 3797 (fusion_p9_p): Delete. 3798 (expand_fusion_p9_load): Delete. 3799 (expand_fusion_p9_store): Delete. 3800 (emit_fusion_p9_load): Delete. 3801 (emit_fusion_p9_store): Delete. 3802 * config/rs6000/rs6000.md (UNSPEC_FUSION_P9): Delete. 3803 (GPR_FUSION): Delete. 3804 (FPR_FUSION): Delete. 3805 (power9 fusion peephole2s): Delete. 3806 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Delete. 3807 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Delete. 3808 (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_load): Delete. 3809 (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_store): Delete. 3810 (fusion_p9_<mode>_constant): Delete. 3811 * config/rs6000/rs6000.opt (-mpower9-fusion): Delete undocumented 3812 power9 fusion switch. 3813 * doc/md.texi (PowerPC constraints): Update wF constraint 3814 documentation for power8 fusion only. 3815 38162018-11-16 Jakub Jelinek <jakub@redhat.com> 3817 3818 PR rtl-optimization/87475 3819 * cfgrtl.c (patch_jump_insn): Allow redirection failure for 3820 CROSSING_JUMP_P insns. 3821 (cfg_layout_redirect_edge_and_branch): Don't ICE if ret is NULL. 3822 38232018-11-16 Uros Bizjak <ubizjak@gmail.com> 3824 3825 PR target/88051 3826 * config/i386/i386.md (floatunsdidf2): Allow only 64bit AVX512F targets. 3827 * config/i386/sse.md (UNSPEC_MOVDI_TO_SSE): New UNSPEC. 3828 (movdi_to_sse): Rewrite using UNSPEC_MOVDI_TO_SSE unspec. 3829 38302018-11-16 Jakub Jelinek <jakub@redhat.com> 3831 3832 PR middle-end/88032 3833 * optabs.c (expand_binop): For op0_mode use GET_MODE (op0), unless it 3834 is VOIDmode, in which case use int_mode. Similarly for op1_mode. 3835 38362018-11-16 Eric Botcazou <ebotcazou@adacore.com> 3837 3838 * tree-switch-conversion.h (switch_decision_tree::emit_case_nodes): Add 3839 location_t parameter. 3840 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise. 3841 (switch_decision_tree::do_jump_if_equal): Likewise. 3842 * tree-switch-conversion.c (switch_decision_tree::emit): Pass location 3843 of switch statement to emit_case_nodes. 3844 (switch_decision_tree::emit_cmp_and_jump_insns): Add LOC parameter and 3845 set it on the newly built GIMPLE comparison statement. 3846 (switch_decision_tree::do_jump_if_equal): Likewise. 3847 (switch_decision_tree::emit_case_nodes): Add LOC parameter and pass it 3848 in calls to do_jump_if_equal as well as recursive calls. 3849 38502018-11-16 Andi Kleen <ak@linux.intel.com> 3851 3852 * config/i386/i386.md: Allow memory operands to ptwrite. 3853 38542018-11-16 Michael Matz <matz@suse.de> 3855 3856 * system.h (PRsa): New macro. 3857 (SIZE_AMOUNT): Cast number to uint64_t. 3858 * alloc-pool.h (pool_usage::dump): Don't use %zu but PRsa. 3859 (pool_usage::dump_footer): Likewise and also use PRIu64. 3860 * bitmap.h (bitmap_usage::dump): Likewise. 3861 * ggc-common.c (ggc_usage::dump): Likewise. 3862 * ggc-page.c (ggc_print_statistics): Likewise. 3863 * input.c (dump_line_table_statistics): Likewise. 3864 * mem-stats.h (mem_usage::dump): Likewise. 3865 (mem_usage::dump_footer): Likewise. 3866 * rtl.c (dump_rtx_statistics): Likewise. 3867 * tree-cfg.c (dump_cfg_stats): Likewise. 3868 * tree-dfa.c (dump_dfa_stats): Likewise. 3869 * tree-phinodes.c (phinodes_print_statistics): Likewise. 3870 * tree-ssanames (ssanames_print_statistics): Likewise. 3871 * vec.c (vec_usage::dump): Likewise. 3872 (vec_usage::dump_footer): Likewise. 3873 38742018-11-16 Richard Biener <rguenther@suse.de> 3875 3876 PR tree-optimization/88011 3877 * tree-vrp.c (extract_range_from_binary_expr): Fix error in 3878 replacing set_value_range_to_undefined and 3879 set_value_range_to_varying with method calls. 3880 38812018-11-16 Ilya Leoshkevich <iii@linux.ibm.com> 3882 3883 * config/s390/s390.md 3884 (*r<noxa>sbg_<mode>_srl_bitmask): Do not delegate arithmetic to 3885 assembler. 3886 (*r<noxa>sbg_<mode>_sll): Likewise. 3887 (*r<noxa>sbg_<mode>_srl): Likewise. 3888 (*r<noxa>sbg_sidi_srl): New pattern. 3889 38902018-11-16 Jerome Lambourg <lambourg@adacore.com> 3891 3892 * config/vxworks.h (TARGET_VXWORKS_HAVE_CTORS_DTORS): New macro. 3893 Default to TARGET_VXWORKS_RTP. 3894 (SUPPORTS_INIT_PRIORITY): Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead 3895 of TARGET_VXWORKS_RTP. 3896 * config/vxworks.c: Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead 3897 of TARGET_VXWORKS_RTP to set targetm.have_ctors_dtors. 3898 38992018-11-15 Sandra Loosemore <sandra@codesourcery.com> 3900 3901 PR c++/25759 3902 * doc/extend.texi (Common Type Attributes): Make it explicit 3903 that attribute "packed" can apply to C++ classes. 3904 39052018-11-15 Martin Sebor <msebor@redhat.com> 3906 3907 PR c++/87541 3908 PR c++/87542 3909 * tree.c (type_argument_type): New function. 3910 * tree.h (type_argument_type): Declare it. 3911 * gcc/doc/extend.texi (alloc_align): Update and clarify. 3912 (alloc_size, nonnull, sentinel): Same. 3913 39142018-11-15 Andrew Stubbs <ams@codesourcery.com> 3915 Kwok Cheung Yeung <kcy@codesourcery.com> 3916 3917 * tree-vect-stmts.c (vectorizable_store): Don't ICE when 3918 int_mode_for_size fails. 3919 (vectorizable_load): Likewise. 3920 39212018-11-15 David Malcolm <dmalcolm@redhat.com> 3922 3923 * doc/ux.texi (Group logically-related diagnostics): Move 3924 discussion of auto_diagnostic_group into this new subsection. 3925 Give an example of where this grouping is used. 3926 39272018-11-15 David Malcolm <dmalcolm@redhat.com> 3928 3929 PR other/19165 3930 * Makefile.in (OBJS): Move json.o to... 3931 (OBJS-libcommon): ...here and add diagnostic-format-json.o. 3932 * common.opt (fdiagnostics-format=): New option. 3933 (diagnostics_output_format): New enum. 3934 * diagnostic-format-json.cc: New file. 3935 * diagnostic.c (default_diagnostic_final_cb): New function, taken 3936 from start of diagnostic_finish. 3937 (diagnostic_initialize): Initialize final_cb to 3938 default_diagnostic_final_cb. 3939 (diagnostic_finish): Move "being treated as errors" messages to 3940 default_diagnostic_final_cb. Call any final_cb. 3941 (default_diagnostic_finalizer): Add diagnostic_t param. 3942 (diagnostic_report_diagnostic): Pass "orig_diag_kind" to 3943 diagnostic_finalizer callback. 3944 * diagnostic.h (enum diagnostics_output_format): New enum. 3945 (diagnostic_finalizer_fn): Reimplement, adding diagnostic_t param. 3946 (struct diagnostic_context): Add "final_cb". 3947 (default_diagnostic_finalizer): Add diagnostic_t param. 3948 (diagnostic_output_format_init): New decl. 3949 * doc/invoke.texi (-fdiagnostics-format): New option. 3950 * dwarf2out.c (gen_producer_string): Ignore 3951 OPT_fdiagnostics_format_. 3952 * gcc.c (driver_handle_option): Handle OPT_fdiagnostics_format_. 3953 * lto-wrapper.c (append_diag_options): Ignore it. 3954 * opts.c (common_handle_option): Handle it. 3955 39562018-11-15 David Malcolm <dmalcolm@redhat.com> 3957 3958 PR tree-optimization/88015 3959 * graphite-isl-ast-to-gimple.c 3960 (translate_isl_ast_to_gimple::scop_to_isl_ast): Add missing check 3961 for dump_enabled_p. 3962 * graphite-sese-to-poly.c (build_poly_scop): Likewise. 3963 39642018-11-15 Richard Biener <rguenther@suse.de> 3965 3966 PR middle-end/88029 3967 * gimple.c (gimple_call_flags): Union flags from decl, type 3968 and call fntype. 3969 * trans-mem.c (is_tm_pure_call): Simplify. 3970 39712018-11-15 Richard Biener <rguenther@suse.de> 3972 3973 PR tree-optimization/88031 3974 * tree-vect-loop.c (vectorizable_reduction): Move check 3975 for multiple types earlier so we get the expected dump. 3976 Simplify calls to vectorizable_condition. 3977 * tree-vect-stmts.h (vectorizable_condition): Update prototype. 3978 * tree-vect-stmts.c (vectorizable_condition): Instead of 3979 reduc_def and reduc_index take just a flag. Simplify 3980 code-generation now that we can rely on the defs being set up. 3981 (vectorizable_comparison): Remove unused argument. 3982 39832018-11-15 Richard Biener <rguenther@suse.de> 3984 3985 PR tree-optimization/88030 3986 * tree-complex.c (need_eh_cleanup): New global. 3987 (update_complex_assignment): Mark blocks that need EH update. 3988 (expand_complex_comparison): Likewise. 3989 (tree_lower_complex): Allocate and deallocate need_eh_cleanup, 3990 perform EH cleanup and schedule CFG cleanup if that did anything. 3991 39922018-11-15 Jakub Jelinek <jakub@redhat.com> 3993 3994 PR rtl-optimization/88018 3995 * cfgrtl.c (fixup_abnormal_edges): Guard moving insns to fallthru edge 3996 on the presence of fallthru edge, rather than if it is a USE or not. 3997 39982018-11-15 Richard Biener <rguenther@suse.de> 3999 4000 PR middle-end/87917 4001 * tree-data-ref.c (analyze_miv_subscript): Guard calls to 4002 analyze_subscript_affine_affine properly. 4003 40042018-11-15 Xianmiao Qu <xianmiao_qu@c-sky.com> 4005 4006 * config/csky/csky-linux-elf.h (LINUX_DYNAMIC_LINKER): Remove. 4007 (GLIBC_DYNAMIC_LINKER): Define. 4008 (LINUX_TARGET_LINK_SPEC): Update the dynamic linker's name. 4009 40102018-11-15 Bin Cheng <bin.cheng@linux.alibaba.com> 4011 4012 PR tree-optimization/84648 4013 * tree-ssa-loop-niter.c (adjust_cond_for_loop_until_wrap): New. 4014 (number_of_iterations_cond): Adjust exit cond for loop-until-wrap case 4015 by calling adjust_cond_for_loop_until_wrap. 4016 40172018-11-15 Sandra Loosemore <sandra@codesourcery.com> 4018 4019 PR other/56334 4020 * doc/extend.texi (Common Function Attributes): Clarify linker 4021 restrictions on "aligned" attribute. 4022 (Common Variable Attributes): Likewise. Mention that linker 4023 restrictions don't apply to stack-allocated variables. 4024 40252018-11-15 Joern Rennecke <joern.rennecke@riscy-ip.com> 4026 4027 * gensupport.c (add_predicate_code): Properly handle ZERO_EXTRACT 4028 as an lvalue. 4029 40302018-11-14 Sandra Loosemore <sandra@codesourcery.com> 4031 4032 PR lto/55102 4033 PR lto/56700 4034 * doc/invoke.texi (Optimize Options): Remove bad example about 4035 interaction between -flto and -O. Replace it with a note that 4036 you need to compile with -O and not just link. Copy-edit -flto 4037 discussion to reduce verbiage and improve flow. 4038 40392018-11-14 Sandra Loosemore <sandra@codesourcery.com> 4040 4041 PR middle-end/59658 4042 * doc/invoke.texi (Optimize Options): Clarify that -O0 and -Og 4043 also suppress many optimizations. Alphabetize option lists for 4044 -O1, -O2, and -Os. Add list of options disabled with -Og, and 4045 correct documentation for those options to say that. 4046 * opts.c (default_options_table): Sort table by level and option 4047 name, to make it easier to correlate to the manual. 4048 40492018-11-14 Uros Bizjak <ubizjak@gmail.com> 4050 4051 * config/i386/i386.c (ix86_print_operand_address_as): Simplify 4052 printing of the names of segment registers. 4053 40542018-11-14 Nathan Sidwell <nathan@acm.org> 4055 4056 PR debug/88006 4057 PR debug/87462 4058 * dwarf2out.c (dwarf2out_finish): Apply resolve_addr to comdat 4059 type list. 4060 40612018-11-14 David Malcolm <dmalcolm@redhat.com> 4062 4063 * Makefile.in (CFLAGS-optinfo-emit-json.o): Add $(ZLIBINC). 4064 40652018-11-14 Jakub Jelinek <jakub@redhat.com> 4066 4067 P1236R1 - Signed integers are two's complement 4068 * doc/invoke.texi (Wshift-overflow): Adjust documentation for c++2a. 4069 4070 PR bootstrap/86739 4071 * hash-map.h (hash_map::iterator::reference_pair): New class. 4072 (hash_map::iterator::operator*): Return it rather than std::pair. 4073 40742018-11-14 Jeff Law <law@redhat.com> 4075 4076 * optabs.c (expand_binop): Pass INT_MODE to operand_subword_force 4077 iff the operand is a constant. 4078 40792018-11-14 Aldy Hernandez <aldyh@redhat.com> 4080 4081 * gimple-ssa-evrp-analyze.c 4082 (evrp_range_analyzer::record_ranges_from_incoming_edge): Rename 4083 ignore_equivs_equal_p to equal_p. 4084 * ipa-cp.c (meet_with_1): Use equal_p instead of 4085 ignore_equivs_equal_p. 4086 * ipa-prop.c (ipa_vr_ggc_hash_traits::equal): Same. 4087 * tree-vrp.c (value_range::ignore_equivs_equal_p): Remove. 4088 (value_range::operator==): Remove. 4089 (value_range::operator!=): Remove. 4090 (vrp_prop::visit_stmt): Use equal_p. 4091 * tree-vrp.h (value_range): Remove operator==, operator!=, 4092 ignore_equivs_equal_p. 4093 * vr-values.c (update_value_range): Use equal_p. 4094 40952018-11-14 Michael Matz <matz@suse.de> 4096 4097 PR middle-end/86575 4098 * gimplify.c (collect_fallthrough_labels): Add new argument, 4099 return location via that, don't modify statements. 4100 (warn_implicit_fallthrough_r): Adjust call, don't use 4101 statement location directly. 4102 41032018-11-14 Richard Biener <rguenther@suse.de> 4104 4105 PR middle-end/87985 4106 * tree-data-ref.c (split_constant_offset): Add wrapper 4107 allocating a cache hash-map. 4108 (split_constant_offset_1): Cache results of expanding 4109 expressions from SSA def stmts. 4110 41112018-11-14 Richard Biener <rguenther@suse.de> 4112 4113 PR middle-end/88021 4114 * tree-data-ref.c (lambda_matrix_row_add): Change const1 argument 4115 to lambda_int. 4116 (lambda_vector_mult_const): Likewise. 4117 (lambda_matrix_right_hermite): Use lambda_int temporaries. 4118 41192018-11-14 Wilco Dijkstra <wdijkstr@arm.com> 4120 Jackson Woodruff <jackson.woodruff@arm.com> 4121 4122 PR 71026/tree-optimization 4123 * match.pd: Simplify floating point comparisons. 4124 41252018-11-14 Jakub Jelinek <jakub@redhat.com> 4126 4127 PR rtl-optimization/87817 4128 * config/i386/i386.c (ix86_fold_builtin): For _bzhi_u{32,64} if 4129 last argument has low 8 bits clear, fold to 0. 4130 41312018-11-14 Iain Sandoe <iain@sandoe.co.uk> 4132 4133 * tree-vect-data-refs.c (vect_can_force_dr_alignment_p): Cast 4134 MAX_OFILE_ALIGNMENT as needed. 4135 41362018-11-14 Richard Biener <rguenther@suse.de> 4137 4138 PR tree-optimization/88019 4139 * graphite-isl-ast-to-gimple.c (ternary_op_to_tree): Rewrite 4140 COND_EXPR arguments to non-trapping overflow. 4141 41422018-11-14 Claudiu Zissulescu <claziss@synopsys.com> 4143 4144 * config/arc/arc.h (reg_class): Reorder registers classes, remove 4145 unused register classes. 4146 (REG_CLASS_NAMES): Likewise. 4147 (REG_CLASS_CONTENTS): Likewise. 4148 (FIXED_REGISTERS): Make lp_count fixed. 4149 (BASE_REG_CLASS): Remove ACC16_BASE_REGS reference. 4150 (PROGRAM_COUNTER_REGNO): Remove. 4151 * config/arc/arc.c (arc_conditional_register_usage): Remove unused 4152 register classes, use constants for register numbers, remove 4153 reg_class_contents references. 4154 (arc_process_double_reg_moves): Add asserts. 4155 (arc_secondary_reload): Remove LPCOUNT_REG reference, use 4156 lra_in_progress predicate. 4157 (arc_init_reg_tables): Remove unused register classes. 4158 (arc_register_move_cost): Likewise. 4159 (arc_preferred_reload_class): Likewise. 4160 (hwloop_optimize): Update rtx patterns involving lp_count register. 4161 (arc_return_address_register): Rename ILINK1, INLINK2 regnums macros. 4162 * config/arc/constraints.md ("c"): Choose between GENERAL_REGS and 4163 CHEAP_CORE_REGS. Former one will be used for LRA. 4164 ("Rac"): Choose between GENERAL_REGS and ALL_CORE_REGS. Former 4165 one will be used for LRA. 4166 ("w"): Choose between GENERAL_REGS and WRITABLE_CORE_REGS. Former 4167 one will be used for LRA. 4168 ("W"): Choose between GENERAL_REGS and MPY_WRITABLE_CORE_REGS. 4169 Former one will be used for LRA. 4170 ("f"): Delete constraint. 4171 ("k"): Likewise. 4172 ("e"): Likewise. 4173 ("l"): Change it from register constraint to constraint. 4174 * config/arc/arc.md (movqi_insn): Remove unsed lp_count constraints. 4175 (movhi_insn): Likewise. 4176 (movsi_insn): Update pattern. 4177 (arc_lp): Likewise. 4178 (dbnz): Likewise. 4179 (stack_tie): Remove 'b' constraint letter. 4180 (R4_REG): Define. 4181 (R9_REG, R15_REG, R16_REG, R25_REG): Likewise. 4182 (R32_REG, R40_REG, R41_REG, R42_REG, R43_REG, R44_REG): Likewise. 4183 (R57_REG, R59_REG, PCL_REG): Likewise. 4184 (ILINK1_REGNUM): Renamed to ILINK1_REG. 4185 (ILINK2_REGNUM): Renamed to ILINK2_REG. 4186 (Rgp): Remove. 4187 (SP_REGS): Likewise. 4188 (Rcw): Remove unused reg classes. 4189 * config/arc/predicates.md (dest_reg_operand): Just default on 4190 register_operand predicate. 4191 (mpy_dest_reg_operand): Likewise. 4192 (move_dest_operand): Use macros instead of constants. 4193 * config/arc/arc.opt (mlra): Switch to lra as default. 4194 41952018-11-14 Richard Biener <rguenther@suse.de> 4196 4197 PR tree-optimization/87974 4198 * tree-vect-loop.c (vectorizable_reduction): When computing 4199 the vectorized reduction PHI vector type ignore constant 4200 and external defs. 4201 42022018-11-14 Jakub Jelinek <jakub@redhat.com> 4203 4204 PR tree-optimization/87977 4205 * tree-ssa-math-opts.c (optimize_recip_sqrt): Don't reuse division 4206 stmt, build a new one and replace the old one with it. Formatting fix. 4207 Call release_ssa_name (x) if !has_other_use and !delete_div. 4208 (pass_cse_reciprocals::execute): Before calling optimize_recip_sqrt 4209 verify lhs of stmt is still def. 4210 42112018-11-13 Peter Bergner <bergner@linux.ibm.com> 4212 4213 PR rtl-optimization/87507 4214 * lower-subreg.c (operand_for_swap_move_operator): New function. 4215 (simple_move): Strip simple operators. 4216 (find_pseudo_copy): Likewise. 4217 (resolve_operand_for_swap_move_operator): New function. 4218 (resolve_simple_move): Strip simple operators and swap operands. 4219 42202018-11-13 Jakub Jelinek <jakub@redhat.com> 4221 4222 PR tree-optimization/87898 4223 * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Formatting fix. 4224 (ipa_simd_modify_function_body): Remove debug stmts where the first 4225 argument was changed into a non-decl. 4226 42272018-11-13 Peter Bergner <bergner@linux.ibm.com> 4228 4229 PR rtl-optimization/87899 4230 * lra-lives.c (start_living): Update white space in comment. 4231 (enum point_type): New. 4232 (sparseset_contains_pseudos_p): New function. 4233 (update_pseudo_point): Likewise. 4234 (make_hard_regno_live): Use HARD_REGISTER_NUM_P macro. 4235 (make_hard_regno_dead): Likewise. Remove ignore_reg_for_conflicts 4236 handling. Move early exit after adding conflicts. 4237 (mark_pseudo_live): Use HARD_REGISTER_NUM_P macro. Add early exit 4238 if regno is already live. Remove all handling of program points. 4239 (mark_pseudo_dead): Use HARD_REGISTER_NUM_P macro. Add early exit 4240 after adding conflicts. Remove all handling of program points and 4241 ignore_reg_for_conflicts. 4242 (mark_regno_live): Use HARD_REGISTER_NUM_P macro. Remove return value 4243 and do not guard call to mark_pseudo_live. 4244 (mark_regno_dead): Use HARD_REGISTER_NUM_P macro. Remove return value 4245 and do not guard call to mark_pseudo_dead. 4246 (check_pseudos_live_through_calls): Use HARD_REGISTER_NUM_P macro. 4247 (process_bb_lives): Use HARD_REGISTER_NUM_P and HARD_REGISTER_P macros. 4248 Use new function update_pseudo_point. Handle register copies by 4249 removing the source register from the live set. Handle INOUT operands. 4250 Update to the next program point using the unused_set, dead_set and 4251 start_dying sets. 4252 (lra_create_live_ranges_1): Use HARD_REGISTER_NUM_P macro. 4253 42542018-11-13 David Malcolm <dmalcolm@redhat.com> 4255 4256 * builtins.c: Replace "source_location" with "location_t". 4257 * diagnostic-show-locus.c: Likewise. 4258 * diagnostic.c: Likewise. 4259 * dumpfile.c: Likewise. 4260 * gcc-rich-location.h: Likewise. 4261 * genmatch.c: Likewise. 4262 * gimple.h: Likewise. 4263 * gimplify.c: Likewise. 4264 * input.c: Likewise. 4265 * input.h: Likewise. Eliminate the typedef. 4266 * omp-expand.c: Likewise. 4267 * selftest.h: Likewise. 4268 * substring-locations.h (get_source_location_for_substring): 4269 Rename to... 4270 (get_location_within_string): ...this. 4271 * tree-cfg.c: Replace "source_location" with "location_t". 4272 * tree-cfgcleanup.c: Likewise. 4273 * tree-diagnostic.c: Likewise. 4274 * tree-into-ssa.c: Likewise. 4275 * tree-outof-ssa.c: Likewise. 4276 * tree-parloops.c: Likewise. 4277 * tree-phinodes.c: Likewise. 4278 * tree-phinodes.h: Likewise. 4279 * tree-ssa-loop-ivopts.c: Likewise. 4280 * tree-ssa-loop-manip.c: Likewise. 4281 * tree-ssa-phiopt.c: Likewise. 4282 * tree-ssa-phiprop.c: Likewise. 4283 * tree-ssa-threadupdate.c: Likewise. 4284 * tree-ssa.c: Likewise. 4285 * tree-ssa.h: Likewise. 4286 * tree-vect-loop-manip.c: Likewise. 4287 42882018-11-13 Michael Ploujnikov <michael.ploujnikov@oracle.com> 4289 4290 * doc/extend.texi: Fix typo in the weakref description. 4291 42922018-11-13 Richard Biener <rguenther@suse.de> 4293 4294 PR tree-optimization/86991 4295 * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction 4296 group building until we have successfully detected the SLP 4297 reduction. 4298 (vect_is_simple_reduction): Remove fixup code here. 4299 43002018-11-13 David Malcolm <dmalcolm@redhat.com> 4301 4302 * dumpfile.c (VERIFY_DUMP_ENABLED_P): New macro. 4303 (dump_gimple_stmt): Use it. 4304 (dump_gimple_stmt_loc): Likewise. 4305 (dump_gimple_expr): Likewise. 4306 (dump_gimple_expr_loc): Likewise. 4307 (dump_generic_expr): Likewise. 4308 (dump_generic_expr_loc): Likewise. 4309 (dump_printf): Likewise. 4310 (dump_printf_loc): Likewise. 4311 (dump_dec): Likewise. 4312 (dump_dec): Likewise. 4313 (dump_hex): Likewise. 4314 (dump_symtab_node): Likewise. 4315 * gimple-loop-interchange.cc (tree_loop_interchange::interchange): 4316 Guard dump call with dump_enabled_p. 4317 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise. 4318 * graphite-optimize-isl.c (optimize_isl): Likewise. 4319 * graphite.c (graphite_transform_loops): Likewise. 4320 * tree-loop-distribution.c (pass_loop_distribution::execute): Likewise. 4321 * tree-parloops.c (parallelize_loops): Likewise. 4322 * tree-ssa-loop-niter.c (number_of_iterations_exit): Likewise. 4323 * tree-vect-data-refs.c (vect_analyze_group_access_1): Likewise. 4324 (vect_prune_runtime_alias_test_list): Likewise. 4325 * tree-vect-loop.c (vect_update_vf_for_slp): Likewise. 4326 (vect_estimate_min_profitable_iters): Likewise. 4327 * tree-vect-slp.c (vect_record_max_nunits): Likewise. 4328 (vect_build_slp_tree_2): Likewise. 4329 (vect_supported_load_permutation_p): Likewise. 4330 (vect_slp_analyze_operations): Likewise. 4331 (vect_slp_analyze_bb_1): Likewise. 4332 (vect_slp_bb): Likewise. 4333 * tree-vect-stmts.c (vect_analyze_stmt): Likewise. 4334 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. 4335 (pass_slp_vectorize::execute): Likewise. 4336 (increase_alignment): Likewise. 4337 43382018-11-13 David Malcolm <dmalcolm@redhat.com> 4339 4340 PR ipa/87955 4341 * ipa-inline.c (report_inline_failed_reason): Guard calls to 4342 cl_target_option_print_diff and cl_optimization_print_diff with 4343 if (dump_file). 4344 43452018-11-13 David Malcolm <dmalcolm@redhat.com> 4346 4347 * doc/invoke.texi (-fsave-optimization-record): Note that the 4348 output is compressed. 4349 * optinfo-emit-json.cc: Include <zlib.h>. 4350 (optrecord_json_writer::write): Compress the output. 4351 43522018-11-13 Aldy Hernandez <aldyh@redhat.com> 4353 4354 * tree-vrp.c (value_range_base::dump): Dump type. 4355 Do not use INF nomenclature for 1-bit types. 4356 (dump_value_range): Group all variants to common dumping code. 4357 (debug): New overloaded functions for value_ranges. 4358 (value_range_base::dump): Remove no argument version. 4359 (value_range::dump): Same. 4360 43612018-11-13 Richard Biener <rguenther@suse.de> 4362 4363 PR tree-optimization/87931 4364 * tree-vect-loop.c (vect_is_simple_reduction): Restrict 4365 nested cycles we support to latch computations vectorizable_reduction 4366 handles. 4367 43682018-11-13 Martin Liska <mliska@suse.cz> 4369 4370 PR tree-optimization/87885 4371 * cfghooks.c (account_profile_record): Rename 4372 to ... 4373 (profile_record_check_consistency): ... this. 4374 Calculate missing num_mismatched_freq_in. 4375 (profile_record_account_profile): New function 4376 that calculates time and size of a function. 4377 * cfghooks.h (struct profile_record): Remove 4378 all tuples. 4379 (struct cfg_hooks): Remove after_pass flag. 4380 (account_profile_record): Rename to ... 4381 (profile_record_check_consistency): ... this. 4382 (profile_record_account_profile): New. 4383 * cfgrtl.c (rtl_account_profile_record): Remove 4384 after_pass flag. 4385 * passes.c (check_profile_consistency): Do only 4386 checking. 4387 (account_profile): Calculate size and time of 4388 function only. 4389 (pass_manager::dump_profile_report): Reformat 4390 output. 4391 (execute_one_ipa_transform_pass): Call 4392 consistency check before clean upand call account_profile 4393 after a clean up is done. 4394 (execute_one_pass): Call check_profile_consistency and 4395 account_profile instead of using after_pass flag.. 4396 * tree-cfg.c (gimple_account_profile_record): Likewise. 4397 43982018-11-13 Martin Liska <mliska@suse.cz> 4399 4400 PR sanitizer/87930 4401 * config/i386/i386.c (ix86_option_override_internal): Error 4402 about usage -mabi=ms and -fsanitize={,kernel-}address. 4403 44042018-11-13 Andre Vieira <andre.simoesdiasvieira@arm.com> 4405 4406 * config/aarch64/aarch64.c 4407 (aarch64_vectorize_preferred_vector_alignment): Change return type to 4408 poly_uint64. 4409 (aarch64_simd_vector_alignment_reachable): Adapt to preferred vector 4410 alignment being a poly int. 4411 * doc/tm.texi (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Change 4412 return type to poly_uint64. 4413 * target.def (default_preferred_vector_alignment): Likewise. 4414 * targhooks.c (default_preferred_vector_alignment): Likewise. 4415 * targhooks.h (default_preferred_vector_alignment): Likewise. 4416 * tree-vect-data-refs.c (vect_calculate_target_alignment): Likewise. 4417 (vect_compute_data_ref_alignment): Adapt to vector alignment being a 4418 poly int. 4419 (vect_update_misalignment_for_peel): Likewise. 4420 (vect_enhance_data_refs_alignment): Likewise. 4421 (vect_find_same_alignment_drs): Likewise. 4422 (vect_duplicate_ssa_name_ptr_info): Likewise. 4423 (vect_setup_realignment): Likewise. 4424 (vect_can_force_dr_alignment_p): Change alignment parameter type to 4425 poly_uint64. 4426 * tree-vect-loop-manip.c (get_misalign_in_elems): Learn to construct a 4427 mask with a compile time variable vector alignment. 4428 (vect_gen_prolog_loop_niters): Adapt to vector alignment being a poly 4429 int. 4430 (vect_do_peeling): Exit early if vector alignment is not constant. 4431 * tree-vect-stmts.c (ensure_base_align): Adapt to vector alignment being 4432 a poly int. 4433 (vectorizable_store): Likewise. 4434 (vectorizable_load): Likweise. 4435 * tree-vectorizer.h (struct dr_vec_info): Make target_alignment field a 4436 poly_uint64. 4437 (vect_known_alignment_in_bytes): Adapt to vector alignment being a 4438 poly int. 4439 (vect_can_force_dr_alignment_p): Change alignment parameter type to 4440 poly_uint64. 4441 44422018-11-13 Richard Biener <rguenther@suse.de> 4443 4444 PR tree-optimization/87962 4445 * tree-vect-loop.c (vect_is_simple_reduction): More reliably 4446 detect outer reduction for disqualifying in-loop uses. 4447 44482018-11-13 Richard Biener <rguenther@suse.de> 4449 4450 PR tree-optimization/87967 4451 * tree-vect-loop.c (vect_transform_loop): Also copy PHIs 4452 for constants for the scalar loop. 4453 44542018-11-13 Alan Modra <amodra@gmail.com> 4455 4456 * config/rs6000/linux64.h (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Exclude 4457 integer constants when -mcmodel=medium. 4458 44592018-11-13 Claudiu Zissulescu <claziss@synopsys.com> 4460 4461 * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty. 4462 * config/arc/elf.h (PROFILE_HOOK): Define. 4463 * config/arc/linux.h (PROFILE_HOOK): Likewise. 4464 44652018-11-13 Claudiu Zissulescu <claziss@synopsys.com> 4466 4467 * config/arc/arc.c (hwloop_optimize): Bailout when detecting a 4468 jump table data in the text section. 4469 44702018-11-13 Claudiu Zissulescu <claziss@synopsys.com> 4471 4472 * config/arc/arc.c (arc_eh_return_address_location): Repurpose it 4473 to fit the eh_return pattern. 4474 * config/arc/arc.md (eh_return): Define. 4475 (VUNSPEC_ARC_EH_RETURN): Likewise. 4476 * config/arc/arc-protos.h (arc_eh_return_address_location): Match 4477 new implementation. 4478 * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it. 4479 4480 44812018-11-13 Claudiu Zissulescu <claziss@synopsys.com> 4482 4483 * common/config/arc/arc-common.c (arc_option_optimization_table): 4484 Millicode optimization is default on for size optimizations. 4485 * config/arc/arc-protos.h (arc_check_multi): New function. 4486 * config/arc/arc.c (RTX_OK_FOR_OFFSET_P): Rearange. 4487 (ENTER_LEAVE_START_REG): Define. 4488 (ENTER_LEAVE_END_REG): Likewise. 4489 (arc_override_options): Disable millicode when long calls option 4490 is on. 4491 (arc_frame_info): Change it from int to bool. 4492 (arc_compute_frame_size): Clean up. 4493 (arc_save_restore): Remove. 4494 (frame_save_reg): New function. 4495 (frame_restore_reg): Likewise. 4496 (arc_enter_leave_p): Likewise. 4497 (arc_save_callee_saves): Likewise. 4498 (arc_restore_callee_saves): Likewise. 4499 (arc_save_callee_enter): Likewise. 4500 (arc_restore_callee_leave): Likewise. 4501 (arc_save_callee_milli): Likewise. 4502 (arc_restore_callee_milli): Likewise. 4503 (arc_expand_prologue): Reimplement to emit enter/leave 4504 instructions. 4505 (arc_expand_epilogue): Likewise. 4506 (arc_check_multi): New function. 4507 * config/arc/arc.md (push_multi_fp): New pattern. 4508 (push_multi_fp_blink): Likewise. 4509 (pop_multi_fp): Likewise. 4510 (pop_multi_fp_blink): Likewise. 4511 (pop_multi_fp_ret): Likewise. 4512 (pop_multi_fp_blink_ret): Likewise. 4513 * config/arc/arc.opt (mmillicode): Update option. 4514 (mcode-density-frame): New option. 4515 * config/arc/predicates.md (push_multi_operand): New predicate. 4516 (pop_multi_operand): Likewise. 4517 * doc/invoke.texi (ARC): Update ARC options information. 4518 4519 45202018-11-13 Claudiu Zissulescu <claziss@synopsys.com> 4521 4522 * config/arc/arc-protos.h (gen_operands_ldd_std): Add. 4523 * config/arc/arc.c (operands_ok_ldd_std): New function. 4524 (mem_ok_for_ldd_std): Likewise. 4525 (gen_operands_ldd_std): Likewise. 4526 * config/arc/arc.md: Add peephole2 rules for std/ldd. 4527 45282018-11-13 Eric Botcazou <ebotcazou@adacore.com> 4529 4530 * toplev.c (output_stack_usage): Turn test on flag_stack_usage into 4531 test on stack_usage_file. 4532 (lang_dependent_init): Do not open the .su file if generating LTO. 4533 45342018-11-13 Jakub Jelinek <jakub@redhat.com> 4535 4536 PR rtl-optimization/87918 4537 * simplify-rtx.c (simplify_merge_mask): For COMPARISON_P, use 4538 simplify_gen_relational rather than simplify_gen_binary. 4539 45402018-11-13 Richard Biener <rguenther@suse.de> 4541 4542 * tree-ssanames.h (set_range_info): Use value_range_base. 4543 (get_range_info): Likewise. 4544 * tree-ssanames.c (set_range_info): Likewise. 4545 (get_range_info): Likewise. 4546 * tree-vrp.c (value_range_base::union_helper): Split 4547 out common parts of value_range[_base]::union_. 4548 (value_range_base::union_): Update. 4549 (value_range::union_): Likewise. 4550 (determine_value_range_1): Use value_range_base. 4551 (determine_value_range): Likewise. 4552 * tree-vrp.h (value_range_base::union_helper): Move ... 4553 (value_range::union_helper): ... from here. 4554 45552018-11-13 Alan Modra <amodra@gmail.com> 4556 4557 * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Negate 4558 offset for PRE_DEC. 4559 (rs6000_secondary_reload_gpr): Don't call find_replacement. 4560 45612018-11-13 Sandra Loosemore <sandra@codesourcery.com> 4562 4563 PR middle-end/59634 4564 * doc/invoke.texi (Optimize Options): Clarify that the 4565 l1-cache-line-size, l1-cache-size, and l2-cache-size parameters 4566 apply to data cache size. 4567 45682018-11-13 Alan Modra <amodra@gmail.com> 4569 4570 * config/rs6000/predicates.md (logical_const_operand), 4571 (logical_operand): Correct comment. 4572 * config/rs6000/rs6000.c (rs6000_emit_prologue): Comment fix. 4573 45742018-11-13 Alan Modra <amodra@gmail.com> 4575 4576 * config/rs6000/rs6000.md (addsi3_high): Prefix with '*'. 4577 45782018-11-13 Alan Modra <amodra@gmail.com> 4579 4580 * gcc/doc/md.texi (Machine Constraints): Correct rs6000 h constraint 4581 description. 4582 * config/rs6000/rs6000.md (movsi_internal1): Delete MT%0 case 4583 covered by alternative. 4584 (movcc_internal1): Ignore h for register preference. 4585 (mov<mode>_hardfloat64): Likewise. 4586 (mov<mode>_softfloat): Ignore c, l, h for register preference. 4587 45882018-11-12 Sandra Loosemore <sandra@codesourcery.com> 4589 4590 PR preprocessor/47823 4591 * doc/cpp.texi (Alternatives to Wrapper #ifndef): Move #pragma once 4592 documentation to... 4593 (Pragmas): ...here. 4594 * doc/extend.texi (Pragmas): Note additional pragmas documented 4595 in the CPP manual. 4596 45972018-11-13 Kugan Vivekanandarajah <kuganv@linaro.org> 4598 4599 PR middle-end/86677 4600 PR middle-end/87528 4601 * tree-scalar-evolution.c (expression_expensive_p): Make 4602 BUILTIN POPCOUNT as expensive when backend does not define it. 4603 46042018-11-12 Fredrik Noring <noring@nocrew.org> 4605 4606 * config.gcc: Update with-llsc defaults for MIPS r5900. 4607 46082018-11-12 Martin Liska <mliska@suse.cz> 4609 4610 PR gcov-profile/87442 4611 * common.opt: Add -fprofile-filter-files and -fprofile-exclude-files 4612 options. 4613 * doc/invoke.texi: Document them. 4614 * tree-profile.c (parse_profile_filter): New. 4615 (parse_profile_file_filtering): Likewise. 4616 (release_profile_file_filtering): Likewise. 4617 (include_source_file_for_profile): Likewise. 4618 (tree_profiling): Filter source files based on the 4619 newly added options. 4620 46212018-11-12 Martin Liska <mliska@suse.cz> 4622 4623 PR target/87903 4624 * doc/extend.texi: Add missing values for __builtin_cpu_is and 4625 __builtin_cpu_supports for x86 target. 4626 46272018-11-12 Sandra Loosemore <sandra@codesourcery.com> 4628 4629 PR middle-end/21110 4630 * doc/rtl.texi (Constants): Clarify that mode of "high" doesn't 4631 have to be Pmode. 4632 (Arithmetic): Likewise for "lo_sum". 4633 46342018-11-12 Renlin Li <renlin.li@arm.com> 4635 4636 PR target/87815 4637 * dse.c (get_stored_val): Add check for compile-time constantness 4638 of gap. 4639 46402018-11-12 Sudakshina Das <sudi.das@arm.com> 4641 4642 * config/arm/arm-cpus.in (armv8_5, sb, predres): New features. 4643 (ARMv8_5a): New fgroup. 4644 (armv8.5-a): New arch. 4645 (armv8-a, armv8.1-a, armv8.2-a, armv8.3-a, armv8.4-a): New options 4646 sb and predres. 4647 * config/arm/arm-tables.opt: Regenerate. 4648 * config/arm/t-aprofile: Add matching rules for -march=armv8.5-a 4649 * config/arm/t-arm-elf (all_v8_archs): Add armv8.5-a. 4650 * config/arm/t-multilib (v8_5_a_simd_variants): New variable. 4651 Add matching rules for -march=armv8.5-a and extensions. 4652 * doc/invoke.texi (ARM options): Document -march=armv8.5-a. 4653 Add sb and predres to all armv8-a except armv8.5-a. 4654 46552018-11-12 Richard Biener <rguenther@suse.de> 4656 4657 * tree-vrp.h (value_range[_base]::set): Make public. Provide 4658 overload for single value. 4659 (value_range[_base]::set_nonnull): New. 4660 (value_range[_base]::set_null): Likewise. 4661 (value_range): Document bitmap copying behavior, mark 4662 copy constructor and assignment operator deleted. 4663 (value_range::move): New. 4664 (value_range::set_and_canonicalize): Default bitmap to zero. 4665 (set_value_range_to_nonnull): Remove. 4666 (set_value_range_to_null): Likewise. 4667 (set_value_range): Likewise. 4668 (set_value_range_to_value): Likewise. 4669 (extract_range_from_unary_expr): Work on value_range_base. 4670 (extract_range_from_binary_expr_1): Likewise. Rename to... 4671 (extract_range_from_binary_expr): ... this. 4672 * tree-vrp.c (value_range::update): Clear equiv bitmap 4673 if required. 4674 (value_range::move): New, move equiv bitmap. 4675 (value_range_base::set_undefined): Avoid assignment. 4676 (value_range::set_undefined): Likewise. 4677 (value_range_base::set_varying): Likewise. 4678 (value_range::set_varying): Likewise. 4679 (set_value_range): Remove. 4680 (value_range_base::set): New overload for value. 4681 (value_range::set): Likewise. 4682 (set_value_range_to_nonnull): Remove. 4683 (value_range_base::set_nonnull): New. 4684 (value_range::set_nonnull): Likewise. 4685 (set_value_range_to_null): Remove. 4686 (value_range_base::set_null): New. 4687 (value_range::set_null): Likewise. 4688 (range_is_null): Work on value_range_base. 4689 (range_is_nonnull): Likewise. 4690 (ranges_from_anti_range): Likewise. 4691 (extract_range_into_wide_ints): Likewise. 4692 (extract_range_from_multiplicative_op): Likewise. 4693 (extract_range_from_binary_expr): Likewise. Update for API changes. 4694 (extract_range_from_unary_expr): Likewise. Remove OBJ_TYPE_REF 4695 handling. 4696 (value_range::intersect_helper): Avoid copy and assignment. 4697 (value_range::union_helper): Likewise. 4698 (determine_value_range_1): Adjust. 4699 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range): 4700 Avoid assignment by using move. 4701 (evrp_range_analyzer::record_ranges_from_stmt): Avoid assignment. 4702 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis): 4703 Likewise. 4704 * tree-ssanames.c (get_range_info): Likewise. 4705 * vr-values.h (vr_values::get_vr_for_comparison): Adjust API. 4706 * vr-values.c (vr_values::get_value_range): Adjust. 4707 (vr_values::update_value_range): Likewise. 4708 (symbolic_range_based_on_p): Work on value_range_base. 4709 (vr_values::extract_range_from_binary_expr): Use value_range_base. 4710 (vr_values::extract_range_from_unary_expr): Likewise. 4711 (vr_values::extract_range_from_cond_expr): Avoid assignment. 4712 (vr_values::extract_range_from_comparison): Adjust. 4713 (vr_values::check_for_binary_op_overflow): Use value_range_base. 4714 (vr_values::extract_range_basic): Adjust. 4715 (vr_values::adjust_range_with_scev): Likewise. 4716 (vr_values::vrp_visit_assignment_or_call): Likewise. 4717 (vr_values::get_vr_for_comparison): Change API to avoid 4718 assignment and copy construction. 4719 (vr_values::compare_name_with_value): Adjust accordingly. 4720 (vr_values::compare_names): Likewise. 4721 (vr_values::extract_range_from_phi_node): Avoid assignment and 4722 bogus in-place modify of equiv bitmap. 4723 (vr_values::simplify_bit_ops_using_ranges): Use value_range_base. 4724 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Adjust 4725 for extract_range_from_unary_expr API change. 4726 * ipa-cp.c (ipa_vr_operation_and_type_effects): Likewise. 4727 47282018-11-12 Eric Botcazou <ebotcazou@adacore.com> 4729 4730 * config/mcore/mcore.h (WORD_REGISTER_OPERATIONS): Remove duplicate. 4731 47322018-11-12 Richard Biener <rguenther@suse.de> 4733 4734 * tree-vrp.h (value_range_base::symbolic_p, 4735 value_range_base::constant_p, value_range_base::zero_p, 4736 value_range_base::singleton_p): Move from value_range. 4737 (value_range::dump): Add. 4738 * gimple-ssa-evrp-analyze.c 4739 (evrp_range_analyzer::record_ranges_from_phis): Use set_varying. 4740 * ipa-cp.c (ipcp_vr_lattice::print): Use dump_value_range. 4741 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis): 4742 Use set_varying. 4743 * tree-vrp.c (value_range::symbolic_p): Move to value_range_base. 4744 (value_range::constant_p): Likewise. 4745 (value_range::singleton_p): Likewise. 4746 (value_range_base::dump): Add. 4747 (set_value_range_to_undefined): Remove. 4748 (set_value_range_to_varying): Likewise. 4749 (range_int_cst_p): Take value_range_base argument. 4750 (range_int_cst_singleton_p): Likewise. 4751 (value_range_constant_singleton): Likewise. 4752 (vrp_set_zero_nonzero_bits): Likewise. 4753 (extract_range_from_multiplicative_op): Use set_varying. 4754 (extract_range_from_binary_expr_1): Likewise. Use set_undefined. 4755 (extract_range_from_unary_expr): Likewise. 4756 (dump_value_range_base): Change to overload of dump_value_range. 4757 (vrp_prop::vrp_initialize): Use set_varying and set_undefined. 4758 (vrp_prop::visit_stmt): Likewise. 4759 (value_range::intersect_helper): Likewise. 4760 (value_range::union_helper): Likewise. 4761 (determine_value_range_1): Likewise. 4762 47632018-11-12 Richard Biener <rguenther@suse.de> 4764 4765 * tree-vrp.c (set_value_range_to_nonnull): Clear equiv. 4766 (set_value_range_to_null): Likewise. 4767 * vr-values.c (vr_values::extract_range_from_comparison): 4768 Clear equiv for constant singleton ranges. 4769 47702018-11-12 Wei Xiao <wei3.xiao@intel.com> 4771 4772 * config/i386/sse.md: Combine VFIXUPIMM* patterns 4773 (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update. 4774 (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update. 4775 (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Remove. 4776 (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update. 4777 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update. 4778 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Remove. 4779 47802018-11-11 Sandra Loosemore <sandra@codesourcery.com> 4781 4782 PR c/69502 4783 * doc/extend.texi (Common Type Attributes): For the align type 4784 attribute, copy language about decreasing alignment from the 4785 corresponding variable attribute. 4786 47872018-11-11 Xianmiao Qu <xianmiao_qu@c-sky.com> 4788 4789 * config/csky/csky.md (*fpuv2_nmulsf3_1, *fpuv2_nmuldf3_1): Handle 4790 -frounding-math. 4791 47922018-11-11 Sandra Loosemore <sandra@codesourcery.com> 4793 4794 PR c++/43105 4795 * doc/invoke.texi (C++ Dialect Options): Add warning about mixing 4796 -frtti and -fno-rtti code. 4797 47982018-11-11 Sandra Loosemore <sandra@codesourcery.com> 4799 4800 PR c/26366 4801 * doc/extend.texi (Other Builtins): Document probability associated 4802 with __builtin_expect. 4803 48042018-11-11 Uros Bizjak <ubizjak@gmail.com> 4805 4806 PR target/87928 4807 * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI 4808 instead of (TARGET_64BIT && ix86_abi == MS_ABI). 4809 * config/i386/darwin.h (STACK_BOUNDARY): Ditto. 4810 * config/i386/cygming.h (STACK_BOUNDARY): Remove. 4811 48122018-11-11 Xianmiao Qu <xianmiao_qu@c-sky.com> 4813 4814 * config/csky/csky-linux-elf.h (CC1_SPEC): Support -profile. 4815 48162018-11-11 Xianmiao Qu <xianmiao_qu@c-sky.com> 4817 4818 * config/csky/csky.h (ASM_PREFERRED_EH_DATA_FORMAT): Define. 4819 48202018-11-11 Richard Biener <rguenther@suse.de> 4821 4822 * tree-vrp.h (class value_range_base): New base class for 4823 value_range containing all but the m_equiv member. 4824 (dump_value_range_base): Add. 4825 (range_includes_zero_p): Work on value_range_base. 4826 * tree-vrp.c (value_range_base::set): Split out base handling 4827 from... 4828 (value_range::set): this. 4829 (value_range::set_equiv): New. 4830 (value_range_base::value_range_base): New constructors. 4831 (value_range_base::check): Split out base handling from... 4832 (value_range::check): this. 4833 (value_range::equal_p): Refactor in terms of 4834 ignore_equivs_equal_p which is now member of the base. 4835 (value_range_base::set_undefined): New. 4836 (value_range_base::set_varying): Likewise. 4837 (value_range_base::dump):Split out base handling from... 4838 (value_range::dump): this. 4839 (value_range_base::set_and_canonicalize): Split out base handling 4840 from... 4841 (value_range::set_and_canonicalize): this. 4842 (value_range_base::union_): New. 4843 * ipa-prop.h (struct ipa_jump_func): Use value_range_base * 4844 for m_vr. 4845 * ipa-cp.c (class ipcp_vr_lattice): Use value_range_base 4846 instead of value_range everywhere. 4847 (ipcp_vr_lattice::print): Use dump_value_range_base. 4848 (ipcp_vr_lattice::meet_with): Adjust. 4849 (ipcp_vr_lattice::meet_with_1): Likewise. 4850 (ipa_vr_operation_and_type_effects): Likewise. 4851 (propagate_vr_across_jump_function): Likewise. 4852 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Likewise. 4853 (ipa_get_value_range): Likewise. 4854 (ipa_set_jfunc_vr): Likewise. 4855 (ipa_compute_jump_functions_for_edge): Likewise. 4856 48572018-11-10 Sandra Loosemore <sandra@codesourcery.com> 4858 4859 PR middle-end/65703 4860 * doc/invoke.texi (Optimize Options): Add @opindex entries 4861 for the positive forms of -fno-xxx and -mno-xxx options 4862 that were lacking them. 4863 48642018-11-10 Segher Boessenkool <segher@kernel.crashing.org> 4865 4866 * combine.c (make_more_copies): Only make an intermediate copy if the 4867 dest of a move is a pseudo. 4868 48692018-11-09 Maya Rashish <coypu@sdf.org> 4870 4871 PR target/87221 4872 * config/netbsd-elf.h (NETBSD_STARTFILE_SPEC): Use crtbeginS.o for PIE. 4873 (NETBSD_ENDFILE_SPEC): Use crtendS.o for PIE. 4874 48752018-11-09 Sandra Loosemore <sandra@codesourcery.com> 4876 4877 PR driver/41179 4878 PR middle-end/65703 4879 * doc/invoke.texi (Optimize Options): Clarify default behavior 4880 for -fno-toplevel-reorder, -fno-defer-pop, and -fno-branch-count-reg. 4881 48822018-11-09 Bernd Edlinger <bernd.edlinger@hotmail.de> 4883 4884 PR tree-optimization/87940 4885 * expr.c (string_constant): Don't strip NOPS in subexpressions. 4886 Fold PLUS_EXPR correctly. 4887 48882018-11-09 Ilya Leoshkevich <iii@linux.ibm.com> 4889 4890 PR target/87762 4891 * config/s390/s390.c (s390_safe_relative_long_p): New function. 4892 (annotate_constant_pool_refs): Skip insns which support 4893 relative addressing. 4894 (annotate_constant_pool_refs_1): New helper function. 4895 (find_constant_pool_ref): Skip insns which support relative 4896 addression. 4897 (find_constant_pool_ref_1): New helper function. 4898 (replace_constant_pool_ref): Skip insns which support 4899 relative addressing. 4900 (replace_constant_pool_ref_1): New helper function. 4901 (s390_mainpool_start): Adapt to the new signature. 4902 (s390_mainpool_finish): Likewise. 4903 (s390_chunkify_start): Likewise. 4904 (s390_chunkify_finish): Likewise. 4905 (pass_s390_early_mach::execute): Likewise. 4906 (s390_prologue_plus_offset): Likewise. 4907 (s390_emit_prologue): Likewise. 4908 (s390_emit_epilogue): Likewise. 4909 49102018-11-09 Jakub Jelinek <jakub@redhat.com> 4911 4912 * gimplify.c (gimplify_scan_omp_clauses): Call sorry_at for valid 4913 but unsupported lastprivate with conditional modifier. 4914 49152018-11-09 Jeff Law <law@redhat.com> 4916 4917 * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): Handle 4918 unused argument better. Add gcc_unreachable to silence warning. 4919 49202018-11-09 Martin Sebor <msebor@redhat.com> 4921 4922 PR middle-end/81824 4923 * attribs.c (has_attribute): New helper function. 4924 (decls_mismatched_attributes, maybe_diag_alias_attributes): Same. 4925 * attribs.h (decls_mismatched_attributes): Declare. 4926 * cgraphunit.c (handle_alias_pairs): Call maybe_diag_alias_attributes. 4927 (maybe_diag_incompatible_alias): Use OPT_Wattribute_alias_. 4928 * common.opt (-Wattribute-alias): Take an argument. 4929 (-Wno-attribute-alias): New option. 4930 * doc/extend.texi (Common Function Attributes): Document copy. 4931 (Common Variable Attributes): Same. 4932 * doc/invoke.texi (-Wmissing-attributes): Document enhancement. 4933 (-Wattribute-alias): Document new option argument. 4934 49352018-11-09 Richard Earnshaw <rearnsha@arm.com> 4936 4937 * config/arm/parsecpu.awk (/alias/): Tighten invisible alias 4938 matching criteria. Remove unused array initializer. 4939 49402018-11-09 Bill Schmidt <wschmidt@linux.ibm.com> 4941 Jinsong Ji <jji@us.ibm.com> 4942 4943 * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix incorrect 4944 constraints by introducing a new temporary. 4945 (_mm_cvtss_si64): Likewise. 4946 49472018-11-09 Martin Liska <mliska@suse.cz> 4948 4949 * common.opt: Add -fipa-stack-alignment flag. 4950 * doc/invoke.texi: Document it. 4951 * final.c (rest_of_clean_state): Guard stack 4952 shrinking with flag. 4953 49542018-11-09 Martin Liska <mliska@suse.cz> 4955 4956 * cgraph.h (ipa_discover_readonly_nonaddressable_vars): Rename 4957 to ... 4958 (ipa_discover_variable_flags): ... this. 4959 * common.opt: Come up with new flag -fipa-reference-addressable. 4960 * doc/invoke.texi: Document it. 4961 * ipa-reference.c (propagate): Call the renamed fn. 4962 * ipa-visibility.c (whole_program_function_and_variable_visibility): 4963 Likewise. 4964 * ipa.c (ipa_discover_readonly_nonaddressable_vars): Renamed to 4965 ... 4966 (ipa_discover_variable_flags): ... this. Discover 4967 non-addressable variables only with the newly added flag. 4968 * opts.c: Enable the newly added flag with -O1 and higher 4969 optimization level. 4970 49712018-11-09 David Malcolm <dmalcolm@redhat.com> 4972 4973 * json.cc (selftest::test_writing_literals): Fix comment. 4974 49752018-11-09 Martin Liska <mliska@suse.cz> 4976 4977 * config/aarch64/aarch64.c (aarch64_parse_arch): Do not copy 4978 string to a stack buffer. 4979 (aarch64_parse_cpu): Likewise. 4980 (aarch64_parse_tune): Likewise. 4981 49822018-11-09 Richard Biener <rguenther@suse.de> 4983 4984 PR tree-optimization/87953 4985 * tree-vect-loop.c (vectorizable_reduction): For analysis 4986 always pass ops[0] to vectorizable_condition. 4987 49882018-11-09 Stafford Horne <shorne@gmail.com> 4989 Richard Henderson <rth@twiddle.net> 4990 Joel Sherrill <joel@rtems.org> 4991 4992 * common/config/or1k/or1k-common.c: New file. 4993 * config/or1k/*: New. 4994 * config.gcc (or1k*-*-*): New. 4995 * configure.ac (or1k*-*-*): New test for openrisc tls. 4996 * configure: Regenerated. 4997 * doc/install.texi: Document OpenRISC triplets. 4998 * doc/invoke.texi: Document OpenRISC arguments. 4999 * doc/md.texi: Document OpenRISC. 5000 50012018-11-09 Richard Earnshaw <rearnsha@arm.com> 5002 5003 * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s. 5004 (arm7tdmi-s): Delete CPU. 5005 (arm710t): Add aliases for arm720t and arm740t. 5006 (arm720t, arm740t): Delete CPUs. 5007 (arm920t): Add aliases for arm920, arm922t and arm940t. 5008 (arm920, arm922t, arm940t): Delete CPUs. 5009 (arm10tdmi): Add alias for arm1020t. 5010 (arm1020t): Delete CPU. 5011 (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s. 5012 (arm946e-s, arm966e-s, arm968e-s): Delete CPUs. 5013 (arm10e): Add aliases for arm1020e and arm1022e. 5014 (arm1020e, arm1022e): Delete CPU. 5015 * config/arm/arm.md (generic_sched): Remove entries that are now 5016 handled by aliases. 5017 (generic_vfp): Likewise. 5018 * config/arm/arm1020e.md: Simplify tuning selection based on alias 5019 changes. 5020 * config/arm/arm-tune.md: Regenerated. 5021 * config/arm/arm-tables.opt: Regenerated. 5022 50232018-11-09 Richard Biener <rguenther@suse.de> 5024 5025 PR tree-optimization/87621 5026 * tree-vect-loop.c (vectorizable_reduction): Handle reduction 5027 op with only phi inputs. 5028 * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h. 5029 (ch_base::copy_headers): Run CSE on copied loop headers. 5030 (pass_ch_vect::process_loop_p): Simplify. 5031 50322018-11-09 Alexandre Oliva <oliva@adacore.com> 5033 5034 * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust 5035 for non-w64 x86_64 biarch. 5036 50372018-11-09 Alexandre Oliva <aoliva@redhat.com> 5038 5039 PR rtl-optimization/86438 5040 * compare-elim.c (try_eliminate_compare): Use SET_SRC instead 5041 of in_b for the compare if in_b is SET_DEST. 5042 5043 PR target/87793 5044 * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject 5045 non-toplevel UNSPEC. 5046 50472018-11-08 Aldy Hernandez <aldyh@redhat.com> 5048 5049 * tree-vrp.c (value_range::check): Do not access internals 5050 directly. 5051 (value_range::singleton_p): Same. 5052 (value_range::type): Same. 5053 (vrp_finalize): Use value_range API. 5054 50552018-11-08 Aldy Hernandez <aldyh@redhat.com> 5056 5057 * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly. 5058 50592018-11-08 Aldy Hernandez <aldyh@redhat.com> 5060 5061 * vr-values.c (vr_values::get_value_range): Use value_range API 5062 instead of piecing together ranges. 5063 (vr_values::update_value_range): Same. 5064 50652018-11-08 Aldy Hernandez <aldyh@redhat.com> 5066 5067 * gimple-fold.c (size_must_be_zero_p): Use value_range API instead 5068 of performing ad-hoc calculations. 5069 * tree-ssanames.c (set_range_info): New overloaded function 5070 accepting value_range &. 5071 (get_range_info): Same. 5072 * tree-ssanames.h (set_range_info_raw): Remove. 5073 (set_range_info): New prototype. 5074 (get_range_info): Same. 5075 * tree-vrp.h (value_range::null_p): Rename to zero_p. 5076 * tree-vrp.c (value_range::null_p): Same. 5077 50782018-11-09 Jan Hubicka <jh@suse.cz> 5079 5080 * tree.c (fld_type_variant_equal_p): Test user align flag. 5081 (flt_type_variant): Copy user align flag. 5082 (fld_incomplete_type_of): Clear it. 5083 50842018-11-09 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org> 5085 5086 * config/arm/neon.md (div<mode>3): New pattern. 5087 50882018-11-08 Andi Kleen <ak@linux.intel.com> 5089 5090 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New. 5091 (OPTION_MASK_ISA_PTWRITE_UNSET): New. 5092 (ix86_handle_option): Handle OPT_mptwrite. 5093 * config/i386/cpuid.h (bit_PTWRITE): Add. 5094 * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite. 5095 * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64. 5096 * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__. 5097 * config/i386/i386.c (ix86_target_string): Handle ptwrite. 5098 (ix86_option_override_internal): Handle PTA_PTWRITE. 5099 (ix86_valid_target_attribute_inner_p): Define ptwrite. 5100 (def_builtin2): Force UINT64 to be 64bit only. 5101 * config/i386/i386.h (TARGET_PTWRITE): Add. 5102 (TARGET_PTWRITE_P): Add. 5103 (PTA_PTWRITE): Add. 5104 * config/i386/i386.md: Define ptwrite. 5105 * config/i386/i386.opt: Add -mptwrite. 5106 * config/i386/immintrin.h (_ptwrite64): Add. 5107 (_ptwrite32): Add 5108 * doc/extend.texi: Document __builtin_ia32_ptwrite*. 5109 * doc/invoke.texi: Document -mptwrite. 5110 51112018-11-08 Peter Bergner <bergner@linux.ibm.com> 5112 5113 PR rtl-optimization/87600 5114 * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage. 5115 * lra-constraints.c (process_alt_operands): Skip illegal hard 5116 register usage. Prefer reloading non hard register operands. 5117 51182018-11-08 Sandra Loosemore <sandra@codesourcery.com> 5119 5120 PR other/36572 5121 * doc/invoke.texi (Optimize Options): Clarify default behavior 5122 for -fno-sched-interblock and -fno-sched-spec. 5123 51242018-11-08 Roman Geissler <roman.geissler@amadeus.com> 5125 5126 * collect2.c (linker_select): Add USE_LLD_LD. 5127 (ld_suffixes): Add ld.lld. 5128 (main): Handle -fuse-ld=lld. 5129 * common.opt (-fuse-ld=lld): New option. 5130 * doc/invoke.texi (-fuse-ld=lld): Document. 5131 * opts.c (common_handle_option): Handle OPT_fuse_ld_lld. 5132 51332018-11-08 Paul Koning <ni1d@arrl.net> 5134 5135 * config/pdp11/constraints.md: Add "Z" series constraints for use 5136 with pre-dec and post-inc addressing. 5137 * config/pdp11/pdp11-protos.m (expand_block_move): Delete. 5138 (pdp11_expand_operands): Add int argument (word count). 5139 (pdp11_sp_frame_offset): Delete. 5140 (pdp11_cmp_length): New function. 5141 (pushpop_regeq): New function. 5142 * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P): 5143 Add hook. 5144 (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new 5145 frame layout. 5146 (pdp11_initial_elimination_offset): Ditto. 5147 (pdp11_expand_operands): Add word count argument. Bugfixes. 5148 (output_move_multiple): Change how pointer adjustment is done. 5149 (pdp11_gen_int_label): Correct format. 5150 (output_ascii): Ditto. 5151 (pdp11_asm_output_var): Add code for DEC assembler case. 5152 (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer 5153 value. 5154 (legitimate_const_double_p): Ditto. 5155 (pdp11_register_move_cost): Adjust for new register classes. 5156 (pdp11_regno_reg_class): Ditto. 5157 (expand_block_move): Delete. 5158 (pushpop_regeq): New function. 5159 (pdp11_legitimate_address_p): Bugfix in check for constant 5160 offset. 5161 (pdp11_sp_frame_offset): Delete. 5162 (pdp11_reg_save_size): New helper function for new frame layout. 5163 (output_addr_const_pdp11): Remove CONST_DOUBLE case. 5164 (pdp11_expand_shift): Bugfix in check for constant shift count. 5165 (pdp11_shift_length): Ditto. 5166 (pdp11_assemble_shift): Copy input to pdp11_expand_operands. 5167 (pdp11_cmp_length): New function. 5168 * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for 5169 some compile options. 5170 (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM. 5171 (CALL_USED_REGISTERS): Ditto. 5172 (ELIMINABLE_REGS): Ditto. 5173 (REGISTER_NAMES): Ditto. 5174 (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z 5175 constraints. 5176 (REG_CLASS_NAMES): Ditto. 5177 (REG_CLASS_CONTENTS): Ditto. Also remove 5178 HARD_FRAME_POINTER_REGNUM. 5179 (CPU_REG_CLASS): New macro. 5180 (CLASS_MAX_NREGS): Adjust for new register classes. 5181 (FUNCTION_PROFILER): Make no-op. 5182 (may_call_alloca): Remove unused declaration. 5183 (ASM_OUTPUT_ALIGN): Add workaround for PR87795. 5184 (ASM_OUTPUT_SKIP): Fix format. 5185 * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM. 5186 (HARD_FRAME_POINTER_REGNUM): Remove. 5187 (return): Delete. 5188 (*rts): Rename. Remove epilogue related checks. 5189 (cmpsi, cmpdi): New insn. 5190 (cbranch<mode>4): Change to apply to SI and DI modes as well. 5191 (mov<mode>): Change constraints to enforce that push/pop 5192 destination cannot use the same register as source. 5193 (*mov<mode><cc_cc>): Ditto. 5194 (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move 5195 at assembly output rather than as RTL expander. 5196 (zero_extendqihi2): Bugfix in check for same registers. 5197 (adddi3_nocc): Bugfix in check for constant operand. 5198 (addsi3_nocc): Ditto. 5199 (subdi3_nocc): Ditto. 5200 (subsi3_nocc): Ditto. 5201 (negdi2_nocc): Copy input to pdp11_expand_operands. 5202 (negsi2_nocc): Ditto. 5203 (bswap2_nocc): Ditto. 5204 * config/pdp11/pdp11.opt (mlra): Fix documentation. 5205 * config/pdp11/t-pdp11: Use -Os. 5206 52072018-11-08 Richard Earnshaw <rearnsha@arm.com> 5208 5209 * config/arm/parsecpu.awk (/alias/): New parsing rule. 5210 (/begin cpu/): Check that the cpu name hasn't been previously defined. 5211 (gen_comm_data): Print out CPU alias tables. 5212 (check_cpu): Match aliases when checking the CPU name. 5213 * config/arm/arm-protos.h (cpu_alias): New structure. 5214 (cpu_option): Add entry for aliases. 5215 * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110 5216 strongarm1100 and strongarm1110. 5217 (strongarm110, strongarm1100, strongarm1110): Delete CPU entries. 5218 (config/arm/arm-generic.md): Remove redundant references to 5219 strongarm110, strongarm1100 and strongarm1110. 5220 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option): 5221 Scan aliases for additional hints. 5222 (arm_parse_cpu_option_name): Also match a cpu name against the list 5223 of aliases. 5224 * config/arm/arm-tables.opt: Regenerated. 5225 * config/arm/arm-tune.md: Regenerated. 5226 52272018-11-08 Jakub Jelinek <jakub@redhat.com> 5228 5229 * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR, 5230 BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT, 5231 BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR, 5232 BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR, 5233 BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR, 5234 BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New. 5235 * gengtype.c (open_base_files): Add omp-general.h. 5236 * gimple.c (gimple_build_omp_critical): 5237 (gimple_build_omp_taskgroup): Add CLAUSES argument. Call 5238 gimple_omp_taskgroup_set_clauses. 5239 (gimple_build_omp_atomic_load): Add mo argument, call 5240 gimple_omp_atomic_set_memory_order. 5241 (gimple_build_omp_atomic_store): Likewise. 5242 (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP. 5243 * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT 5244 instead of GSS_OMP. 5245 (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead 5246 of GSS_OMP_SINGLE_LAYOUT, adjust comments. 5247 * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT 5248 and GF_OMP_ATOMIC_MEMORY_ORDER. Remove GF_OMP_ATOMIC_SEQ_CST, use 5249 different value for GF_OMP_ATOMIC_NEED_VALUE. 5250 (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to 5251 comments. 5252 (struct gimple_statement_omp_single_layout): And remove here. 5253 (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather 5254 than gimple_statement_omp_single_layout. 5255 (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow 5256 GIMPLE_OMP_TEAMS. 5257 (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise. 5258 (gimple_omp_subcode): Formatting fix. 5259 (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr, 5260 gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg, 5261 gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg, 5262 gimple_omp_teams_host, gimple_omp_teams_set_host, 5263 gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p, 5264 gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr, 5265 gimple_omp_taskgroup_set_clauses): New inline functions. 5266 (gimple_build_omp_atomic_load): Add enum omp_memory_order argument. 5267 (gimple_build_omp_atomic_store): Likewise. 5268 (gimple_omp_atomic_seq_cst_p): Remove. 5269 (gimple_omp_atomic_memory_order): New function. 5270 (gimple_omp_atomic_set_seq_cst): Remove. 5271 (gimple_omp_atomic_set_memory_order): New function. 5272 (gimple_build_omp_taskgroup): Add clauses argument. 5273 * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function. 5274 (dump_gimple_omp_task): Print taskwait with depend clauses. 5275 (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use 5276 dump_omp_atomic_memory_order. 5277 (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP. 5278 * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY, 5279 GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL. 5280 (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags, 5281 renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS, 5282 ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and 5283 ORT_UNTIED_TASKLOOP enumerators. 5284 (enum gimplify_defaultmap_kind): New. 5285 (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and 5286 target_map_pointers_as_0len_arrays members, add defaultmap. 5287 (new_omp_context): Initialize defaultmap member. 5288 (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE. 5289 (maybe_fold_stmt): Don't fold even in host teams regions. 5290 (omp_firstprivatize_variable): Handle ORT_TASKGROUP like 5291 ORT_WORKSHARE. Test ctx->defaultmap[GDMK_SCALAR] instead of 5292 ctx->omp_firstprivatize_variable. 5293 (omp_add_variable): Don't add private/firstprivate for VLAs in 5294 ORT_TASKGROUP. 5295 (omp_default_clause): Print "taskloop" rather than "task" if 5296 ORT_*TASKLOOP. 5297 (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE. 5298 Handle new defaultmap clause kinds. 5299 (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE. Allow simd 5300 iterator to be lastprivate or private. Fix up diagnostics if linear 5301 is used on collapse>1 simd iterator. 5302 (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE. 5303 (gimplify_omp_depend): New function. 5304 (gimplify_scan_omp_clauses): Add shared clause on parallel for 5305 combined parallel master taskloop{, simd} if taskloop has 5306 firstprivate, lastprivate or reduction clause. Handle 5307 OMP_CLAUSE_REDUCTION_TASK diagnostics. Adjust tests for 5308 ORT_COMBINED_TEAMS. Gimplify depend clauses with iterators. Handle 5309 cancel and simd OMP_CLAUSE_IF_MODIFIERs. Handle 5310 OMP_CLAUSE_NONTEMPORAL. Handle new defaultmap clause kinds. Handle 5311 OMP_CLAUSE_{TASK,IN}_REDUCTION. Diagnose invalid conditional 5312 lastprivate. 5313 (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL. Handle 5314 GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY. 5315 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL. Handle 5316 OMP_CLAUSE_{TASK,IN}_REDUCTION. 5317 (gimplify_omp_task): Handle taskwait with depend clauses. 5318 (gimplify_omp_for): Add shared clause on parallel for combined 5319 parallel master taskloop{, simd} if taskloop has firstprivate, 5320 lastprivate or reduction clause. Use ORT_TASKLOOP or 5321 ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK. Adjust 5322 tests for ORT_COMBINED_TEAMS. Handle C++ range for loops with 5323 NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS. Firstprivatize 5324 __for_end and __for_range temporaries on OMP_PARALLEL for 5325 distribute parallel for{, simd}. Move OMP_CLAUSE_REDUCTION 5326 and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct 5327 sandwiched in between two taskloops. 5328 (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR] 5329 instead of ctx->omp_firstprivatize_variable. 5330 (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or 5331 ORT_COMBINED_HOST_TEAMS if not inside of target construct. If 5332 host teams, use gimplify_and_return_first etc. for body like 5333 for target or target data constructs, and at the end call 5334 gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object. 5335 (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead 5336 of OMP_ATOMIC_SEQ_CST, pass it as new argument to 5337 gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove 5338 gimple_omp_atomic_set_seq_cst calls. 5339 (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate 5340 case, handle taskgroup clauses. 5341 * lto-streamer-out.c (hash_tree): Handle 5342 OMP_CLAUSE_{TASK,IN}_REDUCTION. 5343 * Makefile.in (GTFILES): Add omp-general.h. 5344 * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND, 5345 BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START, 5346 BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START, 5347 BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START, 5348 BUILT_IN_GOMP_LOOP_DOACROSS_START, 5349 BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT, 5350 BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT, 5351 BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START, 5352 BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START, 5353 BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START, 5354 BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START, 5355 BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT, 5356 BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT, 5357 BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME, 5358 BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME, 5359 BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START, 5360 BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER, 5361 BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER, 5362 BUILT_IN_GOMP_TASK_REDUCTION_REMAP, 5363 BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins. 5364 * omp-expand.c (workshare_safe_to_combine_p): Return false for 5365 non-worksharing loops. 5366 (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero. 5367 (determine_parallel_type): Don't combine parallel with worksharing 5368 which has _reductemp_ clause. 5369 (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or 5370 GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime* 5371 if there is nonmonotonic modifier or if there is no modifier and no 5372 ordered clause. For dynamic and guided schedule without monotonic 5373 and nonmonotonic modifier, default to nonmonotonic. 5374 (expand_omp_for): Likewise. Adjust expand_omp_for_generic caller, use 5375 GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are 5376 task reductions. 5377 (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if 5378 there are any reduction clauses. 5379 (expand_taskwait_call): New function. 5380 (expand_teams_call): New function. 5381 (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call 5382 expand_teams_call for it. Formatting fix. Handle taskwait with 5383 depend clauses. 5384 (expand_omp_for_generic): Add SCHED_ARG argument. Handle expansion 5385 of worksharing loops with task reductions. 5386 (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle 5387 expansion of worksharing loops with task reductions. 5388 (expand_omp_sections): Handle expansion of sections with task 5389 reductions. 5390 (expand_omp_synch): For host teams call expand_omp_taskreg. 5391 (omp_memory_order_to_memmodel): New function. 5392 (expand_omp_atomic_load, expand_omp_atomic_store, 5393 expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order 5394 instead of gimple_omp_atomic_seq_cst_p. 5395 (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with 5396 depend clauses as a standalone directive. 5397 * omp-general.c (enum omp_requires): New variable. 5398 (omp_extract_for_data): Initialize have_reductemp member. Allow 5399 NE_EXPR even in OpenMP loops, transform them into LT_EXPR or 5400 GT_EXPR loops depending on incr sign. Formatting fixes. 5401 * omp-general.h (struct omp_for_data): Add have_reductemp member. 5402 (enum omp_requires): New enum. 5403 (omp_requires_mask): Declare. 5404 * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix. 5405 Fix comment typos. 5406 * omp-low.c (struct omp_context): Add task_reductions and 5407 task_reduction_map fields. 5408 (is_host_teams_ctx): New function. 5409 (is_taskreg_ctx): Return true also if is_host_teams_ctx. 5410 (use_pointer_for_field): Use is_global_var instead of 5411 TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized 5412 in outer contexts. 5413 (build_outer_var_ref): Ignore taskgroup outer contexts. 5414 (delete_omp_context): Release task_reductions and task_reduction_map. 5415 (scan_sharing_clauses): Don't add any fields for reduction clause on 5416 taskloop. Handle OMP_CLAUSE__REDUCTEMP_. Handle 5417 OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task 5418 modifier. Don't ignore shared clauses in is_host_teams_ctx contexts. 5419 Handle OMP_CLAUSE_NONTEMPORAL. 5420 (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if 5421 needed. 5422 (scan_omp_parallel): Add _reductemp_ clause if there are any reduction 5423 clauses with task modifier. 5424 (scan_omp_task): Handle taskwait with depend clauses. 5425 (finish_taskreg_scan): Move field corresponding to _reductemp_ clause 5426 first. Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present. 5427 Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL. 5428 (scan_omp_for): Fix comment formatting. 5429 (scan_omp_teams): Handle host teams constructs. 5430 (check_omp_nesting_restrictions): Allow teams with no outer 5431 OpenMP context. Adjust diagnostics for teams strictly nested into 5432 some explicit OpenMP construct other than target. Allow OpenMP atomics 5433 inside of simd regions. 5434 (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups. 5435 (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump 5436 taskreg_nesting_level while scanning host teams construct. 5437 (task_reduction_read): New function. 5438 (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop 5439 construct. Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_ 5440 clauses. Handle OMP_CLAUSE_REDUCTION with task modifier. Remove 5441 second argument create_tmp_var if it is NULL. Don't ignore shared 5442 clauses in is_host_teams_ctx contexts. Handle 5443 OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE 5444 clauses. 5445 (lower_reduction_clauses): Ignore reduction clauses with task 5446 modifier. Remove second argument create_tmp_var if it is NULL. 5447 Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed. 5448 (lower_send_clauses): Ignore reduction clauses with task modifier. 5449 Handle OMP_CLAUSE__REDUCTEMP_. Don't send anything for 5450 OMP_CLAUSE_REDUCTION on taskloop. Handle OMP_CLAUSE_IN_REDUCTION. 5451 (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't 5452 rely that it is the last stmt in body so far. Ignore outer taskgroup 5453 contexts. 5454 (omp_task_reductions_find_first, omp_task_reduction_iterate, 5455 lower_omp_task_reductions): New functions. 5456 (lower_omp_sections): Handle reduction clauses with taskgroup 5457 modifiers. Adjust maybe_add_implicit_barrier_cancel caller. 5458 (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller. 5459 (lower_omp_for): Likewise. Handle reduction clauses with taskgroup 5460 modifiers. 5461 (lower_omp_taskgroup): Handle taskgroup reductions. 5462 (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer. 5463 Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses. 5464 (lower_depend_clauses): If there are any 5465 OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET 5466 depend clauses, use a new array format. If OMP_CLAUSE_DEPEND_LAST is 5467 seen, assume lowering is done already and return early. Set kind 5468 on artificial depend clause to OMP_CLAUSE_DEPEND_LAST. 5469 (lower_omp_taskreg): Handle reduction clauses with task modifier on 5470 parallel construct. Handle reduction clause on taskloop construct. 5471 Handle taskwait with depend clauses. 5472 (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams 5473 for host teams constructs. 5474 * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction, 5475 nontemporal and _reductemp_ clause entries. 5476 (omp_clause_code_name): Likewise. 5477 (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION, 5478 OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_. 5479 * tree-core.h (enum omp_clause_code): Add 5480 OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}. 5481 (enum omp_clause_defaultmap_kind, enum omp_memory_order): New. 5482 (struct tree_base): Add omp_atomic_memory_order field into union. 5483 Remove OMP_ATOMIC_SEQ_CST comment. 5484 (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET 5485 and OMP_CLAUSE_DEPEND_DEPOBJ. 5486 (struct tree_omp_clause): Add subcode.defaultmap_kind. 5487 * tree.def (OMP_TASKGROUP): Add another operand, move next to other 5488 OpenMP constructs with body and clauses operands. 5489 * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP. 5490 (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE. 5491 (OMP_TASKGROUP_CLAUSES): Define. 5492 (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of 5493 OMP_CLAUSE__LOOPTEMP_. 5494 (OMP_ATOMIC_SEQ_CST): Remove. 5495 (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE, 5496 OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define. 5497 (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT, 5498 OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER, 5499 OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER, 5500 OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle 5501 OMP_CLAUSE_{,IN_,TASK_}REDUCTION. 5502 (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN, 5503 OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY, 5504 OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND): 5505 Define. 5506 * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses. 5507 * tree-nested.c (convert_nonlocal_omp_clauses): Handle 5508 OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL. 5509 (convert_local_omp_clauses): Likewise. Remove useless test. 5510 * tree-parloops.c (create_call_for_reduction_1): Pass 5511 OMP_MEMORY_ORDER_RELAXED as new argument to 5512 dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store. 5513 * tree-pretty-print.c (dump_omp_iterators): New function. 5514 (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_, 5515 OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION. Print 5516 reduction modifiers. Handle OMP_CLAUSE_DEPEND_DEPOBJ and 5517 OMP_CLAUSE_DEPEND_MUTEXINOUTSET. Print iterators in depend clauses. 5518 Print __internal__ for OMP_CLAUSE_DEPEND_LAST. Handle cancel and 5519 simd OMP_CLAUSE_IF_MODIFIERs. Handle new kinds of 5520 OMP_CLAUSE_DEFAULTMAP. Print conditional: for 5521 OMP_CLAUSE_LASTPRIVATE_CONDITIONAL. 5522 (dump_omp_atomic_memory_order): New function. 5523 (dump_generic_node): Use it. Print taskgroup clauses. Print 5524 taskwait with depend clauses. 5525 * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare. 5526 * tree-streamer-in.c (unpack_ts_omp_clause_value_fields): 5527 Handle OMP_CLAUSE_{TASK,IN}_REDUCTION. 5528 * tree-streamer-out.c (pack_ts_omp_clause_value_fields, 5529 write_ts_omp_clause_tree_pointers): Likewise. 5530 55312018-11-08 David Malcolm <dmalcolm@redhat.com> 5532 5533 PR ipa/86395 5534 * doc/invoke.texi (-fdump-ipa-): Document the "-optimized", 5535 "-missed", "-note", and "-all" sub-options. 5536 * ipa-inline.c (caller_growth_limits): Port from fprintf to dump 5537 API. 5538 (can_early_inline_edge_p): Likewise. 5539 (want_early_inline_function_p): Likewise. 5540 (want_inline_self_recursive_call_p): Likewise. 5541 (recursive_inlining): Likewise. 5542 (inline_small_functions): Likewise. 5543 (flatten_function): Likewise. 5544 (ipa_inline): Likewise. 5545 (inline_always_inline_functions): Likewise. 5546 (early_inline_small_functions): Likewise. 5547 (early_inliner): Likewise. 5548 * tree-inline.c (expand_call_inline): Likewise. 5549 55502018-11-08 David Malcolm <dmalcolm@redhat.com> 5551 5552 * pretty-print.c (pp_format): Handle %f. 5553 (selftest::test_pp_format): Add test of %f. 5554 * pretty-print.h (pp_double): New macro. 5555 55562018-11-08 David Malcolm <dmalcolm@redhat.com> 5557 5558 * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro. 5559 * dumpfile.c (make_item_for_dump_cgraph_node): Move to before... 5560 (dump_pretty_printer::decode_format): Implement "%C" for 5561 cgraph_node *. 5562 (selftest::test_capture_of_dump_calls): Rename "where" to 5563 "stmt_loc". Convert test_decl to a function decl and set its 5564 location. Add a symbol_table_test RAII instance and a 5565 cgraph_node, using it to test "%C" and dump_symtab_node. 5566 55672018-11-08 Eric Botcazou <ebotcazou@adacore.com> 5568 5569 PR middle-end/87916 5570 * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P. 5571 55722018-11-08 David Malcolm <dmalcolm@redhat.com> 5573 5574 * cgraph.c: Include "selftest.h". 5575 (saved_symtab): New variable. 5576 (selftest::symbol_table_test::symbol_table_test): New ctor. 5577 (selftest::symbol_table_test::~symbol_table_test): New dtor. 5578 (selftest::test_symbol_table_test): New test. 5579 (selftest::cgraph_c_tests): New. 5580 * cgraph.h (saved_symtab): New decl. 5581 (selftest::symbol_table_test): New class. 5582 * selftest-run-tests.c (selftest::run_tests): Call 5583 selftest::cgraph_c_tests. 5584 * selftest.h (selftest::cgraph_c_tests): New decl. 5585 55862018-11-08 Richard Biener <rguenther@suse.de> 5587 5588 * tree-data-ref.h (lambda_int): New typedef. 5589 (lambda_vector_gcd): Adjust. 5590 (lambda_vector_new): Likewise. 5591 (lambda_matrix_new): Likewise. 5592 * tree-data-ref.c (print_lambda_vector): Adjust. 5593 55942018-11-08 Richard Biener <rguenther@suse.de> 5595 5596 PR tree-optimization/87929 5597 * tree-complex.c (expand_complex_comparison): Clean EH. 5598 55992018-11-08 Martin Liska <mliska@suse.cz> 5600 5601 * doc/extend.texi: Reword. 5602 * predict.c (expr_expected_value_1): Likewise. 5603 56042018-11-08 Richard Biener <rguenther@suse.de> 5605 5606 PR tree-optimization/87913 5607 * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares 5608 of extreme values to ordered comparisons. 5609 56102018-11-07 Sandra Loosemore <sandra@codesourcery.com> 5611 5612 PR middle-end/42726 5613 * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior. 5614 56152018-11-07 Sandra Loosemore <sandra@codesourcery.com> 5616 5617 * doc/invoke.texi: Remove leading dash from @opindex entries 5618 throughout the file. 5619 56202018-11-07 Sandra Loosemore <sandra@codesourcery.com> 5621 5622 PR driver/80828 5623 * doc/invoke.texi (Option Summary): Add -e and --entry. 5624 (Link Options): Likewise. 5625 56262018-11-07 Nathan Sidwell <nathan@acm.org> 5627 5628 PR 87926 5629 * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak 5630 --disable-checking bootstrap. 5631 56322018-11-07 Hafiz Abid Qadeer <abidh@codesourcery.com> 5633 5634 * configure: Regenerated. 5635 56362018-11-07 Jozef Lawrynowicz <jozef.l@mittosystems.com> 5637 5638 PR c/87691 5639 * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE 5640 to the mode of the widest field iff the widest field has mode class 5641 MODE_INT, or MODE_PARTIAL_INT and the union would be passed by 5642 reference. 5643 56442018-11-07 Nikolai Merinov <n.merinov@inango-systems.com> 5645 5646 * common.opt: Add -Wattribute-warning. 5647 * doc/invoke.texi: Add documentation for -Wno-attribute-warning. 5648 * expr.c (expand_expr_real_1): Add new attribute to warning_at 5649 call to allow user configure behavior of "warning" attribute. 5650 56512018-11-07 Segher Boessenkool <segher@kernel.crashing.org> 5652 5653 * target.def: Put @: after every vs., e.g., and i.e. where it is 5654 followed by whitespace. 5655 * doc/extend.texi: Ditto. 5656 * doc/fragments.texi: Ditto. 5657 * doc/gimple.texi: Ditto. 5658 * doc/implement-c.texi: Ditto. 5659 * doc/install.texi: Ditto. 5660 * doc/invoke.texi: Ditto. 5661 * doc/md.texi: Ditto. 5662 * doc/plugins.texi: Ditto. 5663 * doc/rtl.texi: Ditto. 5664 * doc/sourcebuild.texi: Ditto. 5665 * doc/tm.texi.in: Ditto. 5666 * doc/ux.texi: Ditto. 5667 * doc/tm.texi: Regenerate. 5668 56692018-11-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 5670 5671 * config/arm/arm-cpus.in (ares): New entry. 5672 * config/arm/arm-tables.opt: Regenerate. 5673 * config/arm/arm-tune.md: Likewise. 5674 * doc/invoke.texi (ARM Options): Document ares. 5675 56762018-11-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 5677 5678 * config/aarch64/aarch64-cores.def (ares): Define. 5679 * config/aarch64/aarch64-tune.md: Regenerate. 5680 * doc/invoke.texi (AArch64 Options): Document ares value for mtune. 5681 56822018-11-07 Jan Hubicka <jh@suse.cz> 5683 5684 * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE. 5685 (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and 5686 functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for 5687 TYPE_DECL. 5688 56892018-11-07 Richard Biener <rguenther@suse.de> 5690 5691 PR tree-optimization/87914 5692 * tree-vect-loop.c (vect_is_simple_reduction): Improve detection 5693 of nested cycles. 5694 (vectorizable_reduction): Handle shifts and rotates by dispatching 5695 to vectorizable_shift. 5696 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle 5697 in-loop uses of vect_nested_cycle defs. Merge cycle and internal 5698 def cases. 5699 (vectorizable_shift): Export and handle being called as 5700 vect_nested_cycle. 5701 (vect_analyze_stmt): Call vectorizable_shift after 5702 vectorizable_reduction. 5703 * tree-vectorizer.h (vectorizable_shift): Declare. 5704 57052018-11-07 Jan Hubicka <jh@suse.cz> 5706 5707 * ipa-devirt.c (odr_types_equivalent_p): Expect constants 5708 than const decls in TREE_VALUE of enum. 5709 (dump_type_inheritance_graph): Improve duplicate dumping. 5710 (free_enum_values): New. 5711 (build_type_inheritance_graph): Use it. 5712 * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums 5713 which are not main variants or not ODR types. 5714 (verify_type_variant): Expect variants to have no TYPE_VALUES. 5715 57162018-11-07 Richard Biener <rguenther@suse.de> 5717 5718 * ipa-inline.c (want_inline_small_function_p): Compute 5719 big_speedup_p lazily and last. 5720 57212018-11-07 Jan Hubicka <jh@suse.cz> 5722 5723 * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when 5724 building incomplete variant of complete type. 5725 (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete 5726 variant of complete type. 5727 57282018-11-07 Chenghua Xu <paul.hua.gm@gmail.com> 5729 5730 * config/mips/mips.c: Fix typo in documentation of 5731 mips_loongson_ext2_prefetch_cookie. 5732 (mips_option_override): fix brain twister logical. 5733 * config/mips/mips.h: Fix typo in documentation of 5734 ISA_HAS_CTZ_CTO and define pattern. 5735 * config/mips/mips.md (prefetch): Hoist EXT2 above 5736 the 2EF/EXT block. 5737 (prefetch_indexed): Hoist EXT2 above the EXT block. 5738 57392018-11-07 Jan Hubicka <jh@suse.cz> 5740 5741 * tree.c (free_lang_data_in_type): Add fld parameter; simplify 5742 return and parameter types of function and method types. 5743 (free_lang_data_in_cgraph): Update. 5744 57452018-11-07 Martin Liska <mliska@suse.cz> 5746 5747 PR rtl-optimization/87868 5748 * postreload-gcse.c (eliminate_partially_redundant_load): Set 5749 threshold to max_count if we would overflow. 5750 * profile-count.h: Make max_count a public constant. 5751 57522018-11-07 Martin Liska <mliska@suse.cz> 5753 5754 * mem-stats.h: Fix GNU coding style. 5755 57562018-11-07 Chenghua Xu <paul.hua.gm@gmail.com> 5757 5758 * config/mips/gs264e.md: New. 5759 * config/mips/mips-cpus.def: Define gs264e. 5760 * config/mips/mips-tables.opt: Regenerate. 5761 * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for 5762 gs264e. 5763 (mips_issue_rate): Add support for gs264e. 5764 (mips_multipass_dfa_lookahead): Likewise. 5765 * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E. 5766 (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e. 5767 (MIPS_ASE_MSA_SPEC): New. 5768 (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa. 5769 (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E. 5770 (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E. 5771 * config/mips/mips.md: Include gs264e.md. 5772 (processor): Add gs264e. 5773 * config/mips/mips.opt (MSA): Use Mask instead of Var. 5774 * doc/invoke.texi: Add gs264e to supported architectures. 5775 57762018-11-07 Chenghua Xu <paul.hua.gm@gmail.com> 5777 5778 * config/mips/gs464e.md: New. 5779 * config/mips/mips-cpus.def: Define gs464e. 5780 * config/mips/mips-tables.opt: Regenerate. 5781 * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for 5782 gs464e. 5783 (mips_issue_rate): Add support for gs464e. 5784 (mips_multipass_dfa_lookahead): Likewise. 5785 (mips_option_override): Enable MMI, EXT and EXT2 for gs464e. 5786 * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E. 5787 (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e. 5788 (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E. 5789 (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E. 5790 * config/mips/mips.md: Include gs464e.md. 5791 (processor): Add gs464e. 5792 * doc/invoke.texi: Add gs464e to supported architectures. 5793 57942018-11-07 Chenghua Xu <paul.hua.gm@gmail.com> 5795 5796 * config/mips/loongson3a.md: Rename to ... 5797 * config/mips/gs464.md: ... here. 5798 * config/mips/mips-cpus.def: Define gs464; Add loongson3a 5799 as an alias of gs464 processor. 5800 * config/mips/mips-tables.opt: Regenerate. 5801 * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464 5802 instead of PROCESSOR_LOONGSON_3A. 5803 (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of 5804 TUNE_LOONGSON_3A. 5805 (mips_option_override): Enable MMI and EXT for gs464. 5806 * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464; 5807 Rename TUNE_LOONGSON_3A to TUNE_GS464. 5808 (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464. 5809 (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4, 5810 ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of 5811 TARGET_LOONGSON_3A. 5812 * config/mips/mips.md: Include gs464.md instead of loongson3a.md. 5813 (processor): Add gs464; 5814 * doc/invoke.texi: Add gs464 to supported architectures. 5815 58162018-11-07 Chenghua Xu <paul.hua.gm@gmail.com> 5817 5818 * config/mips/mips-protos.h 5819 (mips_loongson_ext2_prefetch_cookie): New prototype. 5820 * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New. 5821 (mips_option_override): Enable TARGET_LOONGSON_EXT when 5822 TARGET_LOONGSON_EXT2 is true. 5823 * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define 5824 __mips_loongson_ext2, __mips_loongson_ext_rev=2. 5825 (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2. 5826 (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and 5827 TARGET_LOONGSON_EXT2. 5828 (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2. 5829 (define_insn "ctz<mode>2"): New insn pattern. 5830 (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2. 5831 (define_insn "prefetch_indexed_<mode>"): Include 5832 TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2. 5833 * config/mips/mips.opt (-mloongson-ext2): Add option. 5834 * gcc/doc/invoke.texi (-mloongson-ext2): Document. 5835 58362018-11-07 Chenghua Xu <paul.hua.gm@gmail.com> 5837 5838 * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add 5839 __mips_loongson_ext. 5840 (MIPS_ASE_LOONGSON_EXT_SPEC): New. 5841 (BASE_DRIVER_SELF_SPECS): march=loongson3a implies 5842 -mloongson-ext. 5843 (ASM_SPEC): Add mloongson-ext and mno-loongson-ext. 5844 * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo, 5845 <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT 5846 instead of TARGET_LOONGSON_3A. 5847 * config/mips/mips.opt (-mloongson-ext): Add option. 5848 * gcc/doc/invoke.texi (-mloongson-ext): Document. 5849 58502018-11-07 Chenghua Xu <paul.hua.gm@gmail.com> 5851 5852 * config.gcc (extra_headers): Add loongson-mmiintrin.h. 5853 * config/mips/loongson.md: Move to ... 5854 * config/mips/loongson-mmi.md: here; Adjustment. 5855 * config/mips/loongson.h: Move to ... 5856 State as deprecated. Include loongson-mmiintrin.h for back 5857 compatibility and warning. 5858 * config/mips/loongson-mmiintrin.h: ... here. 5859 * config/mips/mips.c (mips_hard_regno_mode_ok_uncached, 5860 mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use 5861 TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS. 5862 (mips_option_override): Make sure MMI use hard float; 5863 (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd, 5864 mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast, 5865 mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of 5866 TARGET_LOONGSON_VECTORS. 5867 * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete. 5868 (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi. 5869 (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New. 5870 (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies 5871 -mloongson-mmi. 5872 (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of 5873 TARGET_LOONGSON_VECTORS. 5874 * gcc/config/mips/mips.md (MOVE64, MOVE128): Use 5875 TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS. 5876 (Loongson MMI patterns): Include loongson-mmi.md instead of 5877 loongson.md. 5878 * gcc/config/mips/mips.opt (-mloongson-mmi): New option. 5879 * gcc/doc/invoke.texi (-mloongson-mmi): Document. 5880 58812018-11-07 Richard Biener <rguenther@suse.de> 5882 5883 PR lto/87906 5884 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup 5885 BLOCK_ABSTRACT_ORIGIN to be the ultimate origin. 5886 58872018-11-07 Alexandre Oliva <aoliva@redhat.com> 5888 5889 PR rtl-optimization/87874 5890 * lra.c (lra_substitute_pseudo): Do not create a subreg for 5891 const wide ints. 5892 58932018-11-06 Aaron Sawdey <acsawdey@linux.ibm.com> 5894 5895 * config/rs6000/rs6000.md (bswap<mode>2): Force address into register 5896 if not in indexed or indirect form. 5897 (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand. 5898 (bswap<mode>2_store): Ditto. 5899 59002018-11-06 Richard Earnshaw <rearnsha@arm.com> 5901 5902 * config/aarch64/aarch64.md (speculation_tracker): Set the mode for 5903 the UNSPEC. 5904 59052018-11-06 Richard Biener <rguenther@suse.de> 5906 5907 PR tree-optimization/86850 5908 * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length () 5909 instead of src.m_vec. 5910 59112018-11-06 Jan Hubicka <jh@suse.cz> 5912 5913 * tree.c (fld_simplified_type_name): Break out form ... 5914 (free_lang_data_in_type): ... here. 5915 (fld_type_variant_equal_p): Use it. 5916 59172018-11-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE> 5918 5919 * config/default-d.c: Include memmodel.h. 5920 5921 * config/sol2-d.c: New file. 5922 * config/t-sol2 (sol2-d.o): New rule. 5923 * config.gcc <*-*-solaris2*>: Set d_target_objs, 5924 target_has_targetdm. 5925 59262018-11-06 Jan Hubicka <jh@suse.cz> 5927 5928 * tree.c (fld_type_variant): Also copy alignment; be sure that 5929 new variant is equal. 5930 59312018-11-06 Ilya Leoshkevich <iii@linux.ibm.com> 5932 5933 PR target/87762 5934 * config/s390/s390.md: Add relative_long attribute. 5935 59362018-11-06 Jan Hubicka <jh@suse.cz> 5937 5938 * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING. 5939 * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING. 5940 * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not 5941 stream TYPE_NEEDS_CONSTRUCTING. 5942 * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise. 5943 * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING. 5944 59452018-11-06 Richard Biener <rguenther@suse.de> 5946 5947 * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb 5948 dump-scope ... 5949 (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations. 5950 59512018-11-06 Jozef Lawrynowicz <jozef.l@mittosystems.com> 5952 5953 * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to 5954 REG_CLASS_CONTENTS[GEN_REGS]. 5955 (REGNO_REG_CLASS): Return NO_REGS for R2 and R3. 5956 59572018-11-06 Jan Hubicka <jh@suse.cz> 5958 5959 * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag. 5960 59612018-11-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE> 5962 5963 PR sanitizer/80953 5964 * config/sol2.h (ASAN_CC1_SPEC): Define. 5965 (LD_WHOLE_ARCHIVE_OPTION): Define. 5966 (LD_NO_WHOLE_ARCHIVE_OPTION): Define. 5967 (ASAN_REJECT_SPEC): Provide default. 5968 (LIBASAN_EARLY_SPEC): Define. 5969 (LIBTSAN_EARLY_SPEC): Define. 5970 (LIBLSAN_EARLY_SPEC): Define. 5971 * config/i386/sol2.h (CC1_SPEC): Redefine. 5972 (ASAN_REJECT_SPEC): Define. 5973 5974 * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare. 5975 (TARGET_ASAN_SHADOW_OFFSET): Define. 5976 (sparc_asan_shadow_offset): New function. 5977 * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC. 5978 (ASAN_REJECT_SPEC): Define. 5979 59802018-11-06 Jan Hubicka <jh@suse.cz> 5981 5982 * tree.c (fld_type_variant): Copy canonical type. 5983 (fld_incomplete_type_of): Check that canonical types looks sane; 5984 copy canonical type. 5985 (verify_type): Accept when incomplete type has complete canonical type. 5986 59872018-11-06 Jan Hubicka <jh@suse.cz> 5988 5989 * tree.c (free_lang_data): Reset overwite_assembler_name, 5990 print_xnode, print_decl, print_type and print_identifier of 5991 langhooks. 5992 59932018-11-06 Richard Biener <rguenther@suse.de> 5994 5995 PR tree-optimization/87889 5996 * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges): 5997 Do nothing if old and new arg are the same 5998 59992018-11-06 Andreas Krebbel <krebbel@linux.ibm.com> 6000 6001 PR target/87723 6002 * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode 6003 attributes for operands 3 and 4. 6004 60052018-11-06 Richard Biener <rguenther@suse.de> 6006 6007 PR middle-end/18041 6008 * simplify-rtx.c (simplify_binary_operation_1): Add pattern 6009 matching bitfield insertion. 6010 60112018-11-06 Alexandre Oliva <aoliva@redhat.com> 6012 6013 * auto-inc-dec.c: Include valtrack.h. Improve comments. 6014 (reg_next_debug_use): New. 6015 (attempt_change): Propagate adjusted expression into affected 6016 debug insns. 6017 (merge_in_block): Track uses in debug insns. 6018 (pass_inc_dec::execute): Allocate and release 6019 reg_next_debug_use. 6020 60212018-11-06 Wei Xiao <wei3.xiao@intel.com> 6022 6023 * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics. 6024 (_mm512_fixupimm_round_pd): Update parameters and builtin. 6025 (_mm512_maskz_fixupimm_round_pd): Ditto. 6026 (_mm512_fixupimm_round_ps): Ditto. 6027 (_mm512_maskz_fixupimm_round_ps): Ditto. 6028 (_mm_fixupimm_round_sd): Ditto. 6029 (_mm_maskz_fixupimm_round_sd): Ditto. 6030 (_mm_fixupimm_round_ss): Ditto. 6031 (_mm_maskz_fixupimm_round_ss): Ditto. 6032 (_mm512_fixupimm_pd): Ditto. 6033 (_mm512_maskz_fixupimm_pd): Ditto. 6034 (_mm512_fixupimm_ps): Ditto. 6035 (_mm512_maskz_fixupimm_ps): Ditto. 6036 (_mm_fixupimm_sd): Ditto. 6037 (_mm_maskz_fixupimm_sd): Ditto. 6038 (_mm_fixupimm_ss): Ditto. 6039 (_mm_maskz_fixupimm_ss): Ditto. 6040 (_mm512_mask_fixupimm_round_pd): Update builtin. 6041 (_mm512_mask_fixupimm_round_ps): Ditto. 6042 (_mm_mask_fixupimm_round_sd): Ditto. 6043 (_mm_mask_fixupimm_round_ss): Ditto. 6044 (_mm512_mask_fixupimm_pd): Ditto. 6045 (_mm512_mask_fixupimm_ps): Ditto. 6046 (_mm_mask_fixupimm_sd): Ditto. 6047 (_mm_mask_fixupimm_ss): Ditto. 6048 * config/i386/avx512vlintrin.h: 6049 (_mm256_fixupimm_pd): Update parameters and builtin. 6050 (_mm256_maskz_fixupimm_pd): Ditto. 6051 (_mm256_fixupimm_ps): Ditto. 6052 (_mm256_maskz_fixupimm_ps): Ditto. 6053 (_mm_fixupimm_pd): Ditto. 6054 (_mm_maskz_fixupimm_pd): Ditto. 6055 (_mm_fixupimm_ps): Ditto. 6056 (_mm_maskz_fixupimm_ps): Ditto. 6057 (_mm256_mask_fixupimm_pd): Update builtin. 6058 (_mm256_mask_fixupimm_ps): Ditto. 6059 (_mm_mask_fixupimm_pd): Ditto. 6060 (_mm_mask_fixupimm_ps): Ditto. 6061 * config/i386/i386-builtin-types.def: Add new types and remove useless ones. 6062 * config/i386/i386-builtin.def: Update builtin definitions. 6063 * config/i386/i386.c: Handle new builtin types and remove useless ones. 6064 * config/i386/sse.md: Update VFIXUPIMM* patterns. 6065 (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update. 6066 (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update. 6067 (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update. 6068 (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update. 6069 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update. 6070 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update. 6071 * config/i386/subst.md: 6072 (round_saeonly_sd_mask_operand4): Add new subst_attr. 6073 (round_saeonly_sd_mask_op4): Ditto. 6074 (round_saeonly_expand_operand5): Ditto. 6075 (round_saeonly_expand): Update. 6076 60772018-11-05 Max Filippov <jcmvbkbc@gmail.com> 6078 6079 * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0. 6080 60812018-11-05 Segher Boessenkool <segher@kernel.crashing.org> 6082 6083 PR rtl-optimization/87871 6084 * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx. 6085 60862018-11-05 Paul Koning <ni1d@arrl.net> 6087 6088 * doc/sourcebuild.texi (target attributes): Document new "inf" 6089 effective target keyword. 6090 60912018-11-05 Robin Dapp <rdapp@linux.ibm.com> 6092 6093 * config/s390/s390.c (s390_register_move_cost): Increase costs for 6094 moves involving the CC reg. 6095 60962018-11-05 Richard Biener <rguenther@suse.de> 6097 6098 PR tree-optimization/87873 6099 * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p 6100 argument. 6101 * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise. 6102 * tree-vect-loop.c (vect_transform_loop): When splitting the 6103 loop exit also create forwarder PHIs for constants. 6104 * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges): 6105 Handle constant to_arg, add extra checking we match up the correct 6106 PHIs. 6107 61082018-11-05 Robin Dapp <rdapp@linux.ibm.com> 6109 6110 * config/s390/s390.md: QImode and HImode for load on condition. 6111 61122018-11-05 Robin Dapp <rdapp@linux.ibm.com> 6113 6114 * config/s390/predicates.md: Fix typo. 6115 * config/s390/s390.md: Allow immediates for load on condition. 6116 61172018-11-05 Martin Liska <mliska@suse.cz> 6118 6119 * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT. 6120 * bitmap.h (struct bitmap_usage): Likewise. 6121 * ggc-common.c (SCALE): Remove. 6122 (LABEL): Likewise. 6123 (struct ggc_usage): Use SIZE_AMOUNT. And update 6124 compare method. 6125 * ggc-page.c (SCALE): Remove. 6126 (STAT_LABEL): Remove. 6127 (ggc_print_statistics): Use SIZE_AMOUNT. 6128 * gimple.h (SCALE): Remove. 6129 (LABEL): Likewise. 6130 * input.c (ONE_K): Remove. 6131 (ONE_M): Likewise. 6132 (SCALE): Likewise. 6133 (STAT_LABEL): Likewise. 6134 (FORMAT_AMOUNT): Likewise. 6135 (dump_line_table_statistics): Use SIZE_AMOUNT. 6136 * mem-stats.h (struct mem_usage): Likewise. 6137 * rtl.c (dump_rtx_statistics): Likewise. 6138 (rtx_alloc_counts): Change type to size_t. 6139 (rtx_alloc_sizes): Likewise. 6140 (rtx_count_cmp): New. 6141 (dump_rtx_statistics): Sort first based on counts. 6142 * tree.c (tree_nodes_cmp): New. 6143 (tree_codes_cmp): New. 6144 (dump_tree_statistics): Sort first based on counts. 6145 * system.h (ONE_K): New. 6146 (ONE_M): Likewise. 6147 (SIZE_SCALE): Likewise. 6148 (SIZE_LABEL): Likewise. 6149 (SIZE_AMOUNT): Likewise. 6150 * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT. 6151 * tree-dfa.c (dump_dfa_stats): Likewise. 6152 * tree-phinodes.c (phinodes_print_statistics): Likewise. 6153 * tree-ssanames.c (ssanames_print_statistics): Likewise. 6154 * tree.c (dump_tree_statistics): Likewise. 6155 * vec.c (struct vec_usage): Likewise. 6156 * trans-mem.c (tm_mangle): Enlarge buffer in order to not 6157 trigger a -Werror=format-overflow with 6158 --enable-gather-detailed-stats. 6159 61602018-11-05 Martin Liska <mliska@suse.cz> 6161 6162 * mem-stats.h (mem_alloc_description::release_instance_overhead): 6163 Return T *. 6164 * vec.c (struct vec_usage): Register m_element_size. 6165 (vec_prefix::register_overhead): New arguments: elements and 6166 element_size. 6167 (vec_prefix::release_overhead): Subtract elements. 6168 * vec.h (struct vec_prefix): Change signature. 6169 (va_heap::reserve): Pass proper arguments. 6170 (va_heap::release): Likewise. 6171 61722018-11-05 Martin Liska <mliska@suse.cz> 6173 6174 * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding 6175 style. 6176 * vec.c: Likewise. 6177 61782018-11-05 Richard Biener <rguenther@suse.de> 6179 6180 * tree-scalar-evolution.h (final_value_replacement_loop): Update 6181 prototype. 6182 * tree-scalar-evolution.c (final_value_replacement_loop): Return 6183 whether anything was done. 6184 (scev_const_prop): Remove constant propagation part, fold 6185 remains into ... 6186 * tree-ssa-loop.c (pass_scev_cprop::execute): ... here. 6187 (pass_data_scev_cprop): TODO_cleanup_cfg is now done 6188 conditionally. 6189 61902018-11-05 Jakub Jelinek <jakub@redhat.com> 6191 6192 PR tree-optimization/87859 6193 * gimple-ssa-store-merging.c (struct merged_store_group): Add 6194 only_constants and first_nonmergeable_order members. 6195 (merged_store_group::merged_store_group): Initialize them. 6196 (merged_store_group::do_merge): Clear only_constants member if 6197 adding something other than INTEGER_CST store. 6198 (imm_store_chain_info::coalesce_immediate_stores): Don't merge 6199 stores with order >= first_nonmergeable_order. Use 6200 merged_store->only_constants instead of always recomputing it. 6201 Set merged_store->first_nonmergeable_order if we've skipped any 6202 stores. Attempt to merge overlapping INTEGER_CST stores that 6203 we would otherwise skip. 6204 6205 PR sanitizer/87837 6206 * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED. 6207 62082018-11-05 Xuepeng Guo <xuepeng.guo@intel.com> 6209 6210 PR target/87853 6211 * config/i386/emmintrin.h (__v16qs): New to cope with option 6212 -funsigned-char. 6213 (_mm_cmpeq_epi8): Replace __v16qi with __v16qs. 6214 (_mm_cmplt_epi8): Likewise. 6215 (_mm_cmpgt_epi8): Likewise. 6216 62172018-11-05 Richard Biener <rguenther@suse.de> 6218 6219 PR rtl-optimization/87852 6220 * fwprop.c (use_killed_between): Only consider single-defs of the 6221 use whose definition statement dominates the use. 6222 62232018-11-05 Martin Liska <mliska@suse.cz> 6224 6225 PR web/87829 6226 * doc/invoke.texi: Remove options that are 6227 not disabled with -Os. 6228 62292018-11-05 Martin Liska <mliska@suse.cz> 6230 6231 PR c/87811 6232 * doc/extend.texi: Update constrain about the last argument 6233 of __builtin_expect_with_probability. 6234 62352018-11-05 Martin Liska <mliska@suse.cz> 6236 6237 PR c/87811 6238 * predict.c (expr_expected_value_1): Verify 6239 that last argument is a real constants and emit 6240 error. 6241 62422018-11-05 Martin Liska <mliska@suse.cz> 6243 6244 PR gcov-profile/77698 6245 * ipa-profile.c (ipa_profile): Adjust hotness threshold 6246 only in LTO mode. 6247 62482018-11-04 Bernd Edlinger <bernd.edlinger@hotmail.de> 6249 6250 PR tree-optimization/86572 6251 * builtins.c (c_strlen): Handle negative offsets in a safe way. 6252 62532018-11-04 Bernd Edlinger <bernd.edlinger@hotmail.de> 6254 6255 PR tree-optimization/87672 6256 * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify. 6257 * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size. 6258 62592018-11-04 Uros Bizjak <ubizjak@gmail.com> 6260 6261 PR middle-end/58372 6262 * cfgexpand.c (pass_expand::execute): Move the call to 6263 finish_eh_generation in front of the call to expand_stack_alignment. 6264 62652018-11-04 Venkataramanan Kumar <venkataramanan.kumar@amd.com> 6266 6267 * common/config/i386/i386-common.c (processor_alias_table): Add 6268 znver2 entry. 6269 * config.gcc (i[34567]86-*-linux* | ...): Add znver2. 6270 (case ${target}): Add znver2. 6271 * config/i386/driver-i386.c: (host_detect_local_cpu): Let 6272 -march=native recognize znver2 processors. 6273 * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2. 6274 * config/i386/i386.c (m_znver2): New definition. 6275 (m_ZNVER): New definition. 6276 (m_AMD_MULTIPLE): Includes m_znver2. 6277 (processor_cost_table): Add znver2 entry. 6278 (processor_target_table): Add znver2 entry. 6279 (get_builtin_code_for_version): Set priority for 6280 PROCESSOR_ZNVER2. 6281 (processor_model): Add M_AMDFAM17H_ZNVER2. 6282 (arch_names_table): Ditto. 6283 (ix86_reassociation_width): Include znver2. 6284 * config/i386/i386.h (TARGET_znver2): New definition. 6285 (struct ix86_size_cost): Add TARGET_ZNVER2. 6286 (enum processor_type): Add PROCESSOR_ZNVER2. 6287 * config/i386/i386.md (define_attr "cpu"): Add znver2. 6288 * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs. 6289 * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2. 6290 (ix86_adjust_cost): Add znver2. 6291 * config/i386/x86-tune.def: Replace m_ZNVER1 by m_ZNVER. 6292 * gcc/doc/extend.texi: Add details about znver2. 6293 * gcc/doc/invoke.texi: Add details about znver2. 6294 62952018-11-03 Sandra Loosemore <sandra@codesourcery.com> 6296 6297 PR target/87079 6298 6299 * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3 6300 pattern. 6301 63022018-11-02 Aaron Sawdey <acsawdey@linux.ibm.com> 6303 6304 * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay 6305 attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN. 6306 63072018-11-02 Richard Earnshaw <rearnsha@arm.com> 6308 6309 * config/aarch64/aarch64.c ((aarch64_override_options): Disable 6310 shrink-wrapping when -mtrack-speculation. 6311 63122018-11-02 Richard Biener <rguenther@suse.de> 6313 6314 * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member. 6315 (pop_cost_one_pair): Do not free pair. 6316 (pop_best_coalesce): Likewise. 6317 (create_coalesce_list): Initialize obstack. 6318 (delete_coalesce_list): Free obstack. 6319 (find_coalesce_pair): Obstack-allocate coalesce pairs. 6320 (add_cost_one_coalesce): Likewise. 6321 (struct live_track): Remove bitmap pointer indirections. 6322 (new_live_track): Adjust. 6323 (delete_live_track): Likewise. 6324 (live_track_remove_partition): Likewise. 6325 (live_track_add_partition): Likewise. 6326 (live_track_live_p): Likewise. 6327 (live_track_process_def): Likewise. 6328 (live_track_clear_base_vars): Likewise. 6329 63302018-11-02 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE> 6331 6332 * configure.ac (gcc_cv_as_sparc_register_op): Remove. 6333 * configure: Regenerate. 6334 * config.in: Regenerate. 6335 * config/sparc/sparc.c (sparc_output_scratch_registers): Remove 6336 HAVE_AS_REGISTER_PSEUDO_OP guard. 6337 * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove 6338 !HAVE_AS_REGISTER_PSEUDO_OP support. 6339 (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP 6340 guard. 6341 63422018-11-02 Richard Biener <rguenther@suse.de> 6343 6344 * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call 6345 with is_type_die. 6346 63472018-11-02 Richard Biener <rguenther@suse.de> 6348 6349 PR tree-optimization/87776 6350 * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges 6351 executable when iterating but running into rpo-vn-max-loop-depth 6352 and not eliding the iteration. 6353 63542018-11-30 Jan Hubicka <jh@suse.cz> 6355 6356 * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of 6357 TYPE_DECL. 6358 63592018-11-01 Aaron Sawdey <acsawdey@linux.ibm.com> 6360 6361 * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove 6362 prototype. 6363 * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): 6364 Combine with rs6000_address_for_fpconvert. 6365 (rs6000_address_for_fpconvert): Combine with 6366 rs6000_force_indexed_or_indirect_mem. 6367 (rs6000_expand_vector_init): Change function call from 6368 rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem. 6369 * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from 6370 rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem. 6371 (floatsi<mode>2_lfiwax_mem): Ditto. 6372 (floatunssi<mode>2_lfiwzx): Ditto. 6373 (floatunssi<mode>2_lfiwzx_mem): Ditto. 6374 (float<QHI:mode><FP_ISA3:mode>2): Ditto. 6375 (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto. 6376 (fix_trunc<mode>si2_stfiwx): Ditto. 6377 (fixuns_trunc<mode>si2_stfiwx): Ditto. 6378 (float_<mode>si2_hw): Ditto. 6379 (floatuns_<mode>si2_hw): Ditto. 6380 * config/rs6000/vsx.md (*vsx_extract_si): Ditto. 6381 (vsx_splat_<mode>): Ditto. 6382 63832018-11-01 Joseph Myers <joseph@codesourcery.com> 6384 6385 * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline 6386 at end of assembler input text. 6387 * configure: Regenerate. 6388 63892018-11-01 Jakub Jelinek <jakub@redhat.com> 6390 6391 PR tree-optimization/87826 6392 * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is 6393 negative or larger or equal to type's precision. 6394 63952018-10-31 Alexandre Oliva <aoliva@redhat.com> 6396 6397 * opts.c (default_options_table): Do not enable 6398 OPT_fdelayed_branch at -Og. 6399 * doc/invoke.texi (-fdelayed-branch): Document it. 6400 64012018-10-31 Richard Henderson <richard.henderson@linaro.org> 6402 6403 * optabs-libfuncs.c (build_libfunc_function_visibility): 6404 New, split out from... 6405 (build_libfunc_function): ... here. 6406 (init_one_libfunc_visibility): New, split out from ... 6407 (init_one_libfunc): ... here. 6408 6409 * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): 6410 scratch register need not be early-clobber. Document the reason 6411 why we cannot use ST<OP>. 6412 64132018-10-31 Joseph Myers <joseph@codesourcery.com> 6414 6415 PR bootstrap/82856 6416 * configure.ac: Remove AC_PREREQ. Use AC_LANG_SOURCE. Use single 6417 line for second argument of AC_DEFINE_UNQUOTED. 6418 * doc/install.texi (Tools/packages necessary for modifying GCC): 6419 Update to autoconf 2.69 and automake 1.15.1. 6420 * aclocal.m4, config.in, configure: Regenerate. 6421 64222018-10-31 Pat Haugen <pthaugen@us.ibm.com> 6423 6424 * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead 6425 initialization. 6426 64272018-10-31 Martin Liska <mliska@suse.cz> 6428 6429 PR driver/83193 6430 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension): 6431 Add new argument invalid_extension. 6432 (aarch64_get_all_extension_candidates): New function. 6433 (aarch64_rewrite_selected_cpu): Add NULL to function call. 6434 * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add 6435 new argument. 6436 (aarch64_get_all_extension_candidates): New function. 6437 * config/aarch64/aarch64.c (aarch64_parse_arch): Add new 6438 argument invalid_extension. 6439 (aarch64_parse_cpu): Likewise. 6440 (aarch64_print_hint_for_extensions): New function. 6441 (aarch64_validate_mcpu): Provide hint about invalid extension. 6442 (aarch64_validate_march): Likewise. 6443 (aarch64_handle_attr_arch): Pass new argument. 6444 (aarch64_handle_attr_cpu): Provide hint about invalid extension. 6445 (aarch64_handle_attr_isa_flags): Likewise. 6446 64472018-10-31 Richard Biener <rguenther@suse.de> 6448 6449 PR middle-end/70359 6450 PR middle-end/86270 6451 * tree-outof-ssa.c (insert_backedge_copies): Restrict 6452 copy generation to useful cases. Place the copy before 6453 the definition of the backedge value when possible. 6454 64552018-10-31 Claudiu Zissulescu <claziss@synopsys.com> 6456 6457 * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value. 6458 * config/arc/arc.c (arc_active_insn): New function. 6459 (check_store_cacheline_hazard): Likewise. 6460 (workaround_arc_anomaly): Use check_store_cacheline_hazard. 6461 (arc_override_options): Disable delay slot scheduler for older 6462 A7. 6463 (arc_store_addr_hazard_p): New implementation, old one renamed to 6464 ... 6465 (arc_store_addr_hazard_internal_p): Renamed. 6466 (arc_reorg): Don't combine into brcc instructions which are part 6467 of hardware hazard solution. 6468 * config/arc/arc.md (attr tune): Consider new arc7xx tune value. 6469 (tune_arc700): Likewise. 6470 * config/arc/arc.opt (arc7xx): New tune value. 6471 * config/arc/arc700.md: Improve A7 scheduler. 6472 64732018-10-31 Claudiu Zissulescu <claziss@synopsys.com> 6474 6475 * config/arc/arc.c (arc_override_options): Remove 6476 TARGET_COMPACT_CASESI. 6477 * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update. 6478 (CASE_VECTOR_MODE): Likewise. 6479 (CASE_VECTOR_PC_RELATIVE): Likewise. 6480 (CASE_VECTOR_SHORTEN_MODE): Likewise. 6481 (CASE_VECTOR_SHORTEN_MODE1): Delete. 6482 (ADDR_VEC_ALIGN): Update. 6483 (ASM_OUTPUT_CASE_LABEL): Undefine. 6484 (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine. 6485 (TARGET_BI_BIH): Define. 6486 (DEFAULT_BRANCH_INDEX): Likewise. 6487 * config/arc/arc.md (casesi): Rework to accept BI/BIH 6488 instructions, remove compact_casesi use case. 6489 (casesi_compact_jump): Remove. 6490 (casesi_dispatch): New pattern. 6491 * config/arc/arc.opt: Add mbranch-index option. Deprecate 6492 compact_casesi option. 6493 * doc/invoke.texi: Document mbranch-index option. 6494 64952018-10-31 Claudiu Zissulescu <claziss@synopsys.com> 6496 6497 * config/arc/arc.c (arc_get_tp): Remove function. 6498 (arc_emit_call_tls_get_addr): Likewise. 6499 (arc_call_tls_get_addr): New function. 6500 (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr. 6501 * config/arc/arc.md (tls_load_tp_soft): Remove. 6502 (tls_gd_get_addr): Likewise. 6503 65042018-10-31 Claudiu Zissulescu <claziss@synopsys.com> 6505 6506 * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib. 6507 (mulsi3_600_lib): Remove pattern. 6508 (umulsi3_highpart_600_lib_le): Likewise. 6509 (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le. 6510 (umulsidi3): Remove call to umulsidi3_600_lib. 6511 (umulsidi3_600_lib): Remove pattern. 6512 (peephole2): Remove peephole using the above deprecated patterns. 6513 65142018-10-31 Thomas Preud'homme <thomas.preudhomme@linaro.org> 6515 6516 PR target/87374 6517 * config/arm/arm.c (arm_option_check_internal): Disable the combined 6518 use of -mslow-flash-data and -mword-relocations. 6519 (arm_option_override): Enable -mword-relocations if -fpic or -fPIC. 6520 * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for 6521 flag_pic. 6522 * doc/invoke.texi (-mword-relocations): Mention conflict with 6523 -mslow-flash-data. 6524 (-mslow-flash-data): Reciprocally. 6525 65262018-10-31 Richard Henderson <richard.henderson@linaro.org> 6527 6528 * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force 6529 16-byte modes held in GP registers to use an even regno. 6530 6531 * config/aarch64/aarch64.c (aarch64_emit_bic): Remove. 6532 (aarch64_atomic_ldop_supported_p): Remove. 6533 (aarch64_gen_atomic_ldop): Remove. 6534 * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>): 6535 Fully expand LSE operations here. 6536 (atomic_fetch_<atomic_optab><ALLI>): Likewise. 6537 (atomic_<atomic_optab>_fetch<ALLI>): Likewise. 6538 (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator 6539 and use ATOMIC_LDOP instead; use register_operand for the input; 6540 drop the split and emit insns directly. 6541 (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise. 6542 (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove. 6543 (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove. 6544 6545 * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove. 6546 (aarch64_gen_atomic_ldop): Don't call it. 6547 * config/aarch64/atomics.md (atomic_exchange<ALLI>): 6548 Use aarch64_reg_or_zero. 6549 (aarch64_atomic_exchange<ALLI>): Likewise. 6550 (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from 6551 operand 0; use aarch64_reg_or_zero for input; merge ... 6552 (@aarch64_atomic_swp<ALLI>): ... this and remove. 6553 6554 * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New. 6555 (aarch64_split_compare_and_swap): Use it. 6556 (aarch64_expand_compare_and_swap): Likewise. Remove convert_modes; 6557 test oldval against the proper predicate. 6558 * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>): 6559 Use nonmemory_operand for expected. 6560 (cas_short_expected_pred): New. 6561 (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match. 6562 (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected. 6563 * config/aarch64/predicates.md (aarch64_plushi_immediate): New. 6564 (aarch64_plushi_operand): New. 6565 6566 * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap): 6567 Force oldval into the rval register for TARGET_LSE; emit the compare 6568 during initial expansion so that it may be deleted if unused. 6569 (aarch64_gen_atomic_cas): Remove. 6570 * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse): 6571 Change =&r to +r for operand 0; use match_dup for operand 2; 6572 remove is_weak and mod_f operands as unused. Drop the split 6573 and merge with... 6574 (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove. 6575 (@aarch64_compare_and_swap<GPI>_lse): Similarly. 6576 (@aarch64_atomic_cas<GPI>): Similarly. 6577 65782018-10-31 Richard Biener <rguenther@suse.de> 6579 6580 * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case 6581 using ABSU_EXPR. 6582 65832018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org> 6584 6585 * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline. 6586 * config/aarch64/aarch64.md: Include saphira.md 6587 * config/aarch64/saphira.md: New file for pipeline description. 6588 65892018-10-30 Martin Sebor <msebor@redhat.com> 6590 6591 PR middle-end/87041 6592 * gimple-ssa-sprintf.c (format_directive): Use %G to include 6593 inlining context. 6594 (sprintf_dom_walker::compute_format_length): 6595 Avoid setting POSUNDER4K here. 6596 (get_destination_size): Handle null argument values. 6597 (get_user_idx_format): New function. 6598 (sprintf_dom_walker::handle_gimple_call): Handle all printf-like 6599 functions, including user-defined with attribute format printf. 6600 Use %G to include inlining context. 6601 Set POSUNDER4K here. 6602 66032018-10-30 Jan Hubicka <jh@suse.cz> 6604 6605 * params.def (lto-partitions): Bump from 32 to 128. 6606 66072018-10-30 Jan Hubicka <jh@suse.cz> 6608 6609 * tree.c 6610 (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move 6611 head in file. 6612 (free_lang_data_in_type): Forward declare. 6613 (fld_type_variant_equal_p): New function. 6614 (fld_type_variant): New function 6615 (fld_incomplete_types): New hash. 6616 (fld_incomplete_type_of): New function 6617 (fld_simplfied-type): New function. 6618 (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL 6619 (free_lang_data): Allocate and free fld_incomplete_type; update call 6620 of free_lang_data_in_decl. 6621 66222018-10-30 Eric Botcazou <ebotcazou@adacore.com> 6623 6624 * gcov.c (output_lines): Remove duplicate line. 6625 66262018-10-30 Aaron Sawdey <acsawdey@linux.ibm.com> 6627 6628 * config/rs6000/rs6000.md (bswapdi2): Force address into register 6629 if not in indexed or indirect form. 6630 (bswapdi2_load): Change predicate to indexed_or_indirect_operand. 6631 (bswapdi2_store): Ditto. 6632 * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New 6633 helper function. 6634 * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem): 6635 Prototype for helper function. 6636 66372018-10-30 Martin Sebor <msebor@redhat.com> 6638 6639 * doc/extend.texi (optimize): Clarify/expand attribute documentation. 6640 (target, pragma GCC optimize, pragma GCC target): Ditto. 6641 66422018-10-30 Jonathan Wakely <jwakely@redhat.com> 6643 6644 * doc/extend.texi: Fix prototype and description of 6645 __builtin_expect_with_probability. 6646 66472018-10-30 Michael Ploujnikov <michael.ploujnikov@oracle.com> 6648 6649 * cgraph.h (clone_function_name_1): Replaced by new 6650 clone_function_name_numbered that takes name as string; for 6651 privatize_symbol_name_1 use only. 6652 (clone_function_name): Renamed to 6653 clone_function_name_numbered to be explicit about numbering. 6654 (clone_function_name): New two-argument function that does 6655 not number its output. 6656 (clone_function_name): New three-argument function that 6657 takes a number to append to its output. 6658 * cgraphclones.c (duplicate_thunk_for_node): 6659 (clone_function_name_1): Renamed. 6660 (clone_function_name_numbered): Two new functions. 6661 (clone_function_name): Improved documentation. 6662 (cgraph_node::create_virtual_clone): Use clone_function_name_numbered. 6663 * config/rs6000/rs6000.c (make_resolver_func): Ditto. 6664 * final.c (final_scan_insn_1): Use the new clone_function_name 6665 without numbering. 6666 * multiple_target.c (create_dispatcher_calls): Ditto. 6667 (create_target_clone): Ditto. 6668 * omp-expand.c (grid_expand_target_grid_body): Ditto. 6669 * omp-low.c (create_omp_child_function_name): Ditto. 6670 * omp-simd-clone.c (simd_clone_create): Ditto. 6671 * symtab.c (simd_symtab_node::noninterposable_alias): Use the 6672 new clone_function_name without numbering. 6673 66742018-10-30 Richard Earnshaw <rearnsha@arm.com> 6675 6676 * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize): 6677 Assert that the allocation size is not zero. 6678 66792018-10-30 Richard Biener <rguenther@suse.de> 6680 6681 PR tree-optimization/87800 6682 * tree-vect-slp.c (vect_build_slp_tree_2): Reject any 6683 non-induction or reduction PHIs. 6684 66852018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org> 6686 6687 * config/aarch64/falkor-tag-collision-avoidance.c 6688 (execute_tag_collision_avoidance): Call df_note_add_problem. 6689 66902018-10-30 Martin Liska <mliska@suse.cz> 6691 6692 * doc/extend.texi: Fix typo in documentation 6693 of __builtin_expect_with_probability. 6694 66952018-10-29 David Malcolm <dmalcolm@redhat.com> 6696 6697 PR c++/87721 6698 * input.c (get_substring_ranges_for_loc): Detect if 6699 linemap_resolve_location gives us a NULL map, and reject 6700 this case. 6701 67022018-10-29 Iain Buclaw <ibuclaw@gdcproject.org> 6703 6704 * config.gcc (xstormy16-*-elf): Set tm_d_file. 6705 67062018-10-29 David Malcolm <dmalcolm@redhat.com> 6707 Martin Sebor <msebor@redhat.com> 6708 Manuel Lopez-Ibanez <manu@gcc.gnu.org> 6709 6710 * doc/ux.texi (Quoting): New subsection, adapted from material at 6711 https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by 6712 MartinSebor and ManuelLopezIbanez. 6713 (Fix-it hints): Note that fix-it hints shouldn't be marked for 6714 translation. 6715 67162018-10-29 Kugan Vivekanandarajah <kuganv@linaro.org> 6717 6718 PR middle-end/87469 6719 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter 6720 max value. 6721 67222018-10-29 Paul A. Clarke <pc@us.ibm.com> 6723 6724 * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment. 6725 67262018-10-29 Paul Koning <ni1d@arrl.net> 6727 6728 * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name. 6729 67302018-10-29 Paul A. Clarke <pc@us.ibm.com> 6731 6732 * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32, 6733 _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8, 6734 _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32, 6735 _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32, 6736 _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8, 6737 _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8, 6738 _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16, 6739 _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8): 6740 Change 'vector' to '__vector'. 6741 * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32, 6742 _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16, 6743 _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8, 6744 _mm_avg_pu16): Likewise. And, whitespace corrections. 6745 67462018-10-29 Richard Biener <rguenther@suse.de> 6747 6748 PR tree-optimization/87785 6749 * tree-vect-slp.c (vect_gather_slp_loads): Only gather 6750 internal defs. 6751 67522018-10-29 Olivier Hainque <hainque@adacore.com> 6753 6754 * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP. 6755 67562018-10-29 Olivier Hainque <hainque@adacore.com> 6757 6758 * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks 6759 ports configuration macro, defaults to "gnu". 6760 (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and 6761 TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY. 6762 67632018-10-29 Olivier Hainque <hainque@adacore.com> 6764 6765 * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define. 6766 67672018-10-29 Richard Biener <rguenther@suse.de> 6768 6769 PR tree-optimization/87790 6770 * tree-vect-slp.c (vect_mark_slp_stmts): Simplify. 6771 (vect_make_slp_decision): Adjust. 6772 (vect_slp_analyze_bb_1): Likewise. 6773 (vect_detect_hybrid_slp_stmts): Properly union SLP type over 6774 edges. 6775 67762018-10-29 Richard Biener <rguenther@suse.de> 6777 6778 PR tree-optimization/87785 6779 * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument 6780 and processing. 6781 (vect_build_slp_tree): Likewise. 6782 (vect_gather_slp_loads): New function. 6783 (vect_analyze_slp_instance): Gather loads separately from the 6784 SLP tree build. 6785 67862018-10-29 Martin Liska <mliska@suse.cz> 6787 6788 * Makefile.in: Make dependency to json.o. 6789 * doc/gcov.texi: Document new JSON format, remove 6790 old intermediate format documentation. 6791 * gcov.c (struct function_info): Come up with m_name and 6792 m_demangled_name. 6793 (function_info::function_info): Initialize it. 6794 (function_info::~function_info): Release it. 6795 (main): Rename flag_intermediate_format to flag_json_format. 6796 (print_usage): Describe --json-format. 6797 (process_args): Set flag_json_format. 6798 (output_intermediate_line): Remove. 6799 (output_intermediate_json_line): Likewise. 6800 (get_gcov_intermediate_filename): Return new extension 6801 ".gcov.json.gz". 6802 (output_intermediate_file): Implement JSON emission. 6803 (output_json_intermediate_file): Implement JSON emission. 6804 (generate_results): Use ::get_name for function name. 6805 Handle JSON output file. 6806 (read_graph_file): Use ::get_name instead of cplus_demangle. 6807 (read_count_file): Likewise. 6808 (solve_flow_graph): Likewise. 6809 (add_line_counts): Likewise. 6810 (accumulate_line_counts): Use new flag_json_format. 6811 (output_function_details): Use ::get_name instead of cplus_demangle. 6812 (output_lines): Likewise. 6813 * json.cc (test_writing_literals): Add new tests. 6814 * json.h (class literal): Add new boolean constructor. 6815 68162018-10-29 Segher Boessenkool <segher@kernel.crashing.org> 6817 6818 PR rtl-optimization/87701 6819 PR rtl-optimization/87780 6820 * combine.c (make_more_copies): Rewrite. 6821 68222018-10-28 Kugan Vivekanandarajah <kuganv@linaro.org> 6823 6824 * doc/generic.texi (ABSU_EXPR): Document. 6825 * match.pd (absu(x)*absu(x) -> x*x): Handle. 6826 (absu(absu(X)) -> absu(X)): Likewise. 6827 (absu(-X) -> absu(X)): Likewise. 6828 (absu(X) where X is nonnegative -> X): Likewise. 6829 68302018-10-28 Iain Buclaw <ibuclaw@gdcproject.org> 6831 6832 * Makefile.in (tm_d_file_list, tm_d_include_list): New variables. 6833 (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables. 6834 (tm_d.h, cs-tm_d.h, default-d.o): New rules. 6835 (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules. 6836 (s-tm-texi): Also check timestamp on d-target.def. 6837 (generated_files): Add TM_D_H and d-target-hooks-def.h. 6838 (build/genhooks.o): Also depend on D_TARGET_DEF. 6839 * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New 6840 variables. 6841 * config/aarch64/aarch64-d.c: New file. 6842 * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE): 6843 Define. 6844 * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New 6845 prototype. 6846 * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define. 6847 * config/aarch64/t-aarch64 (aarch64-d.o): New rule. 6848 * config/arm/arm-d.c: New file. 6849 * config/arm/arm-protos.h (arm_d_target_versions): New prototype. 6850 * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define. 6851 * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define. 6852 * config/arm/t-arm (arm-d.o): New rule. 6853 * config/default-d.c: New file. 6854 * config/glibc-d.c: New file. 6855 * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define. 6856 * config/i386/i386-d.c: New file. 6857 * config/i386/i386-protos.h (ix86_d_target_versions): New prototype. 6858 * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define. 6859 * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define. 6860 (GNU_USER_TARGET_D_CRITSEC_SIZE): Define. 6861 * config/i386/t-i386 (i386-d.o): New rule. 6862 * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define. 6863 * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define. 6864 * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define. 6865 * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define. 6866 * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define. 6867 * config/mips/mips-d.c: New file. 6868 * config/mips/mips-protos.h (mips_d_target_versions): New prototype. 6869 * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define. 6870 * config/mips/t-mips (mips-d.o): New rule. 6871 * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define. 6872 * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define. 6873 * config/powerpcspe/powerpcspe-d.c: New file. 6874 * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions): 6875 New prototype. 6876 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue): 6877 Support GNU D by using 0 as the language type. 6878 * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define. 6879 * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule. 6880 * config/riscv/riscv-d.c: New file. 6881 * config/riscv/riscv-protos.h (riscv_d_target_versions): New 6882 prototype. 6883 * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define. 6884 * config/riscv/t-riscv (riscv-d.o): New rule. 6885 * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define. 6886 * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define. 6887 * config/rs6000/rs6000-d.c: New file. 6888 * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New 6889 prototype. 6890 * config/rs6000/rs6000.c (rs6000_output_function_epilogue): 6891 Support GNU D by using 0 as the language type. 6892 * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define. 6893 * config/rs6000/t-rs6000 (rs6000-d.o): New rule. 6894 * config/s390/s390-d.c: New file. 6895 * config/s390/s390-protos.h (s390_d_target_versions): New prototype. 6896 * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define. 6897 * config/s390/t-s390 (s390-d.o): New rule. 6898 * config/sparc/sparc-d.c: New file. 6899 * config/sparc/sparc-protos.h (sparc_d_target_versions): New 6900 prototype. 6901 * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define. 6902 * config/sparc/t-sparc (sparc-d.o): New rule. 6903 * config/t-glibc (glibc-d.o): New rule. 6904 * configure: Regenerated. 6905 * configure.ac (tm_d_file): New variable. 6906 (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes. 6907 * doc/contrib.texi (Contributors): Add self for the D frontend. 6908 * doc/frontends.texi (G++ and GCC): Mention D as a supported language. 6909 * doc/install.texi (Configuration): Mention libphobos as an option for 6910 --enable-shared. Mention d as an option for --enable-languages. 6911 (Testing): Mention check-d as a target. 6912 * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file 6913 name suffixes. Mention d as a -x option. 6914 * doc/sourcebuild.texi (Top Level): Mention libphobos. 6915 * doc/standards.texi (Standards): Add section on D language. 6916 * doc/tm.texi: Regenerated. 6917 * doc/tm.texi.in: Add @node for D language and ABI, and @hook for 6918 TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE. 6919 * dwarf2out.c (is_dlang): New function. 6920 (gen_compile_unit_die): Use DW_LANG_D for D. 6921 (declare_in_namespace): Return module die for D, instead of adding 6922 extra declarations into the namespace. 6923 (gen_namespace_die): Generate DW_TAG_module for D. 6924 (gen_decl_die): Handle CONST_DECLSs for D. 6925 (dwarf2out_decl): Likewise. 6926 (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type. 6927 (prune_unused_types_walk): Handle DW_tag_interface_type same as other 6928 kinds of aggregates. 6929 * gcc.c (default_compilers): Add entries for .d, .dd and .di. 6930 * genhooks.c: Include d/d-target.def. 6931 69322018-10-28 Iain Sandoe <iain@sandoe.co.uk> 6933 6934 PR target/85669 6935 * config/rs6000/darwin.h (STACK_BOUNDARY): New. 6936 (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment. 6937 (STACK_DYNAMIC_OFFSET): Likewise. 6938 69392018-10-27 Sandra Loosemore <sandra@codesourcery.com> 6940 6941 PR target/80024 6942 * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix 6943 error message. 6944 69452018-10-26 Tulio Magno Quites Machado Filho <tuliom@linux.ibm.com> 6946 6947 * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128 6948 return type and other typos. 6949 69502018-10-26 Aaron Sawdey <acsawdey@linux.ibm.com> 6951 6952 * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to 6953 a shorter sequence with fewer branches. 6954 (emit_final_str_compare_gpr): Ditto. 6955 69562018-10-26 Paul A. Clarke <pc@us.ibm.com> 6957 6958 * config/rs6000/tmmintrin.h: New file. 6959 * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers. 6960 69612018-10-26 Paul A. Clarke <pc@us.ibm.com> 6962 6963 * config/rs6000/mmintrin.h: Enable 32bit compilation. 6964 * config/rs6000/xmmintrin.h: Likewise. 6965 69662018-10-26 Paul A. Clarke <pc@us.ibm.com> 6967 6968 * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian. 6969 69702018-10-26 Richard Biener <rguenther@suse.de> 6971 6972 * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set 6973 and wrapper. 6974 (vect_mark_slp_stmts_relevant): Likewise. 6975 (vect_detect_hybrid_slp_stmts): Likewise. 6976 (vect_bb_slp_scalar_cost): Likewise. 6977 (vect_remove_slp_scalar_calls): Likewise. 6978 69792018-10-26 Jan Hubicka <jh@suse.cz> 6980 6981 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion. 6982 (warn_types_mismatch): Fix walk of DECL_NAME. 6983 (odr_types_equivalent_p): Fix overactive assert. 6984 69852018-10-26 Richard Biener <rguenther@suse.de> 6986 6987 PR tree-optimization/87105 6988 * tree-vectorizer.h (_slp_tree::refcnt): New member. 6989 * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor 6990 refcnt. 6991 (vect_create_new_slp_node): Initialize refcnt to one. 6992 (bst_traits): Move. 6993 (scalar_stmts_set_t, bst_fail): Remove. 6994 (vect_build_slp_tree_2): Add bst_map argument and adjust calls. 6995 (vect_build_slp_tree): Add bst_map argument and lookup 6996 already created SLP nodes. 6997 (vect_print_slp_tree): Handle a SLP graph, print SLP node 6998 addresses. 6999 (vect_slp_rearrange_stmts): Handle a SLP graph. 7000 (vect_analyze_slp_instance): Adjust and free SLP nodes from 7001 the CSE map. Fix indenting. 7002 (vect_schedule_slp_instance): Add short-cut. 7003 70042018-10-26 Martin Liska <mliska@suse.cz> 7005 7006 PR testsuite/86158 7007 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with 7008 addr_expr and not with pointers. 7009 70102018-10-26 Jan Hubicka <jh@suse.cz> 7011 7012 * tree.c (free_lang_data_in_type): Only check main variants. 7013 * ipa-devirt.c (warn_odr): Make static. 7014 (types_same_for_odr): Drop strict variant. 7015 (types_odr_comparable): Likewise. 7016 (odr_or_derived_type_p): Look for main variants. 7017 (odr_name_hasher::equal): Cleanup comment. 7018 (odr_subtypes_equivalent): Add warn and warned arguments; check main 7019 variants. 7020 (type_variants_equivalent_p): break out from ... 7021 (odr_types_equivalent): ... here; go for main variants where needed. 7022 (warn_odr): ... here; turn static. 7023 (warn_types_mismatch): Compare mangled names of main variants. 7024 * ipa-utils.h (types_odr_comparable): Drop strict parameter. 7025 (type_with_linkage_p): Sanity check that we look at main variant. 7026 * lto.c (lto_read_decls): Only consider main variant to be ODR type. 7027 * tree.h (types_same_for_odr): Drop strict argument. 7028 70292018-10-26 Richard Biener <rguenther@suse.de> 7030 7031 PR tree-optimization/87746 7032 * tree-vect-data-refs.c (vect_update_misalignment_for_peel): 7033 Simplify and fix WRT strided store groups with size not 7034 equal to step in element count. 7035 (vect_analyze_group_access_1): Dump the whole group. 7036 70372018-10-25 Carl Love <cel@us.ibm.com> 7038 7039 * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT, 7040 P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ, 7041 P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name. Add quad 7042 precicion entry for each overloaded builtin. 7043 * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ, 7044 VSCEDPUO): Rename overloaded name. 7045 (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ, 7046 VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins. 7047 * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add 7048 define_expand for xscmpexqp instruction. 7049 (*xscmpexpqp): Add define_insn for the xscmpexqp instruction. 7050 70512018-10-25 Bill Schmidt <wschmidt@linux.ibm.com> 7052 Jinsong Ji <jji@us.ibm.com> 7053 7054 * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated 7055 function with vec_sl. 7056 (_mm_slli_epi32): Likewise. 7057 (_mm_slli_epi64): Likewise. 7058 (_mm_srai_epi16): Replace deprecated function with vec_sra. 7059 (_mm_srai_epi32): Likewise. 7060 (_mm_srli_epi16): Replace deprecated function with vec_sr. 7061 (_mm_srli_epi32): Likewise. 7062 (_mm_srli_epi64): Likewise. 7063 (_mm_sll_epi16): Replace deprecated function with vec_sl. 7064 (_mm_sll_epi32): Likewise. 7065 (_mm_sll_epi64): Likewise. 7066 (_mm_sra_epi16): Replace deprecated function with vec_sra. 7067 (_mm_sra_epi32): Likewise. 7068 (_mm_srl_epi16): Replace deprecated function with vec_sr. 7069 (_mm_srl_epi32): Likewise. 7070 (_mm_srl_epi64): Likewise. 7071 70722018-10-25 Bill Schmidt <wschmidt@linux.ibm.com> 7073 Jinsong Ji <jji@us.ibm.com> 7074 7075 * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace 7076 comparison operators with vec_cmp* for compatibility due to 7077 unfortunate history; clean up formatting and use types more 7078 appropriately. 7079 (_mm_sll_epi32): Likewise. 7080 (_mm_sll_epi64): Likewise. 7081 (_mm_srl_epi16): Likewise. 7082 (_mm_srl_epi32): Likewise. 7083 (_mm_srl_epi64): Likewise. 7084 70852018-10-25 Bill Schmidt <wschmidt@linux.ibm.com> 7086 Jinsong Ji <jji@us.ibm.com> 7087 7088 * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast. 7089 * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to 7090 __vector __bool int. Use vec_cmpgt in preference to deprecated 7091 function vec_vcmpgtfp. 7092 (_mm_max_ps): Likewise. 7093 70942018-10-25 Jeff Law <law@redhat.com> 7095 7096 * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn 7097 if returning false. 7098 70992018-10-25 Martin Sebor <msebor@redhat.com> 7100 7101 * doc/extend.texi (aligned): Expand attribute description. 7102 (Alignment): Rename section. Discuss function arguments. 7103 71042018-10-25 Jan Hubicka <jh@suse.cz> 7105 7106 * ipa-devirt.c (main_odr_variant): Remove. 7107 (hash_odr_name, types_same_for_odr, types_odr_comparable, 7108 odr_name_hasher::equal, odr_subtypes_equivalent_p): 7109 Drop use of main_odr_variant. 7110 (add_type_duplicate): Silence confused warnings on integer types. 7111 (get_odr_type): Always look for main variant. 7112 (register_odr_type): Simplify. 7113 71142018-10-25 Richard Biener <rguenther@suse.de> 7115 7116 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): 7117 Initialize ng to silence error with release checking bootstrap. 7118 71192018-10-25 Richard Biener <rguenther@suse.de> 7120 7121 * tree-if-conv.c: Include tree-ssa-sccvn.h. 7122 (tree_if_conversion): Run CSE on the if-converted loop body. 7123 71242018-10-25 Ilya Leoshkevich <iii@linux.ibm.com> 7125 7126 * config/s390/constraints.md (ZL): New constraint. 7127 * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL 7128 operands. 7129 * config/s390/s390.md (movdi_larl): Remove. 7130 (movdi_64): Add the LARL alternative. 7131 71322018-10-25 Ilya Leoshkevich <iii@linux.ibm.com> 7133 7134 PR bootstrap/87747 7135 * rtl.c (RTX_CODE_HWINT_P_1): New helper macro. 7136 (RTX_CODE_HWINT_P): New macro. 7137 (rtx_code_size): Use RTX_CODE_HWINT_P (). 7138 71392018-10-25 Jan Hubicka <jh@suse.cz> 7140 7141 * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types 7142 is anonymous. 7143 71442018-10-25 Richard Biener <rguenther@suse.de> 7145 7146 PR tree-optimization/87665 7147 PR tree-optimization/87745 7148 * tree-vectorizer.h (get_earlier_stmt): Remove. 7149 (get_later_stmt): Pick up UID from the original non-pattern stmt. 7150 71512018-10-25 Sam Tebbs <sam.tebbs@arm.com> 7152 7153 * options.texi (Deprecated): Move list to Var section. 7154 71552018-10-24 Bill Schmidt <wschmidt@linux.ibm.com> 7156 Jinsong Ji <jji@us.ibm.com> 7157 7158 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated 7159 __vector long to __vector long long. 7160 (_mm_cvtpd_ps): Likewise. 7161 (_mm_cvttpd_epi32): Likewise. 7162 (_mm_cvtpi32_pd): Likewise. 7163 (_mm_unpackhi_epi64): Likewise. 7164 (_mm_unpacklo_epi64): Likewise. 7165 71662018-10-24 Segher Boessenkool <segher@kernel.crashing.org> 7167 7168 PR rtl-optimization/87720 7169 * combine.c (make_more_copies): Skip if the dest is pc_rtx. 7170 71712018-10-24 Alexandre Oliva <aoliva@redhat.com> 7172 7173 * gimple-ssa-isolate-paths.c 7174 (find_implicit_erroneous_behavior): Do not change code if the 7175 pass is running for warnings only. 7176 (find_explicit_erroneous_behavior): Likewise. 7177 71782018-10-24 Michael Meissner <meissner@linux.ibm.com> 7179 7180 * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME): 7181 Define as rs6000_mangle_decl_assembler_name. 7182 (rs6000_mangle_decl_assembler_name): If the user switched from IBM 7183 long double to IEEE long double, switch the names of the long 7184 double built-in functions to be <func>f128 instead of <func>l. 7185 71862018-10-24 Martin Sebor <msebor@redhat.com> 7187 7188 * doc/extend.texi (nonnull): List no-argument form. Reference 7189 -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute. 7190 71912018-10-24 Richard Biener <rguenther@suse.de> 7192 7193 * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state. 7194 71952018-10-24 Martin Liska <mliska@suse.cz> 7196 7197 PR tree-optimization/84436 7198 * tree-switch-conversion.c (switch_conversion::contains_same_values_p): 7199 Remove. 7200 (switch_conversion::contains_linear_function_p): New. 7201 (switch_conversion::build_one_array): Support linear 7202 transformation on input. 7203 * tree-switch-conversion.h (struct switch_conversion): Add 7204 contains_linear_function_p declaration. 7205 72062018-10-24 Richard Biener <rguenther@suse.de> 7207 7208 * varasm.c (const_hash_1): Return hash of ADDR_EXPR 7209 if its argument is CONSTANT_CLASS_P. 7210 72112018-10-11 Jan Hubicka <hubicka@ucw.cz> 7212 7213 * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL; 7214 it is wrong for forward declarations. 7215 72162018-10-24 Ilya Leoshkevich <iii@linux.ibm.com> 7217 7218 * config/s390/s390.c (s390_check_qrst_address): Add the missing 7219 SYMBOL_REF_P () check. 7220 72212018-10-24 Richard Biener <rguenther@suse.de> 7222 7223 PR tree-optimization/87105 7224 * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust 7225 dump classification. 7226 (vect_analyze_data_ref_accesses): Handle duplicate loads and 7227 stores by splitting the affected group after the fact. 7228 * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we 7229 fail the SLP build because of size constraints. 7230 72312018-10-24 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE> 7232 7233 * configure.ac (gcc_cv_ld_aligned_shf_merge): New test. 7234 * configure: Regenerate. 7235 * config.in: Regenerate. 7236 * varasm.c (mergeable_string_section): Use readonly_data_section 7237 if linker doesn't support SHF_MERGE with alignment > 8. 7238 (mergeable_constant_section): Likewise. 7239 72402018-10-24 Richard Biener <rguenther@suse.de> 7241 7242 PR tree-optimization/84013 7243 * tree-ssa-structalias.c (struct msdi_data): New struct for 7244 marshalling data to walk_stmt_load_store_ops. 7245 (maybe_set_dependence_info): Refactor as callback for 7246 walk_stmt_load_store_ops. 7247 (compute_dependence_clique): Set restrict info on all stmt kinds. 7248 72492018-10-24 Martin Liska <mliska@suse.cz> 7250 7251 * cgraph.c (cgraph_node::dump): 7252 Remove reduntant dumps and make tp_first_run dump more compact. 7253 72542018-10-24 Richard Biener <rguenther@suse.de> 7255 7256 PR tree-optimization/87665 7257 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust 7258 to reflect reality. 7259 72602018-10-12 Jeff Law <law@redhat.com> 7261 7262 * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation 7263 for H8/S. 7264 72652018-10-23 Richard Biener <rguenther@suse.de> 7266 7267 * tree-vrp.c (add_assert_info): Guard dump_printf with 7268 dump_enabled_p. 7269 * gimple-ssa-evrp-analyze.c 7270 (evrp_range_analyzer::record_ranges_from_incoming_edge): 7271 Use value_range::ignore_equivs_equal_p. 7272 72732018-10-23 Richard Biener <rguenther@suse.de> 7274 7275 PR tree-optimization/87105 7276 PR tree-optimization/87608 7277 * passes.def (pass_all_early_optimizations): Add early phi-opt 7278 after dce. 7279 * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in 7280 addition to debug stmts. 7281 (tree_ssa_phiopt_worker): Add early_p argument, do only min/max 7282 and abs replacement early. 7283 * tree-cfg.c (gimple_empty_block_p): Likewise. 7284 72852018-10-23 Richard Earnshaw <rearnsha@arm.com> 7286 7287 PR target/86383 7288 * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU 7289 specified to configure. 7290 (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided. 7291 72922018-10-23 Richard Biener <rguenther@suse.de> 7293 7294 PR tree-optimization/87700 7295 * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic. 7296 72972018-10-23 Jakub Jelinek <jakub@redhat.com> 7298 7299 PR target/87674 7300 * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of 7301 second argument from __mmask16 to __mmask8. 7302 * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32, 7303 _mm_mask_packs_epi32): Likewise. 7304 * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps): 7305 Likewise. 7306 (_mm512_mask_prefetch_i64scatter_pd): Likewise. Formatting fix. 7307 73082018-10-23 Richard Biener <rguenther@suse.de> 7309 7310 * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment. 7311 73122018-10-23 Richard Biener <rguenther@suse.de> 7313 7314 PR tree-optimization/86144 7315 * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi 7316 over simd attribute. 7317 73182018-10-23 Richard Biener <rguenther@suse.de> 7319 7320 PR tree-optimization/87693 7321 * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle 7322 the case we do not find the taken edge. 7323 73242018-10-22 Bill Schmidt <wschmidt@linux.ibm.com> 7325 Jinsong Ji <jji@us.ibm.com> 7326 7327 * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup. 7328 (_mm_store_pd): Use unaligned vector type for pointer cast. 7329 (_mm_maskmoveu_si128): Likewise. 7330 * config/rs6000/xmmintrin.h (__m128_u): New typedef. 7331 (_mm_store_ps): Use unaligned vector type for pointer cast. 7332 73332018-10-22 Paul Koning <ni1d@arrl.net> 7334 7335 * symtab.c (symtab_node::increase_alignment): Correct max 7336 alignment check. 7337 73382018-10-22 Yury Gribov <tetra2005@gmail.com> 7339 7340 PR tree-optimization/87633 7341 * match.pd: Do not generate unordered integer comparisons. 7342 73432018-10-22 Segher Boessenkool <segher@kernel.crashing.org> 7344 7345 PR rtl-optimization/87600 7346 * combine.c: Add include of expr.h. 7347 (cant_combine_insn_p): Do not combine moves from any hard non-fixed 7348 register to a pseudo. 7349 (make_more_copies): New function, add a copy to a new pseudo after 7350 the moves from hard registers into pseudos. 7351 (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine 7352 later. Call make_more_copies. 7353 73542018-10-22 Andrew Stubbs <ams@codesourcery.com> 7355 7356 * lra-constraints.c (process_alt_operands): New local array, 7357 matching_early_clobber. Check matching_early_clobber before 7358 decrementing reject, and set matching_early_clobber after. 7359 73602018-10-22 Segher Boessenkool <segher@kernel.crashing.org> 7361 7362 PR target/87598 7363 * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL 7364 call output_addr_const and hope for the best. 7365 73662018-10-22 Richard Biener <rguenther@suse.de> 7367 7368 * gimple-ssa-evrp-analyze.c 7369 (evrp_range_analyzer::record_ranges_from_incoming_edge): Be 7370 smarter about what ranges to use. 7371 * tree-vrp.c (add_assert_info): Dump here. 7372 (register_edge_assert_for_2): Instead of here at multiple but 7373 not all places. 7374 7375 * gcc.dg/tree-ssa/evrp12.c: New testcase. 7376 * gcc.dg/predict-6.c: Adjust. 7377 * gcc.dg/tree-ssa/vrp33.c: Disable EVRP. 7378 * gcc.dg/tree-ssa/vrp02.c: Likewise. 7379 * gcc.dg/tree-ssa/cunroll-9.c: Likewise. 7380 73812018-10-22 Steven Bosscher <steven@gcc.gnu.org> 7382 Richard Biener <rguenther@suse.de> 7383 7384 * bitmap.h: Update data structure documentation, including a 7385 description of bitmap views as either linked-lists or splay trees. 7386 (struct bitmap_element_def): Update comments for splay tree bitmaps. 7387 (struct bitmap_head_def): Likewise. 7388 (bitmap_list_view, bitmap_tree_view): New prototypes. 7389 (bitmap_initialize_stat): Initialize a bitmap_head's indx and 7390 tree_form fields. 7391 (bmp_iter_set_init): Assert the iterated bitmaps are in list form. 7392 (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise. 7393 * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a 7394 released bitmap element here. 7395 (bitmap_element_free): Remove. 7396 (bitmap_elt_clear_from): Work on splay tree bitmaps. 7397 (bitmap_list_link_element): Renamed from bitmap_element_link. Move 7398 this function similar ones such that linked-list bitmap implementation 7399 functions are grouped. 7400 (bitmap_list_unlink_element): Renamed from bitmap_element_unlink, 7401 and moved for grouping. 7402 (bitmap_list_insert_element_after): Renamed from 7403 bitmap_elt_insert_after, and moved for grouping. 7404 (bitmap_list_find_element): New function spliced from bitmap_find_bit. 7405 (bitmap_tree_link_left, bitmap_tree_link_right, 7406 bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay, 7407 bitmap_tree_link_element, bitmap_tree_unlink_element, 7408 bitmap_tree_find_element): New functions for splay-tree bitmap 7409 implementation. 7410 (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after): 7411 Renamed and moved, see above entries. 7412 (bitmap_tree_listify_from): New function to convert part of a splay 7413 tree bitmap to a linked-list bitmap. 7414 (bitmap_list_view): Convert a splay tree bitmap to linked-list form. 7415 (bitmap_tree_view): Convert a linked-list bitmap to splay tree form. 7416 (bitmap_find_bit): Remove. 7417 (bitmap_clear, bitmap_clear_bit, bitmap_set_bit, 7418 bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit): 7419 Handle splay tree bitmaps. 7420 (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into, 7421 bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into, 7422 bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into, 7423 bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p, 7424 bitmap_intersect_compl_p, bitmap_ior_and_compl, 7425 bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range, 7426 bitmap_hash): Reject trying to act on splay tree bitmaps. Make 7427 corresponding changes to use linked-list specific bitmap_element 7428 manipulation functions as applicable for efficiency. 7429 (bitmap_tree_to_vec): New function. 7430 (debug_bitmap_elt_file): New function split out from ... 7431 (debug_bitmap_file): ... here. Handle splay tree bitmaps. 7432 (bitmap_print): Likewise. 7433 7434 PR tree-optimization/63155 7435 * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the 7436 SSA edge worklists. 7437 * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies 7438 in tree-view. 7439 74402018-10-22 Martin Liska <mliska@suse.cz> 7441 7442 PR tree-optimization/87686 7443 Revert 7444 2018-08-29 Martin Liska <mliska@suse.cz> 7445 7446 * tree-switch-conversion.c (switch_conversion::expand): 7447 Strenghten assumption about gswitch statements. 7448 74492018-10-22 Martin Liska <mliska@suse.cz> 7450 7451 * ipa-icf.c (sem_item::compare_attributes): Remove. 7452 (sem_item::compare_referenced_symbol_properties): Use 7453 attribute_list_equal instead. 7454 (sem_function::equals_wpa): Likewise. 7455 * ipa-icf.h: Remove compare_attributes. 7456 74572018-10-22 Richard Biener <rguenther@suse.de> 7458 7459 PR middle-end/87682 7460 * mem-stats.h (mem_usage::operator==): Fix pasto. 7461 74622018-10-22 Richard Biener <rguenther@suse.de> 7463 7464 PR tree-optimization/87640 7465 * tree-vrp.c (set_value_range_with_overflow): Decompose 7466 incomplete result. 7467 (extract_range_from_binary_expr_1): Adjust. 7468 74692018-10-22 Martin Jambor <mjambor@suse.cz> 7470 7471 * tree-eh.h (stmt_could_throw_p): Add function parameter. 7472 (stmt_can_throw_external): Likewise. 7473 (stmt_can_throw_internal): Likewise. 7474 * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p. 7475 (lower_eh_constructs_2): Likewise. 7476 (stmt_could_throw_p): Add fun parameter, use it instead of cfun. 7477 (stmt_can_throw_external): Likewise. 7478 (stmt_can_throw_internal): Likewise. 7479 (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p. 7480 (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p. 7481 (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p. 7482 (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p. 7483 (pass_lower_eh_dispatch::execute): Pass cfun to 7484 stmt_can_throw_external. 7485 (cleanup_empty_eh): Likewise. 7486 (verify_eh_edges): Pass cfun to stmt_could_throw_p. 7487 * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to 7488 stmt_can_throw_external instead of pushing it to cfun. 7489 (symbol_table::create_edge): Likewise. 7490 * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to 7491 stmt_can_throw_internal. 7492 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun 7493 to stmt_could_throw_p. 7494 * gimple-ssa-store-merging.c (handled_load): Pass cfun to 7495 stmt_can_throw_internal. 7496 (pass_store_merging::execute): Likewise. 7497 * gimple-ssa-strength-reduction.c 7498 (find_candidates_dom_walker::before_dom_children): Pass cfun to 7499 stmt_could_throw_p. 7500 * gimplify-me.c (gimple_regimplify_operands): Pass cfun to 7501 stmt_can_throw_internal. 7502 * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and 7503 to stmt_can_throw_external. 7504 (check_stmt): Pass cfun to stmt_could_throw_p. 7505 (check_stmt): Pass cfun to stmt_can_throw_external. 7506 (pass_nothrow::execute): Likewise. 7507 * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal. 7508 * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to 7509 stmt_can_throw_internal. 7510 (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p. 7511 (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external. 7512 (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal. 7513 * tree-complex.c (expand_complex_libcall): Pass cfun to 7514 stmt_could_throw_p and to stmt_can_throw_internal. 7515 (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal. 7516 * tree-inline.c (copy_edges_for_bb): Likewise. 7517 (maybe_move_debug_stmts_to_successors): Likewise. 7518 * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to 7519 stmt_could_throw_p. 7520 * tree-parloops.c (oacc_entry_exit_ok_1): Likewise. 7521 * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external. 7522 * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to 7523 stmt_can_throw_internal. 7524 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise. 7525 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to 7526 stmt_could_throw_p. 7527 (mark_aliased_reaching_defs_necessary_1): Pass cfun to 7528 stmt_can_throw_internal. 7529 * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise. 7530 * tree-ssa-loop-im.c (movement_possibility): Pass cfun to 7531 stmt_could_throw_p. 7532 * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise. 7533 (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal. 7534 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise. 7535 (convert_mult_to_fma_1): Likewise. 7536 (convert_to_divmod): Likewise. 7537 * tree-ssa-phiprop.c (propagate_with_phi): Likewise. 7538 * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p. 7539 * tree-ssa-propagate.c 7540 (substitute_and_fold_dom_walker::before_dom_children): Likewise. 7541 * tree-ssa-reassoc.c (suitable_cond_bb): Likewise. 7542 (maybe_optimize_range_tests): Likewise. 7543 (linearize_expr_tree): Likewise. 7544 (reassociate_bb): Likewise. 7545 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise. 7546 * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise. 7547 * tree-ssa-strlen.c (adjust_last_stmt): Likewise. 7548 (handle_char_store): Likewise. 7549 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to 7550 stmt_can_throw_internal. 7551 * tree-vect-patterns.c (check_bool_pattern): Pass cfun to 7552 stmt_could_throw_p. 7553 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise. 7554 (vectorizable_call): Pass cfun to stmt_can_throw_internal. 7555 (vectorizable_simd_clone_call): Likewise. 7556 * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p. 7557 (gimple_stringop_fixed_value): Likewise. 7558 75592018-10-22 Ilya Leoshkevich <iii@linux.ibm.com> 7560 7561 * config/s390/s390.c (s390_loadrelative_operand_p): Accept 7562 literal pool references. 7563 (s390_check_qrst_address): Adapt to the new behavior of 7564 s390_loadrelative_operand_p (). 7565 75662018-10-22 H.J. Lu <hongjiu.lu@intel.com> 7567 7568 PR target/72782 7569 * config/i386/sse.md (*andnot<mode>3_bcst): New. 7570 75712018-10-22 H.J. Lu <hongjiu.lu@intel.com> 7572 7573 PR target/72782 7574 * config/i386/sse.md (*<code><mode>3_bcst): New. 7575 75762018-10-22 H.J. Lu <hongjiu.lu@intel.com> 7577 7578 PR target/72782 7579 * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI, 7580 V4DI, V16SI and V8DI. 7581 (*sub<mode>3<mask_name>_bcst): New. 7582 (*add<mode>3<mask_name>_bcst): Likewise. 7583 75842018-10-21 Bill Schmidt <wschmidt@linux.ibm.com> 7585 Jinsong Ji <jji@us.ibm.com> 7586 7587 * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector 7588 __m64 with __vector unsigned long long for compatibility. 7589 (_mm_movemask_epi8): Likewise. 7590 * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise. 7591 (_mm_cvttps_pi32): Likewise. 7592 (_mm_cvtpi32_ps): Likewise. 7593 (_mm_cvtps_pi16): Likewise. 7594 (_mm_loadh_pi): Likewise. 7595 (_mm_storeh_pi): Likewise. 7596 (_mm_movehl_ps): Likewise. 7597 (_mm_movelh_ps): Likewise. 7598 (_mm_loadl_pi): Likewise. 7599 (_mm_storel_pi): Likewise. 7600 (_mm_movemask_ps): Likewise. 7601 (_mm_shuffle_pi16): Likewise. 7602 76032018-10-21 H.J. Lu <hongjiu.lu@intel.com> 7604 7605 PR target/72782 7606 * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use 7607 __builtin_ia32_vfnmsubpd512_mask. 7608 (_mm512_mask_fnmsub_round_pd): Likewise. 7609 (_mm512_fnmsub_pd): Likewise. 7610 (_mm512_mask_fnmsub_pd): Likewise. 7611 (_mm512_maskz_fnmsub_round_pd): Use 7612 __builtin_ia32_vfnmsubpd512_maskz. 7613 (_mm512_maskz_fnmsub_pd): Likewise. 7614 (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask. 7615 (_mm512_mask_fnmsub_round_ps): Likewise. 7616 (_mm512_fnmsub_ps): Likewise. 7617 (_mm512_mask_fnmsub_ps): Likewise. 7618 (_mm512_maskz_fnmsub_round_ps): Use 7619 __builtin_ia32_vfnmsubps512_maskz. 7620 (_mm512_maskz_fnmsub_ps): Likewise. 7621 * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use 7622 __builtin_ia32_vfnmsubpd256_mask. 7623 (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz. 7624 (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask 7625 (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz. 7626 (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask. 7627 (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask. 7628 (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz. 7629 (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask. 7630 (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz. 7631 * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use 7632 __builtin_ia32_vfnmsubpd. 7633 (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256. 7634 (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps. 7635 (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256. 7636 (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3. 7637 (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3. 7638 * config/i386/i386-builtin.def: Add 7639 __builtin_ia32_vfnmsubpd256_mask, 7640 __builtin_ia32_vfnmsubpd256_maskz, 7641 __builtin_ia32_vfnmsubpd128_mask, 7642 __builtin_ia32_vfnmsubpd128_maskz, 7643 __builtin_ia32_vfnmsubps256_mask, 7644 __builtin_ia32_vfnmsubps256_maskz, 7645 __builtin_ia32_vfnmsubps128_mask, 7646 __builtin_ia32_vfnmsubps128_maskz, 7647 __builtin_ia32_vfnmsubpd512_mask, 7648 __builtin_ia32_vfnmsubpd512_maskz, 7649 __builtin_ia32_vfnmsubps512_mask, 7650 __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3, 7651 __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps, 7652 __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and. 7653 __builtin_ia32_vfnmsubpd256. 7654 * config/i386/sse.md (fma4i_fnmsub_<mode>): New. 7655 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise. 7656 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): 7657 Likewise. 7658 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): 7659 Likewise. 7660 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): 7661 Likewise. 7662 (fmai_vmfnmsub_<mode><round_name>): Likewise. 7663 76642018-10-21 H.J. Lu <hongjiu.lu@intel.com> 7665 7666 PR target/72782 7667 * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use 7668 __builtin_ia32_vfnmaddpd512_mask. 7669 (_mm512_mask_fnmadd_round_pd): Likewise. 7670 (_mm512_fnmadd_pd): Likewise. 7671 (_mm512_mask_fnmadd_pd): Likewise. 7672 (_mm512_maskz_fnmadd_round_pd): Use 7673 __builtin_ia32_vfnmaddpd512_maskz. 7674 (_mm512_maskz_fnmadd_pd): Likewise. 7675 (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask. 7676 (_mm512_mask_fnmadd_round_ps): Likewise. 7677 (_mm512_fnmadd_ps): Likewise. 7678 (_mm512_mask_fnmadd_ps): Likewise. 7679 (_mm512_maskz_fnmadd_round_ps): Use 7680 __builtin_ia32_vfnmaddps512_maskz. 7681 (_mm512_maskz_fnmadd_ps): Likewise. 7682 * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use 7683 __builtin_ia32_vfnmaddpd256_mask. 7684 (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz. 7685 (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask 7686 (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz. 7687 (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask. 7688 (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask. 7689 (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz. 7690 (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask. 7691 (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz. 7692 * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use 7693 __builtin_ia32_vfnmaddpd. 7694 (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256. 7695 (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps. 7696 (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256. 7697 (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3. 7698 (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3. 7699 * config/i386/i386-builtin.def: Add 7700 __builtin_ia32_vfnmaddpd256_mask, 7701 __builtin_ia32_vfnmaddpd256_maskz, 7702 __builtin_ia32_vfnmaddpd128_mask, 7703 __builtin_ia32_vfnmaddpd128_maskz, 7704 __builtin_ia32_vfnmaddps256_mask, 7705 __builtin_ia32_vfnmaddps256_maskz, 7706 __builtin_ia32_vfnmaddps128_mask, 7707 __builtin_ia32_vfnmaddps128_maskz, 7708 __builtin_ia32_vfnmaddpd512_mask, 7709 __builtin_ia32_vfnmaddpd512_maskz, 7710 __builtin_ia32_vfnmaddps512_mask, 7711 __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3, 7712 __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps, 7713 __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and. 7714 __builtin_ia32_vfnmaddpd256. 7715 * config/i386/sse.md (fma4i_fnmadd_<mode>): New. 7716 (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise. 7717 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): 7718 Likewise. 7719 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): 7720 Likewise. 7721 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): 7722 Likewise. 7723 (fmai_vmfnmadd_<mode><round_name>): Likewise. 7724 77252018-10-21 H.J. Lu <hongjiu.lu@intel.com> 7726 7727 PR target/72782 7728 * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use 7729 __builtin_ia32_vfmsubpd512_mask. 7730 (_mm512_mask_fmsub_round_pd): Likewise. 7731 (_mm512_fmsub_pd): Likewise. 7732 (_mm512_mask_fmsub_pd): Likewise. 7733 (_mm512_maskz_fmsub_round_pd): Use 7734 __builtin_ia32_vfmsubpd512_maskz. 7735 (_mm512_maskz_fmsub_pd): Likewise. 7736 (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask. 7737 (_mm512_mask_fmsub_round_ps): Likewise. 7738 (_mm512_fmsub_ps): Likewise. 7739 (_mm512_mask_fmsub_ps): Likewise. 7740 (_mm512_maskz_fmsub_round_ps): Use 7741 __builtin_ia32_vfmsubps512_maskz. 7742 (_mm512_maskz_fmsub_ps): Likewise. 7743 * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use 7744 __builtin_ia32_vfmsubpd256_mask. 7745 (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz. 7746 (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask 7747 (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz. 7748 (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask. 7749 (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask. 7750 (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz. 7751 (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask. 7752 (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz. 7753 * config/i386/fmaintrin.h (_mm_fmsub_pd): Use 7754 __builtin_ia32_vfmsubpd. 7755 (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256. 7756 (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps. 7757 (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256. 7758 (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3. 7759 (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3. 7760 * config/i386/i386-builtin.def: Add 7761 __builtin_ia32_vfmsubpd256_mask, 7762 __builtin_ia32_vfmsubpd256_maskz, 7763 __builtin_ia32_vfmsubpd128_mask, 7764 __builtin_ia32_vfmsubpd128_maskz, 7765 __builtin_ia32_vfmsubps256_mask, 7766 __builtin_ia32_vfmsubps256_maskz, 7767 __builtin_ia32_vfmsubps128_mask, 7768 __builtin_ia32_vfmsubps128_maskz, 7769 __builtin_ia32_vfmsubpd512_mask, 7770 __builtin_ia32_vfmsubpd512_maskz, 7771 __builtin_ia32_vfmsubps512_mask, 7772 __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3, 7773 __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps, 7774 __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and. 7775 __builtin_ia32_vfmsubpd256. 7776 * config/i386/sse.md (fma4i_fmsub_<mode>): New. 7777 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise. 7778 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): 7779 Likewise. 7780 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): 7781 Likewise. 7782 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): 7783 Likewise. 7784 (fmai_vmfmsub_<mode><round_name>): Likewise. 7785 77862018-10-21 H.J. Lu <hongjiu.lu@intel.com> 7787 7788 * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1): 7789 Remove plus. Renamed to ... 7790 (*sub<mode>3<mask_name>_bcst): This. 7791 (*add<mode>3<mask_name>_bcst_2): Renamede to ... 7792 (*add<mode>3<mask_name>_bcst): This. 7793 77942018-10-21 H.J. Lu <hongjiu.lu@intel.com> 7795 7796 PR target/72782 7797 * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New. 7798 77992018-10-21 H.J. Lu <hongjiu.lu@intel.com> 7800 7801 PR target/87662 7802 * i386/avx512vlintrin.h (_mm256_or_epi32): New. 7803 (_mm_or_epi32): Likewise. 7804 (_mm256_xor_epi32): Likewise. 7805 (_mm_xor_epi32): Likewise. 7806 (_mm256_or_epi64): Likewise. 7807 (_mm_or_epi64): Likewise. 7808 (_mm256_xor_epi64): Likewise. 7809 (_mm_xor_epi64): Likewise. 7810 78112018-10-20 H.J. Lu <hongjiu.lu@intel.com> 7812 7813 PR target/72782 7814 * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New. 7815 78162018-10-20 Jakub Jelinek <jakub@redhat.com> 7817 7818 PR middle-end/87647 7819 * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR. 7820 78212018-10-20 Andreas Schwab <schwab@linux-m68k.org> 7822 7823 * doc/ux.texi: Move @section directly after @node. 7824 78252018-10-19 Jakub Jelinek <jakub@redhat.com> 7826 7827 PR middle-end/85488 7828 PR middle-end/87649 7829 * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without 7830 depend closely nested inside of loop with ordered clause with 7831 a parameter. 7832 78332018-10-19 David Malcolm <dmalcolm@redhat.com> 7834 7835 * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi. 7836 * doc/gccint.texi: Include ux.texi and use it in top-level menu. 7837 * doc/ux.texi: New file. 7838 78392018-10-19 Segher Boessenkool <segher@kernel.crashing.org> 7840 7841 * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to 7842 be the first CR field allocated. 7843 78442018-10-19 Richard Biener <rguenther@suse.de> 7845 7846 PR target/87657 7847 * config/i386/i386.c (ix86_builtin_vectorization_cost): Use 7848 TYPE_VECTOR_SUBPARTS and avoid relying on vector mode. 7849 78502018-10-19 H.J. Lu <hongjiu.lu@intel.com> 7851 7852 PR target/72782 7853 * config/i386/sse.md 7854 (*<plusminus_insn><mode>3<mask_name>_bcst_1): New. 7855 (*add<mode>3<mask_name>_bcst_2): Likewise. 7856 78572018-10-19 H.J. Lu <hongjiu.lu@intel.com> 7858 7859 * config/i386/sse.md 7860 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): 7861 Replace nonimmediate_operand with register_operand. 7862 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): 7863 Likewise. 7864 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): 7865 Likewise. 7866 78672018-10-19 Ilya Leoshkevich <iii@linux.ibm.com> 7868 7869 PR rtl-optimization/87596 7870 * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () + 7871 lra_get_insn_recog_data () instead of lra_insn_recog_data[] 7872 for instructions in FROM..TO range. 7873 78742018-10-19 Eric Botcazou <ebotcazou@adacore.com> 7875 7876 * cfgexpand.c (expand_one_var): Use specific wording in error message 7877 for non-local frame variables. 7878 * stor-layout.c (layout_decl): Do not issue a warning for them. 7879 78802018-10-19 Robin Dapp <rdapp@linux.ibm.com> 7881 7882 * haifa-sched.c (priority): Add force_recompute parameter. 7883 (apply_replacement): Call priority () with force_recompute = true. 7884 (restore_pattern): Likewise. 7885 78862018-10-18 H.J. Lu <hongjiu.lu@intel.com> 7887 7888 * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to 7889 HOST_BITS_PER_WIDE_INT. 7890 (test_vector_ops_duplicate): Likewise. 7891 78922018-10-18 H.J. Lu <hongjiu.lu@intel.com> 7893 7894 PR target/72782 7895 * config/i386/sse.md (VF_AVX512): New. 7896 (avx512bcst): Likewise. 7897 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): 7898 Likewise. 7899 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): 7900 Likewise. 7901 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): 7902 Likewise. 7903 79042018-10-18 Jonathan Wakely <jwakely@redhat.com> 7905 7906 * doc/invoke.texi (-dumpversion): Improve grammar. 7907 (-dumpfullversion): Make more consistent with -dumpversion. 7908 79092018-10-18 Uros Bizjak <ubizjak@gmail.com> 7910 7911 * config/i386/i386.c (ix86_emit_fp_unordered_jump): 7912 Set JUMP_LABEL to the jump insn. 7913 (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch. 7914 Predict emitted jump and add label to jump insn. 7915 79162018-10-18 David Malcolm <dmalcolm@redhat.com> 7917 7918 PR tree-optimization/87562 7919 * input.c (get_substring_ranges_for_loc): Use 7920 LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when 7921 getting the linemap for the endpoint. Verify that it's either 7922 in the same linemap as the start point's spelling location, or 7923 at least in the same file. 7924 79252018-10-18 Richard Biener <rguenther@suse.de> 7926 7927 * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not 7928 feed width-specific load/store costs through ix86_vec_cost. 7929 * config/i386/x86-tune-costs.h (athlon_cost): Adjust. 7930 (k8_cost): Likewise. 7931 (bdver_cost): Likewise. 7932 (znver1_cost): Likewise. 7933 (btver1_cost): Likewise. 7934 (btver2_cost): Likewise. 7935 79362018-10-18 H.J. Lu <hongjiu.lu@intel.com> 7937 7938 * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg 7939 to simplify subreg of vec_merge. 7940 79412018-10-18 Richard Biener <rguenther@suse.de> 7942 7943 * config/i386/i386.c: Fix costing of vector FMA. 7944 79452018-10-18 Richard Biener <rguenther@suse.de> 7946 7947 * config/i386/i386.c (ix86_vec_cost): Remove !parallel path 7948 and argument. 7949 (ix86_builtin_vectorization_cost): For vec_construct properly 7950 cost insertion into SSE regs. 7951 (...): Adjust calls to ix86_vec_cost. 7952 79532018-10-18 Richard Biener <rguenther@suse.de> 7954 7955 PR middle-end/87087 7956 Revert 7957 2018-02-07 Richard Biener <rguenther@suse.de> 7958 7959 PR tree-optimization/84204 7960 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in 7961 this place. 7962 79632018-10-18 H.J. Lu <hongjiu.lu@intel.com> 7964 7965 PR target/87537 7966 * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge 7967 of vec_duplicate. 7968 (test_vector_ops_duplicate): Add test for a scalar subreg of a 7969 VEC_MERGE of a VEC_DUPLICATE. 7970 79712018-10-17 Joseph Myers <joseph@codesourcery.com> 7972 7973 * doc/cpp.texi (__STDC_VERSION__): Document C2X handling. 7974 * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options. 7975 * doc/standards.texi (C Language): Document C2X. 7976 * dwarf2out.c (highest_c_language), config/rl78/rl78.c 7977 (rl78_option_override): Handle "GNU C2X" language name. 7978 79792018-10-17 Joseph Myers <joseph@codesourcery.com> 7980 7981 * doc/invoke.texi (-std=c17), doc/standards.texi (C Language): 7982 Document C17 as published in 2018. 7983 79842018-10-17 Eric Botcazou <ebotcazou@adacore.com> 7985 7986 PR middle-end/87623 7987 * fold-const.c (fold_truth_andor_1): If the right side is not constant, 7988 bail out if both sides do not have the same storage order. 7989 79902018-10-17 Aldy Hernandez <aldyh@redhat.com> 7991 7992 * bitmap.c (bitmap_head::dump): New. 7993 * bitmap.h (bitmap_head): Add dump(). 7994 * gimple-ssa-evrp-analyze.c 7995 (evrp_range_analyzer::try_find_new_range): Adjust for value_range API. 7996 (evrp_range_analyzer::set_ssa_range_info): Same. 7997 (evrp_range_analyzer::record_ranges_from_phis): Same. 7998 (evrp_range_analyzer::record_ranges_from_stmt): Same. 7999 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same. 8000 * gimple-ssa-sprintf.c (get_int_range): Same. 8001 (format_integer): Same. 8002 (sprintf_dom_walker::handle_gimple_call): Same. 8003 * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same. 8004 (ipcp_vr_lattice::top_p): Same. 8005 (ipcp_vr_lattice::bottom_p): Same. 8006 (ipcp_vr_lattice::set_to_bottom): Same. 8007 (ipa_vr_operation_and_type_effects): Same. 8008 (propagate_vr_across_jump_function): Same. 8009 (ipcp_store_vr_results): Same. 8010 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same. 8011 (ipa_print_node_jump_functions_for_edge): Same. 8012 (ipa_get_value_range): Same. 8013 (ipa_compute_jump_functions_for_edge): Same. 8014 (ipa_write_jump_function): Same. 8015 * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same. 8016 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis): 8017 Same. 8018 * vr-values.c (set_value_range_to_nonnegative): Same. 8019 (set_value_range_to_truthvalue): Same. 8020 (vr_values::get_value_range): Same. 8021 (vr_values::set_defs_to_varying): Same. 8022 (vr_values::update_value_range): Same. 8023 (symbolic_range_based_on_p): Same. 8024 (vr_values::op_with_boolean_value_range_p): Same. 8025 (vr_values::extract_range_for_var_from_comparison_expr): Same. 8026 (vr_values::extract_range_from_ssa_name): Same. 8027 (vr_values::extract_range_from_binary_expr): Same. 8028 (vr_values::extract_range_from_unary_expr): Same. 8029 (vr_values::extract_range_from_cond_expr): Same. 8030 (vr_values::extract_range_from_comparison): Same. 8031 (vr_values::check_for_binary_op_overflow): Same. 8032 (vr_values::extract_range_basic): Same. 8033 (vr_values::extract_range_from_assignment): Same. 8034 (compare_ranges): Same. 8035 (compare_range_with_value): Same. 8036 (vr_values::adjust_range_with_scev): Same. 8037 (vrp_valueize): Same. 8038 (vrp_valueize_1): Same. 8039 (vr_values::get_vr_for_comparison): Same. 8040 (vr_values::compare_name_with_value): Same. 8041 (vr_values::compare_names): Same. 8042 (vr_values::vrp_evaluate_conditional): Same. 8043 (find_case_label_ranges): Same. 8044 (vr_values::vrp_visit_switch_stmt): Same. 8045 (vr_values::extract_range_from_phi_node): Same. 8046 (vr_values::simplify_div_or_mod_using_ranges): Same. 8047 (vr_values::simplify_bit_ops_using_ranges): Same. 8048 (test_for_singularity): Same. 8049 (range_fits_type_p): Same. 8050 (vr_values::simplify_cond_using_ranges_1): Same. 8051 (vr_values::simplify_switch_using_ranges): Same. 8052 (vr_values::simplify_float_conversion_using_ranges): Same. 8053 (vr_values::two_valued_val_range_p): Same. 8054 (vr_values::add_equivalence): Move to value_range::equiv_add. 8055 * vr-values.h (vr_values::add_equivalence): Remove. 8056 (VR_INITIALIZER): Remove. 8057 * tree-vrp.c (value_range::set): New. 8058 (value_range::equiv_add): New. 8059 (value_range::value_range): New. 8060 (value_range::deep_copy): New. 8061 (value_range::check): New. 8062 (value_range::equal_p): New. 8063 (value_range::ignore_equivs_equal_p): New. 8064 (value_range::operator==): New. 8065 (value_range::operator!=): New. 8066 (value_range::symbolic_p): New. 8067 (value_range::numeric_p): New. 8068 (value_range::set_undefined): New. 8069 (value_range::set_varying): New. 8070 (value_range::may_contain_p): New. 8071 (value_range::equiv_clear): New. 8072 (value_range::singleton_p): New. 8073 (value_range::intersect): New. 8074 (value_range::dump): New. 8075 (value_range::set_and_canonicalize): New. 8076 (set_value_range): Adjust for value_range API. 8077 (set_value_range_to_undefined): Same. 8078 (set_value_range_to_varying): Same. 8079 (set_and_canonicalize_value_range): Same. 8080 (set_value_range_to_nonnull): Same. 8081 (set_value_range_to_null): Same. 8082 (range_is_null): Same. 8083 (range_is_nonnull): Same. 8084 (range_int_cst_p): Same. 8085 (range_int_cst_singleton_p): Same. 8086 (symbolic_range_p): Same. 8087 (range_includes_zero_p): Same. 8088 (value_range_constant_singleton): Same. 8089 (vrp_set_zero_nonzero_bits): Same. 8090 (ranges_from_anti_range): Same. 8091 (extract_range_into_wide_ints): Same. 8092 (extract_range_from_multiplicative_op): Same. 8093 (set_value_range_with_overflow): Same. 8094 (extract_range_from_binary_expr_1): Same. 8095 (extract_range_from_unary_expr): Same. 8096 (dump_value_range): Same. 8097 (debug_value_range): Same. 8098 (vrp_prop::check_array_ref): Same. 8099 (vrp_prop::check_mem_ref): Same. 8100 (vrp_prop::vrp_initialize): Same. 8101 (vrp_prop::visit_stmt): Same. 8102 (intersect_ranges): Same. 8103 (vrp_prop::visit_phi): Same. 8104 (vrp_prop::vrp_finalize): Same. 8105 (determine_value_range_1): Same. 8106 (determine_value_range): Same. 8107 (vrp_intersect_ranges_1): Rename to... 8108 (vrp_intersect_1): this. 8109 (vrp_intersect_ranges): Rename to... 8110 (value_range::intersect_helper): ...this. 8111 (vrp_meet_1): Rename to... 8112 (value_range::union_helper): ...this. 8113 (vrp_meet): Rename to... 8114 (value_range::union_): ...this. 8115 (copy_value_range): Remove. 8116 * tree-vrp.h (struct value_range): Rewrite into a proper class. 8117 (value_range::vrtype): New. 8118 (value_range::type): New. 8119 (value_range::equiv): New. 8120 (value_range::min): New. 8121 (value_range::max): New. 8122 (value_range::varying_p): New. 8123 (value_range::undefined_p): New. 8124 (value_range::null_p): New. 8125 (value_range::equiv_add): New. 8126 (copy_value_range): Remove. 8127 81282018-10-17 David Malcolm <dmalcolm@redhat.com> 8129 8130 * Makefile.in (SELFTEST_TARGETS): New. 8131 (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS). 8132 (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb) 8133 (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to 8134 c/Make-lang.in. 8135 (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++) 8136 (selftest-c++-gdb, selftest-c++-valgrind): Move to 8137 cp/Make-lang.in. 8138 * configure: Regenerate. 8139 * configure.ac (selftest_languages): New. 8140 81412018-10-17 Aldy Hernandez <aldyh@redhat.com> 8142 8143 * tree-vrp.c (extract_range_from_multiplicative_op): Remove 8144 overflow wraps argument. 8145 (extract_range_from_binary_expr_1): Do not pass overflow wraps to 8146 wide_int_range_multiplicative_op. 8147 * wide-int-range.cc (wide_int_range_mult_wrapping): Remove 8148 overflow wraps argument. 8149 (wide_int_range_multiplicative_op): Same. 8150 (wide_int_range_lshift): Same. 8151 (wide_int_range_div): Same. 8152 * wide-int-range.h (wide_int_range_multiplicative_op): Same. 8153 (wide_int_range_lshift): Same. 8154 (wide_int_range_div): Same. 8155 81562018-10-17 Aldy Hernandez <aldyh@redhat.com> 8157 8158 * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to 8159 use sign as argument. 8160 * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to 8161 wide_int_range_shift_undefined_p. 8162 81632018-10-16 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 8164 8165 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): 8166 Rename to... 8167 (@despeculate_copy<ALLI_TI:mode>): ... This. 8168 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove 8169 switch statement. 8170 81712018-10-16 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE> 8172 8173 * config.gcc: Obsolete *-*-solaris2.10*. 8174 * doc/install.texi (Specific, *-*-solaris2*): Document it. 8175 81762018-10-12 Jeff Law <law@redhat.com> 8177 8178 * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable 8179 reg + sym +- const_int addressing modes. 8180 81812018-10-15 David Malcolm <dmalcolm@redhat.com> 8182 8183 * common.opt (fdiagnostics-minimum-margin-width=): New option. 8184 * diagnostic-show-locus.c (layout::layout): Apply the minimum 8185 margin width. 8186 (layout::start_annotation_line): Only print up to 3 of the 8187 margin character, to avoid touching the left-hand side. 8188 (selftest::test_diagnostic_show_locus_fixit_lines): Update for 8189 minimum margin width, as set by test_diagnostic_context's ctor. 8190 (selftest::test_fixit_insert_containing_newline): Likewise. 8191 (selftest::test_fixit_insert_containing_newline_2): Likewise. 8192 (selftest::test_line_numbers_multiline_range): Clear 8193 dc.min_margin_width. 8194 * diagnostic.c (diagnostic_initialize): Initialize 8195 min_margin_width. 8196 * diagnostic.h (struct diagnostic_context): Add field 8197 "min_margin_width". 8198 * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=. 8199 * opts.c (common_handle_option): Handle 8200 OPT_fdiagnostics_minimum_margin_width_. 8201 * selftest-diagnostic.c 8202 (selftest::test_diagnostic_context::test_diagnostic_context): 8203 Initialize min_margin_width to 6. 8204 * toplev.c (general_init): Initialize global_dc->min_margin_width. 8205 82062018-10-15 David Malcolm <dmalcolm@redhat.com> 8207 8208 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby): 8209 Fix usage of "error_at_rich_loc" in the comment. 8210 82112018-10-15 Renlin Li <renlin.li@arm.com> 8212 8213 PR target/87563 8214 * tree-vectorizer.c (try_vectorize_loop_1): Don't use 8215 if-conversioned loop when it contains ifn with types not 8216 supported by backend. 8217 * internal-fn.c (expand_direct_optab_fn): Add an assert. 8218 (direct_internal_fn_supported_p): New helper function. 8219 * internal-fn.h (direct_internal_fn_supported_p): Declare. 8220 82212018-10-15 Jakub Jelinek <jakub@redhat.com> 8222 8223 PR target/87572 8224 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET): 8225 Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and 8226 OPTION_MASK_ISA_AVX5124VNNIW_UNSET. 8227 82282018-10-15 Bin Cheng <bin.cheng@linux.alibaba.com> 8229 8230 PR tree-optimization/87022 8231 * tree-loop-distribution.c (pg_add_dependence_edges): Check all 8232 bits in dist vector rather than the first one. 8233 82342018-10-15 Richard Biener <rguenther@suse.de> 8235 8236 PR middle-end/87610 8237 * tree-ssa-structalias.c (struct vls_data): Add escaped_p member. 8238 (visit_loadstore): When a used restrict tag escaped verify that 8239 the points-to solution of "other" pointers do not include 8240 escaped. 8241 (compute_dependence_clique): If a used restrict tag escaped 8242 communicated that down to visit_loadstore. 8243 82442018-10-15 Andreas Krebbel <krebbel@linux.ibm.com> 8245 8246 * config/s390/s390.c (s390_expand_vec_init): Force vector element 8247 into reg if it isn't a general operand. 8248 82492018-10-14 H.J. Lu <hongjiu.lu@intel.com> 8250 8251 PR target/87599 8252 * config/i386/sse.md (*vec_dupv2di): Add register source to 8253 movddup. 8254 82552018-10-14 H.J. Lu <hongjiu.lu@intel.com> 8256 8257 PR target/87572 8258 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET): 8259 Add OPTION_MASK_ISA_AVX512IFMA_UNSET, 8260 OPTION_MASK_ISA_AVX5124FMAPS_UNSET and 8261 OPTION_MASK_ISA_AVX5124VNNIW_UNSET. 8262 82632018-10-13 Eric Botcazou <ebotcazou@adacore.com> 8264 8265 * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh. 8266 (notice_args_size): Set it in the current trace if no insn that can 8267 throw internally has been seen yet. 8268 (connect_traces): When connecting args_size between traces, allow the 8269 incoming values not to match if there is an insn setting it before the 8270 first insn that can throw internally; in that case, force the creation 8271 of a CFI note on this latter insn. 8272 82732018-10-13 Jonathan Wakely <jwakely@redhat.com> 8274 8275 * opt-problem.h (opt_wrapper): Use template-argument-list when naming 8276 the base class, because using the injected-class-name was not clearly 8277 specified until DR 176. 8278 82792018-10-12 Paul Koning <ni1d@arrl.net> 8280 8281 * config/pdp11/pdp11.md (doloop_end): New expander. 8282 (doloop_end_insn): renamed from "doloop_end". 8283 (addqi3): New pattern. 8284 (subqi3): New pattern. 8285 * config/pdp11/predicates.md (incdec_operand): New predicate. 8286 82872018-10-12 Yury Gribov <tetra2005@gmail.com> 8288 8289 PR middle-end/81376 8290 * real.c (format_helper::can_represent_integral_type_p): New function 8291 * real.h (format_helper::can_represent_integral_type_p): Ditto. 8292 * match.pd: New pattern. 8293 82942018-10-12 Alexandre Oliva <oliva@adacore.com> 8295 8296 * configure.ac: Introduce --enable-large-address-aware 8297 to define MINGW_DEFAULT_LARGE_ADDR_AWARE. 8298 * doc/install.texi: Document it. 8299 * configure, config.in: Rebuilt. 8300 * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define, 8301 based on MINGW_DEFAULT_LARGE_ADDR_AWARE. 8302 (LINK_SPEC): Insert it. 8303 * config/i386/mingw-w64.h: Likewise. 8304 8305 * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define. 8306 83072018-10-12 Peter Bergner <bergner@linux.ibm.com> 8308 8309 PR rtl-optimization/87600 8310 * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets. 8311 83122018-10-12 Paul Koning <ni1d@arrl.net> 8313 8314 * doc/md.texi (doloop_end): Document that the pattern code may 8315 need to check operand mode. 8316 83172018-10-12 Wilco Dijkstra <wdijkstr@arm.com> 8318 8319 * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives 8320 to zero-extend between int and floating-point registers. 8321 (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended 8322 ldp into floating-point registers. Add type and arch attributes. 8323 (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute. 8324 Use f_loads for type attribute. 8325 83262018-10-11 Martin Sebor <msebor@redhat.com> 8327 8328 * doc/extend.texi (attribute packed): Correct typos. 8329 83302018-10-11 Martin Sebor <msebor@redhat.com> 8331 8332 * doc/extend.texi (attribute flatten): Mention interaction with 8333 noinline. 8334 83352018-10-11 Jan Hubicka <hubicka@ucw.cz> 8336 8337 PR target/87156 8338 * cgraphclones.c (cgraph_node::create_version_clone_with_body): 8339 Set new_decl virtual flag to zero. 8340 83412018-10-11 Martin Sebor <msebor@redhat.com> 8342 8343 PR middle-end/87593 8344 * doc/extend.texi (attribute format_arg): Discuss using multiple 8345 attributes on a single function. 8346 83472018-10-11 Giuliano Belinassi <giuliano.belinassi@usp.br> 8348 8349 PR tree-optimization/86829 8350 * match.pd (sin (atan (x))): New simplification rules. 8351 (cos (atan (x))): Likewise. 8352 * real.c (build_sinatan_real): New function. 8353 * real.h (build_sinatan_real): Prototype. 8354 83552018-10-11 Will Schmidt <will_schmidt@vnet.ibm.com> 8356 8357 * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper 8358 function. 8359 (fold_mergeeo_helper): New helper function. 8360 (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo 8361 intrinsics. Correct some whitespace indentation issues. 8362 83632018-10-11 Wilco Dijkstra <wdijkstr@arm.com> 8364 8365 PR target/87511 8366 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p): 8367 Use HOST_WIDE_INT_1U for shift. 8368 83692018-10-11 Doug Rupp <rupp@adacore.com> 8370 Olivier Hainque <hainque@adacore.com> 8371 8372 * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro. 8373 Pass --relax to the linker for RTPs. 8374 (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC. 8375 83762018-10-11 Andrew Stubbs <ams@codesourcery.com> 8377 Jan Hubicka <jh@suse.cz> 8378 Martin Jambor <mjambor@suse.cz> 8379 8380 * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times 8381 the same elements are repeated rather than printing all of them. 8382 * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand 8383 "repeated" elements. 8384 * read-rtl-function.c (test_loading_repeat): New function. 8385 (read_rtl_function_c_tests): Call test_loading_repeat. 8386 * rtl-tests.c (test_dumping_repeat): New function. 8387 (rtl_tests_c_tests): Call test_dumping_repeat. 8388 83892018-10-11 Richard Biener <rguenther@suse.de> 8390 8391 * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset, 8392 bdver?_cost): Unify to ... 8393 (bdver_memcpy, bdver_memset, bdver_cost): ... this. 8394 * config/i386/i386.c (processor_cost_table): Adjust. 8395 83962018-10-10 Eric Botcazou <ebotcazou@adacore.com> 8397 8398 PR middle-end/87574 8399 * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on 8400 the thunk when expanding to GIMPLE. 8401 84022018-10-10 Bernd Edlinger <bernd.edlinger@hotmail.de> 8403 8404 PR bootstrap/87551 8405 * varasm.c (mergeable_string_section): Don't try to move zero-length 8406 strings to the merge section. 8407 84082018-10-10 Uros Bizjak <ubizjak@gmail.com> 8409 8410 PR target/87573 8411 * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter. 8412 84132018-10-10 Jakub Jelinek <jakub@redhat.com> 8414 8415 PR target/87550 8416 * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set 8417 to special_args set. 8418 84192018-10-10 Richard Biener <rguenther@suse.de> 8420 8421 * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df, 8422 reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf, 8423 reduc_plus_scal_v4sf): Merge into pattern reducing to half width 8424 and recursing and pattern terminating the recursion on SSE 8425 vector width using ix86_expand_reduc. 8426 (reduc_sminmax_scal_<mode>): Split into part reducing to half 8427 width and recursing and SSE2 vector variant doing the final 8428 reduction with ix86_expand_reduc. 8429 (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants 8430 with terminating the recursion at AVX level, splitting that 8431 to SSE there. 8432 84332018-10-09 David Malcolm <dmalcolm@redhat.com> 8434 8435 * genmatch.c (error_cb): Rename to... 8436 (diagnostic_cb): ...this, converting int params to enums. 8437 (fatal_at): Update for renaming. 8438 (warning_at): Likewise. 8439 (main): Likewise. 8440 * input.c (selftest::ebcdic_execution_charset::apply): 8441 Update for renaming of... 8442 (selftest::ebcdic_execution_charset::on_error): ...this, renaming 8443 to... 8444 (selftest::ebcdic_execution_charset::on_diagnostic): ...this, 8445 converting level and reason to enums. 8446 (class selftest::lexer_error_sink): Rename to... 8447 (class selftest::lexer_test_options): ...this, renaming field 8448 "m_errors" to "m_diagnostics". 8449 (selftest::lexer_test_options::apply): Update for renaming of... 8450 (selftest::lexer_test_options::on_error): ...this, renaming to... 8451 (selftest::lexer_test_options::on_diagnostic): ...this 8452 converting level and reason to enums. 8453 (selftest::test_lexer_string_locations_raw_string_unterminated): 8454 Update for renamings. 8455 * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for 8456 "reason". 8457 84582018-10-09 Paul A. Clarke <pc@us.ibm.com> 8459 8460 * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers. 8461 * config/rs6000/pmmintrin.h: New file. 8462 84632018-10-09 Eric Botcazou <ebotcazou@adacore.com> 8464 8465 PR tree-optimization/86659 8466 * gimple-match.h (gimple_match_op constructors): Initialize reverse. 8467 84682018-10-09 Richard Biener <rguenther@suse.de> 8469 8470 PR tree-optimization/63155 8471 * tree-ssa-structalias.c: Include tree-ssa.h. 8472 (get_constraint_for_ssa_var): For undefs return nothing_id. 8473 (find_func_aliases): Cleanup PHI handling. 8474 84752018-10-09 Richard Biener <rguenther@suse.de> 8476 8477 * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing 8478 replacements. 8479 84802018-10-09 Martin Liska <mliska@suse.cz> 8481 8482 * asan.c (asan_emit_stack_protection): If a stack variable 8483 is located in a same file as current function, then emit 8484 line info into variable definition string. 8485 84862018-10-08 Eric Botcazou <ebotcazou@adacore.com> 8487 8488 * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column 8489 information. 8490 84912018-10-08 Eric Botcazou <ebotcazou@adacore.com> 8492 8493 * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P 8494 on the thunk. 8495 84962018-10-08 Eric Botcazou <ebotcazou@adacore.com> 8497 8498 PR tree-optimization/86659 8499 * gimple-match.h (struct gimple_match_op): Add reverse field. 8500 (gimple_match_op::set_op): New overloaded method. 8501 * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set 8502 the REF_REVERSE_STORAGE_ORDER flag on the value. 8503 (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the 8504 REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set. 8505 85062018-10-08 Richard Sandiford <richard.sandiford@arm.com> 8507 8508 PR middle-end/63155 8509 * gimple-ssa-backprop.c (backprop::intersect_uses): Use 8510 FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT. 8511 85122018-10-08 H.J. Lu <hongjiu.lu@intel.com> 8513 8514 PR target/87517 8515 * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd): 8516 Defined with __builtin_ia32_vfmaddsubpd512_mask. 8517 85182018-10-08 Richard Biener <rguenther@suse.de> 8519 8520 * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads 8521 cost the same as AVX128 ones. 8522 85232018-10-08 Paul Koning <ni1d@arrl.net> 8524 8525 * config/pdp11/pdp11-protos.h (output_block_move): Remove. 8526 (expand_block_move): New function. 8527 * config/pdp11/pdp11.c (output_block_move): Remove. 8528 (expand_block_move): New function. 8529 * config/pdp11/pdp11.h (MOVE_RATIO): New definition. 8530 * config/pdp11/pdp11.md (movmemhi): Use expand_block_move. 8531 (*movmemhi1): Remove. 8532 85332018-10-08 Robin Dapp <rdapp@linux.ibm.com> 8534 8535 * config/s390/2827.md: Increase latencies for some FP instructions. 8536 85372018-10-08 Richard Biener <rguenther@suse.de> 8538 8539 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost): 8540 Open a dump scope. 8541 * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it. 8542 * tree-vectorizer.h (dump_stmt_cost): Adjust. 8543 (add_stmt_cost): Dump return value of the hook. 8544 85452018-10-08 Richard Biener <rguenther@suse.de> 8546 8547 PR tree-optimization/63155 8548 * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first. 8549 (ssa_propagation_engine::ssa_propagate): Remove redundant 8550 bitmap bit clearing. 8551 85522018-10-05 Peter Bergner <bergner@linux.ibm.com> 8553 8554 PR rtl-optimization/86939 8555 PR rtl-optimization/87479 8556 * ira.h (non_conflicting_reg_copy_p): New prototype. 8557 * ira-lives.c (ignore_reg_for_conflicts): New static variable. 8558 (make_hard_regno_dead): Don't add conflicts for register 8559 ignore_reg_for_conflicts. 8560 (make_object_dead): Likewise. 8561 (non_conflicting_reg_copy_p): New function. 8562 (process_bb_node_lives): Set ignore_reg_for_conflicts for copies. 8563 Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM. 8564 * lra-lives.c (ignore_reg_for_conflicts): New static variable. 8565 (make_hard_regno_dead): Don't add conflicts for register 8566 ignore_reg_for_conflicts. Remove special conflict handling of 8567 REAL_PIC_OFFSET_TABLE_REGNUM. Remove now unused argument 8568 check_pic_pseudo_p and update callers. 8569 (mark_pseudo_dead): Don't add conflicts for register 8570 ignore_reg_for_conflicts. 8571 (process_bb_lives): Set ignore_reg_for_conflicts for copies. 8572 85732018-10-05 Andrew Waterman <andrew@sifive.com> 8574 Jim Wilson <jimw@sifive.com> 8575 8576 * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4): 8577 Add define_expand. Add ! HONOR_SNANS check to current pattern. Add 8578 new pattern using HONOR_SNANS that emits one extra instruction. 8579 85802018-10-05 Segher Boessenkool <segher@kernel.crashing.org> 8581 8582 * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator 8583 patterns): Merge SI and DI patterns to a GPR pattern. 8584 (unnamed define_insn and define_split for record form of that): Merge 8585 to a single define_insn_and_split pattern. 8586 85872018-10-05 David Malcolm <dmalcolm@redhat.com> 8588 8589 PR c++/56856 8590 * input.c (expand_location_to_spelling_point): Add param "aspect" 8591 and use rather than hardcoding LOCATION_ASPECT_CARET. 8592 (get_substring_ranges_for_loc): Handle the case of a single token 8593 within a macro expansion. 8594 * input.h (expand_location_to_spelling_point): Add "aspect" param, 8595 defaulting to LOCATION_ASPECT_CARET. 8596 85972018-10-05 Paul Koning <ni1d@arrl.net> 8598 8599 * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define. 8600 (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define. 8601 (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define. 8602 (pdp11_guard_type): New function. 8603 86042018-10-05 Paul Koning <ni1d@arrl.net> 8605 8606 * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32. 8607 * config/pdp11/pdp11.opt (mfloat32): Remove. 8608 (mfloat64): Remove. 8609 * doc/invoke.texi (pdp11 -mfloat32): Remove: 8610 (pdp11 -mfloat64): Remove. 8611 86122018-10-05 Uros Bizjak <ubizjak@gmail.com> 8613 8614 * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern. 8615 (*cmp<mode>_cc_i387): Ditto. 8616 (*cmpu<mode>_cc_i387): Ditto. 8617 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto. 8618 * config/i386/i386.c (ix86_expand_fp_compare): Remove 8619 "scratch" argument. 8620 <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber. 8621 Emit x86_sahf_1 pattern. 8622 (ix86_expand_compare): Update call to ix86_expand_fp_compare. 8623 (ix86_expand_carry_flag_compare): Ditto. 8624 86252018-10-05 Uros Bizjak <ubizjak@gmail.com> 8626 8627 * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate 8628 to reg_or_0_operand. Add "C" constraint. 8629 (*cmpxf_cc_i387): Ditto. 8630 (*cmp<mode>_i387): Change operand 2 predicate 8631 to nonimm_or_0_operand. Add "C" constraint. 8632 (*cmp<mode>_cc_i387): Ditto. 8633 (*cmp<mode>_0_i387): Remove insn pattern. 8634 (*cmp<mode>_0_cc_i387): Ditto. 8635 86362018-10-05 Uros Bizjak <ubizjak@gmail.com> 8637 8638 * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE. 8639 * config/i386/predicates.md (nonimm_or_0_operand): Rename 8640 from vector_move_operand. Update all uses. 8641 86422018-10-05 Martin Sebor <msebor@redhat.com> 8643 8644 PR tree-optimization/87490 8645 * builtins.c (expand_builtin_strnlen): Handle a null data.decl 8646 consistently. 8647 86482018-10-05 Richard Biener <rguenther@suse.de> 8649 8650 PR tree-optimization/63155 8651 * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess 8652 vertical space in dumpfiles. 8653 * tree-ssa-propagate.h 8654 (ssa_propagation_engine::process_ssa_edge_worklist): Remove. 8655 * tree-ssa-propagate.c (cfg_blocks_back): New global. 8656 (ssa_edge_worklist_back): Likewise. 8657 (curr_order): Likewise. 8658 (cfg_blocks_get): Remove abstraction. 8659 (cfg_blocks_add): Likewise. 8660 (cfg_blocks_empty_p): Likewise. 8661 (add_ssa_edge): Add to current or next worklist based on 8662 RPO index. 8663 (add_control_edge): Likewise. 8664 (ssa_propagation_engine::process_ssa_edge_worklist): Fold 8665 into ... 8666 (ssa_propagation_engine::ssa_propagate): ... here. Unify 8667 iteration from CFG and SSA edge worklist so we process 8668 everything in RPO order, prioritizing forward progress 8669 over iteration. 8670 (ssa_prop_init): Allocate new worklists, do not dump 8671 immediate uses. 8672 (ssa_prop_fini): Free new worklists. 8673 86742018-10-05 Richard Biener <rguenther@suse.de> 8675 8676 * tree-core.h (tree_block::abstract_flag): Remove. 8677 (tree_block::block_num): Make full 32bits. 8678 * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT. 8679 * tree.h (BLOCK_ABSTRACT): Remove. 8680 * dwarf2out.c (gen_lexical_block_die): Remove dead code 8681 resulting from BLOCK_ABSTRACT being always false. 8682 (gen_inlined_subroutine_die): Likewise. 8683 (gen_block_die): Likewise. 8684 * tree.c (block_ultimate_origin): Likewise. 8685 * tree-pretty-print.c (dump_block_node): Remove code dealing 8686 with BLOCK_ABSTRACT. 8687 * tree-ssa-live.c (dump_scope_block): Likewise. 8688 * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise. 8689 * tree-streamer-out.c (pack_ts_block_value_fields): Likewise. 8690 86912018-10-05 Richard Biener <rguenther@suse.de> 8692 8693 * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost 8694 is asked for initialize mode to the component mode of the 8695 vector type. 8696 86972018-10-05 H.J. Lu <hongjiu.lu@intel.com> 8698 8699 PR target/87522 8700 * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to 8701 assembler for -mavx. 8702 * config/i386/gnu-user64.h (ASM_SPEC): Likewise. 8703 87042018-10-05 Segher Boessenkool <segher@kernel.crashing.org> 8705 8706 PR target/87509 8707 * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use 8708 RS6000_BTM_DFP. 8709 * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand 8710 to be DImode. When using mffscrn, force the operand to a register. 8711 87122018-10-04 Uros Bizjak <ubizjak@gmail.com> 8713 8714 * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn 8715 from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using 8716 X87MODEF mode iterator. 8717 (*fop_<X87MODEF:mode>_3_i387): Macroize insn from 8718 *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using 8719 X87MODEF mode iterator. 8720 87212018-10-04 Vinay Kumar <vinay.kumar@blackfigtech.com> 8722 8723 * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning 8724 -Wno-prio-ctor-dtor. 8725 87262018-10-04 David Malcolm <dmalcolm@redhat.com> 8727 8728 * Makefile.in (OBJS): Add opt-problem.o. 8729 * dump-context.h: Include "selftest.h. 8730 (selftest::temp_dump_context): New forward decl. 8731 (class dump_context): Make friend of class 8732 selftest::temp_dump_context. 8733 (dump_context::dump_loc_immediate): New decl. 8734 (class dump_pretty_printer): Move here from dumpfile.c. 8735 (class temp_dump_context): Move to namespace selftest. 8736 (temp_dump_context::temp_dump_context): Add param 8737 "forcibly_enable_dumping". 8738 (selftest::verify_dumped_text): 8739 (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c. 8740 (selftest::verify_item): 8741 (ASSERT_IS_TEXT): Move here from dumpfile.c. 8742 (ASSERT_IS_TREE): Likewise. 8743 (ASSERT_IS_GIMPLE): Likewise. 8744 * dumpfile.c (dump_context::dump_loc): Move immediate dumping 8745 to... 8746 (dump_context::dump_loc_immediate): ...this new function. 8747 (class dump_pretty_printer): Move to dump-context.h. 8748 (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED. 8749 (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED. 8750 (temp_dump_context::temp_dump_context): Move to "selftest" 8751 namespace. Add param "forcibly_enable_dumping", and use it to 8752 conditionalize the use of m_pp; 8753 (selftest::verify_dumped_text): Make non-static. 8754 (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h. 8755 (selftest::verify_item): Make non-static. 8756 (ASSERT_IS_TEXT): Move to dump-context.h. 8757 (ASSERT_IS_TREE): Likewise. 8758 (ASSERT_IS_GIMPLE): Likewise. 8759 (selftest::test_capture_of_dump_calls): Pass "true" for new 8760 param of temp_dump_context. 8761 * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding 8762 it to MSG_ALL_PRIORITIES. Update values of TDF_COMPARE_DEBUG and 8763 TDF_COMPARE_DEBUG. 8764 * opt-problem.cc: New file. 8765 * opt-problem.h: New file. 8766 * optinfo-emit-json.cc 8767 (selftest::test_building_json_from_dump_calls): Pass "true" for 8768 new param of temp_dump_context. 8769 * optinfo.cc (optinfo_kind_to_dump_flag): New function. 8770 (optinfo::emit_for_opt_problem): New function. 8771 (optinfo::emit): Clarity which emit_item is used. 8772 * optinfo.h (optinfo::get_dump_location): New accessor. 8773 (optinfo::emit_for_opt_problem): New decl. 8774 (optinfo::emit): Make const. 8775 * selftest-run-tests.c (selftest::run_tests): Call 8776 selftest::opt_problem_cc_tests. 8777 * selftest.h (selftest::opt_problem_cc_tests): New decl. 8778 * tree-data-ref.c (dr_analyze_innermost): Convert return type from 8779 bool to opt_result, converting fprintf messages to 8780 opt_result::failure_at calls. Add "stmt" param for use by the 8781 failure_at calls. 8782 (create_data_ref): Pass "stmt" to the dr_analyze_innermost call. 8783 (runtime_alias_check_p): Convert return type from bool to 8784 opt_result, converting dump_printf calls to 8785 opt_result::failure_at, using the statement DDR_A for their 8786 location. 8787 (find_data_references_in_stmt): Convert return type from bool to 8788 opt_result, converting "return false" to opt_result::failure_at 8789 with a new message. 8790 * tree-data-ref.h: Include "opt-problem.h". 8791 (dr_analyze_innermost): Convert return type from bool to opt_result, 8792 and add a const gimple * param. 8793 (find_data_references_in_stmt): Convert return type from bool to 8794 opt_result. 8795 (runtime_alias_check_p): Likewise. 8796 * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to 8797 dr_analyze_innermost. 8798 * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test): 8799 Convert return type from bool to opt_result, adding a message for 8800 the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case. 8801 (vect_analyze_data_ref_dependence): Convert return type from bool 8802 to opt_result. Change sense of return type from "false" 8803 effectively meaning "no problems" to "false" meaning a problem, 8804 so that "return false" becomes "return opt_result::success". 8805 Convert "return true" calls to opt_result::failure_at, using 8806 the location of statement A rather than vect_location. 8807 (vect_analyze_data_ref_dependences): Convert return type from bool 8808 to opt_result. 8809 (verify_data_ref_alignment): Likewise, converting dump_printf_loc 8810 calls to opt_result::failure_at, using the stmt location rather 8811 than vect_location. 8812 (vect_verify_datarefs_alignment): Convert return type from bool 8813 to opt_result. 8814 (vect_enhance_data_refs_alignment): Likewise. Split local "stat" 8815 into multiple more-tightly-scoped copies. 8816 (vect_analyze_data_refs_alignment): Convert return type from bool 8817 to opt_result. 8818 (vect_analyze_data_ref_accesses): Likewise, converting a 8819 "return false" to a "return opt_result::failure_at", adding a 8820 new message. 8821 (vect_prune_runtime_alias_test_list): Convert return type from 8822 bool to opt_result, converting dump_printf_loc to 8823 opt_result::failure_at. Add a %G to show the pertinent statement, 8824 and use the stmt's location rather than vect_location. 8825 (vect_find_stmt_data_reference): Convert return type from 8826 bool to opt_result, converting dump_printf_loc to 8827 opt_result::failure_at, using stmt's location. 8828 (vect_analyze_data_refs): Convert return type from bool to 8829 opt_result. Convert "return false" to "return 8830 opt_result::failure_at", adding messages as needed. 8831 * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return 8832 type from bool to opt_result. 8833 (vect_determine_vf_for_stmt): Likewise. 8834 (vect_determine_vectorization_factor): Likewise, converting 8835 dump_printf_loc to opt_result::failure_at, using location of phi 8836 rather than vect_location. 8837 (vect_analyze_loop_form_1): Convert return type from bool to 8838 opt_result, converting dump_printf_loc calls, retaining the use of 8839 vect_location. 8840 (vect_analyze_loop_form): Convert return type from loop_vec_info 8841 to opt_loop_vec_info. 8842 (vect_analyze_loop_operations): Convert return type from bool to 8843 opt_result, converting dump_printf_loc calls, using the location 8844 of phi/stmt rather than vect_location where available. Convert 8845 various "return false" to "return opt_result::failure_at" with 8846 "unsupported phi" messages. 8847 (vect_get_datarefs_in_loop): Convert return type from bool to 8848 opt_result. Add a message for the 8849 PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure. 8850 (vect_analyze_loop_2): Convert return type from bool to 8851 opt_result. Ensure "ok" is set to a opt_result::failure_at before 8852 each "goto again;", adding new messages where needed. 8853 Add "unsupported grouped {store|load}" messages. 8854 (vect_analyze_loop): Convert return type from loop_vec_info to 8855 opt_loop_vec_info. 8856 * tree-vect-slp.c (vect_analyze_slp): Convert return type from 8857 bool to opt_result. 8858 * tree-vect-stmts.c (process_use): Likewise, converting 8859 dump_printf_loc call and using stmt location, rather than 8860 vect_location. 8861 (vect_mark_stmts_to_be_vectorized): Likeise. 8862 (vect_analyze_stmt): Likewise, adding a %G. 8863 (vect_get_vector_types_for_stmt): Convert return type from bool to 8864 opt_result, converting dump_printf_loc calls and using stmt 8865 location, rather than vect_location. 8866 (vect_get_mask_type_for_stmt): Convert return type from tree to 8867 opt_tree, converting dump_printf_loc calls and using stmt location. 8868 * tree-vectorizer.c: Include "opt-problem.h. 8869 (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as 8870 MSG_PRIORITY_INTERNALS. Convert local "loop_vinfo" from 8871 loop_vec_info to opt_loop_vec_info. If if fails, and dumping is 8872 enabled, use it to report at the top level "couldn't vectorize 8873 loop" followed by the problem. 8874 * tree-vectorizer.h (opt_loop_vec_info): New typedef. 8875 (vect_mark_stmts_to_be_vectorized): Convert return type from bool 8876 to opt_result. 8877 (vect_analyze_stmt): Likewise. 8878 (vect_get_vector_types_for_stmt): Likewise. 8879 (tree vect_get_mask_type_for_stmt): Likewise. 8880 (vect_analyze_data_ref_dependences): Likewise. 8881 (vect_enhance_data_refs_alignment): Likewise. 8882 (vect_analyze_data_refs_alignment): Likewise. 8883 (vect_verify_datarefs_alignment): Likewise. 8884 (vect_analyze_data_ref_accesses): Likewise. 8885 (vect_prune_runtime_alias_test_list): Likewise. 8886 (vect_find_stmt_data_reference): Likewise. 8887 (vect_analyze_data_refs): Likewise. 8888 (vect_analyze_loop): Convert return type from loop_vec_info to 8889 opt_loop_vec_info. 8890 (vect_analyze_loop_form): Likewise. 8891 (vect_analyze_slp): Convert return type from bool to opt_result. 8892 88932018-10-04 David Malcolm <dmalcolm@redhat.com> 8894 8895 * doc/invoke.texi (-fopt-info): Document new "internals" 8896 sub-option. 8897 * dump-context.h (dump_context::apply_dump_filter_p): New decl. 8898 * dumpfile.c (dump_options): Update for renaming of MSG_ALL to 8899 MSG_ALL_KINDS. 8900 (optinfo_verbosity_options): Add "internals". 8901 (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS. 8902 (dump_context::apply_dump_filter_p): New member function. 8903 (dump_context::dump_loc): Use apply_dump_filter_p rather than 8904 explicitly masking the dump_kind. 8905 (dump_context::begin_scope): Increment the scope depth first. Use 8906 apply_dump_filter_p rather than explicitly masking the dump_kind. 8907 (dump_context::emit_item): Use apply_dump_filter_p rather than 8908 explicitly masking the dump_kind. 8909 (dump_dec): Likewise. 8910 (dump_hex): Likewise. 8911 (dump_switch_p_1): Default to MSG_ALL_PRIORITIES. 8912 (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING. 8913 (opt_info_switch_p): Update handling of default 8914 MSG_OPTIMIZED_LOCATIONS to cope with default of 8915 MSG_PRIORITY_USER_FACING. 8916 (dump_basic_block): Use apply_dump_filter_p rather than explicitly 8917 masking the dump_kind. 8918 (selftest::test_capture_of_dump_calls): Update test_dump_context 8919 instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather 8920 than MSG_ALL. Generalize scope test to be run at all four 8921 combinations of with/without MSG_PRIORITY_USER_FACING and 8922 MSG_PRIORITY_INTERNALS, adding examples of explicit priority 8923 for each of the two values. 8924 * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags. 8925 Rename MSG_ALL to MSG_ALL_KINDS. Add MSG_PRIORITY_USER_FACING, 8926 MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the 8927 values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES. 8928 (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction 8929 with MSG_PRIORITY_*. 8930 * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning 8931 dump messages as MSG_PRIORITY_USER_FACING. 8932 * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment 8933 about the interaction with MSG_PRIORITY_*. 8934 89352018-10-04 Bernd Edlinger <bernd.edlinger@hotmail.de> 8936 8937 * varasm.c (output_constant): Add new parameter merge_strings. 8938 Make strings properly zero terminated in merge string sections. 8939 (mergeable_string_section): Don't fail if the last char is non-zero. 8940 (assemble_variable_contents): Handle merge string sections. 8941 (assemble_variable): Likewise. 8942 (assemble_constant_contents): Likewise. 8943 (output_constant_def_contents): Likewise. 8944 (output_constructor_array_range, 8945 output_constructor_regular_field): Adjust call to output_constant. 8946 (output_object_block): Adjust call to assemble_constant_contents 8947 and assemble_variable_contents. 8948 89492018-10-04 Martin Liska <mliska@suse.cz> 8950 8951 PR c/87483 8952 * cgraphunit.c (process_function_and_variable_attributes): 8953 Warn about a function with alias attribute and a body. 8954 89552018-10-04 Martin Liska <mliska@suse.cz> 8956 8957 PR ipa/82625 8958 * multiple_target.c (redirect_to_specific_clone): New function. 8959 (ipa_target_clone): Use it. 8960 * tree-inline.c: Fix comment. 8961 89622018-10-04 David Malcolm <dmalcolm@redhat.com> 8963 8964 * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new 8965 fields. 8966 (gcc::dump_manager::~dump_manager): Free m_optinfo_filename. 8967 (gcc::dump_manager::register_pass): New member function, adapted 8968 from loop body in gcc::pass_manager::register_pass, adding a 8969 call to update_dfi_for_opt_info. 8970 (gcc::dump_manager::opt_info_enable_passes): Store the 8971 -fopt-info options into the new fields. Move the loop 8972 bodies into... 8973 (gcc::dump_manager::update_dfi_for_opt_info): ...this new member 8974 function. 8975 * dumpfile.h (struct opt_pass): New forward decl. 8976 (gcc::dump_manager::register_pass): New decl. 8977 (gcc::dump_manager::update_dfi_for_opt_info): New decl. 8978 (class gcc::dump_manager): Add fields "m_optgroup_flags", 8979 "m_optinfo_flags", and "m_optinfo_filename". 8980 * passes.c (gcc::pass_manager::register_pass): Move all of the 8981 dump-handling code to gcc::dump_manager::register_pass. 8982 89832018-10-04 Peter Bergner <bergner@linux.ibm.com> 8984 8985 PR rtl-optimization/87466 8986 * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook. 8987 * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook. 8988 * doc/tm.texi: Regenerate. 8989 * ira-lives.c (process_bb_node_lives): Use the new target hook. 8990 * lra-lives.c (process_bb_lives): Likewise. 8991 * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): 8992 Define. 8993 89942018-10-04 Tamar Christina <tamar.christina@arm.com> 8995 8996 * params.c (add_params): Fix initialization. 8997 89982018-10-04 Martin Liska <mliska@suse.cz> 8999 9000 PR gcov-profile/84107 9001 * tree-profile.c (init_ic_make_global_vars): 9002 Remove ic_void_ptr_var and ic_gcov_type_ptr_var. 9003 Come up with new ic_tuple* variables. Emit 9004 __gcov_indirect_call{,_topn} variables. 9005 (gimple_gen_ic_profiler): Access the variable 9006 and emit gimple. 9007 (gimple_gen_ic_func_profiler): Access 9008 __gcov_indirect_call.callee field. 9009 (gimple_init_gcov_profiler): Use ptr_type_node. 9010 * value-prof.c (gimple_ic): Use ptr_type_node. 9011 90122018-10-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org> 9013 9014 PR tree-optimization/85787 9015 * ipa-pure-const.c (malloc_candidate_p_1): Move most of 9016 malloc_candidate_p into this function and add support for 9017 detecting multiple phis. 9018 (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro. 9019 90202018-10-04 Martin Liska <mliska@suse.cz> 9021 9022 PR ipa/87491 9023 * ipa-inline.c (inline_to_all_callers_1): 9024 Call ultimate_alias_target for node being inlined. 9025 90262018-10-03 Jeff Law <law@redhat.com> 9027 9028 * gimple-ssa-sprintf.c (format_string): Do not hardcode size of 9029 target's wchar_t. 9030 * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c. 9031 * tree.h (get_typenode_from_name): Prototype. 9032 90332018-10-03 Uros Bizjak <ubizjak@gmail.com> 9034 9035 * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387): 9036 Change operand 2 predicate to nonimmediate_operand. 9037 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto. 9038 90392018-10-03 Martin Sebor <msebor@redhat.com> 9040 Jeff Law <law@redhat.com> 9041 9042 * gimple-ssa-sprintf.c (struct fmtresult): Add new member and 9043 initialize it. 9044 (get_string_length): Detect unterminated arrays. 9045 (format_string): Same. 9046 (format_directive): Warn about unterminated arrays. 9047 (handle_gimple_call): Mark statements with no_warning as needed. 9048 90492018-10-03 Jim Wilson <jimw@sifive.com> 9050 9051 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E, 9052 also define __riscv_abi_rve. Delete trailing white space. 9053 90542018-10-03 Paul Koning <ni1d@arrl.net> 9055 9056 Enable LRA register allocator for PDP11. 9057 * config/pdp11/constraints.md (Q): Use define_memory_constraint. 9058 (R): Likewise. 9059 (D): Likewise. 9060 * config/pdp11/pdp11.c (pdp11_lra_p): New function. 9061 * config/pdp11/pdp11.opt (-mlra): New option. 9062 * doc/invoke.texi (PDP-11 Options): Document -mlra. 9063 90642018-10-03 Uros Bizjak <ubizjak@gmail.com> 9065 9066 * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove. 9067 (*<absneg:code>extend<mode>xf2): Ditto. 9068 90692018-10-03 Aldy Hernandez <aldyh@redhat.com> 9070 9071 PR tree-optimization/87415 9072 * tree-vrp.c (set_value_range_with_overflow): Special case one bit 9073 precision fields. 9074 90752018-10-02 Jeff Law <law@redhat.com> 9076 9077 * gimple-fold.c (get_range_strlen): Only set *nonstr when 9078 an unterminated string is discovered. Bubble up range 9079 even for unterminated strings. 9080 (gimple_fold_builtin_strlen): Do not fold if get_range_strlen 9081 indicates the string was not terminated via NONSTR. 9082 90832018-10-03 Aldy Hernandez <aldyh@redhat.com> 9084 9085 * tree-vrp.c (extract_range_from_unary_expr): Special case all 9086 pointer conversions. 9087 Do not do anything special for anti-ranges. 9088 90892018-10-03 Jérôme Lambourg <lambourg@adacore.com> 9090 9091 * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to 9092 DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs. 9093 90942018-10-03 Martin Liska <mliska@suse.cz> 9095 9096 PR gcov-profile/86109 9097 * coverage.c (coverage_begin_function): Do not 9098 mark lambdas as artificial. 9099 * tree-core.h (struct GTY): Remove tm_clone_flag 9100 and introduce new lambda_function. 9101 * tree.h (DECL_LAMBDA_FUNCTION): New macro. 9102 91032018-10-02 Aaron Sawdey <acsawdey@linux.ibm.com> 9104 9105 PR target/87474 9106 * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both 9107 P8_VECTOR and VSX are enabled. 9108 91092018-10-02 Andreas Krebbel <krebbel@linux.ibm.com> 9110 9111 * config/s390/driver-native.c (s390_host_detect_local_cpu): Add 9112 0x3907 as CPU model number. 9113 91142018-10-02 Andreas Krebbel <krebbel@linux.ibm.com> 9115 9116 * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14. 9117 * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to 9118 PF_Z14. Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14, 9119 TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to 9120 TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P. 9121 * config/s390/s390.md: Likewise. Rename also the cpu attribute 9122 value from arch12 to z14. 9123 91242018-10-02 Uros Bizjak <ubizjak@gmail.com> 9125 9126 * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove. 9127 (isinfxf2): Ditto. 9128 (isinf<mode>2): Ditto. 9129 91302018-10-02 Uros Bizjak <ubizjak@gmail.com> 9131 9132 * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode 9133 before emitting fxam. Perform calculations in XFmode. 9134 91352018-10-02 Marc Glisse <marc.glisse@inria.fr> 9136 9137 * match.pd (((X /[ex] A) +- B) * A): New transformation. 9138 91392018-10-02 Marc Glisse <marc.glisse@inria.fr> 9140 9141 PR middle-end/87319 9142 * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors. 9143 * tree.c (signed_or_unsigned_type_for): Handle complex. 9144 91452018-10-02 Jeff Law <law@redhat.com> 9146 9147 * gimple-fold.c (get_range_strlen): Remove dead code. 9148 91492018-10-02 Martin Sebor <msebor@redhat.com> 9150 Jeff Law <law@redhat.com> 9151 9152 * builtins.c (unterminated_array): Add new arguments. 9153 If argument is not terminated, bubble up size and exact 9154 state to callers. 9155 (expand_builtin_strnlen): Detect, avoid expanding 9156 and diagnose unterminated arrays. 9157 (c_strlen): Fill in offset of start of unterminated strings. 9158 * builtins.h (unterminated_array): Update prototype. 9159 91602018-10-02 Richard Biener <rguenther@suse.de> 9161 9162 * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use 9163 of haddv4df, first reduce to SSE width and exploit the fact 9164 that we only need element zero with the reduction result. 9165 (reduc_plus_scal_v2df): Likewise. 9166 91672018-10-02 Eric Botcazou <ebotcazou@adacore.com> 9168 9169 * dojump.h (do_jump): Delete. 9170 (do_jump_1): Likewise. 9171 (split_comparison): Move around. 9172 * dojump.c (do_jump): Make static. 9173 (do_jump_1): Likewise. 9174 (jumpifnot): Move around. 9175 (jumpifnot_1): Likewise. 9176 (jumpif): Likewise. 9177 (jumpif_1): Likewise. 9178 * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump. 9179 91802018-10-02 Eric Botcazou <ebotcazou@adacore.com> 9181 9182 * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the 9183 insns in the delay slot and add_insn_after for the jump insn. 9184 91852018-10-02 Richard Biener <rguenther@suse.de> 9186 9187 * tree-inline.c (expand_call_inline): Use the location of 9188 the callee declaration for the inline-entry marker. 9189 * final.c (notice_source_line): Remove special-casing of 9190 NOTE_INSN_INLINE_ENTRY. 9191 91922018-10-01 Carl Love <cel@us.ibm.com> 9193 9194 PR 69431 9195 * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New. 9196 (__builtin_mtfsb0): New. 9197 (__builtin_mtfsb1): New. 9198 ( __builtin_set_fpscr_rn): New. 9199 (__builtin_set_fpscr_drn): New. 9200 * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add. 9201 (rs6000_expand_set_fpscr_rn_builtin): Add. 9202 (rs6000_expand_set_fpscr_drn_builtin): Add. 9203 (rs6000_expand_builtin): Add case statement entries for 9204 RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1, 9205 RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN, 9206 RS6000_BUILTIN_MFFSL. 9207 (rs6000_init_builtins): Add ftype initialization and def_builtin 9208 calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1, 9209 __builtin_set_fpscr_rn, __builtin_set_fpscr_drn. 9210 * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn, 9211 rs6000_mffscdrn): Add define_insn. 9212 (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand. 9213 * doc/extend.texi: Add documentation for the builtins. 9214 92152018-10-01 Richard Biener <rguenther@suse.de> 9216 9217 PR tree-optimization/87465 9218 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo 9219 causing branch miscounts. 9220 92212018-10-01 Tamar Christina <tamar.christina@arm.com> 9222 9223 * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM, 9224 aarch64_option_default_param): New. 9225 (params.h): Include. 9226 (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New. 9227 * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify 9228 stack-clash protection validation code. 9229 92302018-10-01 Tamar Christina <tamar.christina@arm.com> 9231 9232 * params.c (validate_param): New. 9233 (add_params): Use it. 9234 (set_param_value): Refactor param validation into validate_param. 9235 (diagnostic.h): Include. 9236 * diagnostic.h (diagnostic_ready_p): New. 9237 92382018-10-01 Tamar Christina <tamar.christina@arm.com> 9239 9240 * params.c (set_param_value): 9241 Add index of parameter being validated. 9242 * common/common-target.def (option_validate_param): New. 9243 * common/common-targhooks.h (default_option_validate_param): New. 9244 * common/common-targhooks.c (default_option_validate_param): New. 9245 * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New. 9246 * doc/tm.texi: Regenerate. 9247 92482018-10-01 Tamar Christina <tamar.christina@arm.com> 9249 9250 PR target/86486 9251 * config/aarch64/aarch64.c (aarch64_override_options_internal): 9252 Add validation for stack-clash parameters and set defaults. 9253 92542018-10-01 Tamar Christina <tamar.christina@arm.com> 9255 9256 PR target/86486 9257 * configure.ac: Add stack-clash-protection-guard-size. 9258 * doc/install.texi: Document it. 9259 * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New. 9260 * params.def: Update comment for guard-size. 9261 (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE, 9262 PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description. 9263 * configure: Regenerate. 9264 92652018-10-01 Tamar Christina <tamar.christina@arm.com> 9266 9267 PR target/86486 9268 * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS, 9269 STACK_DYNAMIC_OFFSET): New. 9270 * config/aarch64/aarch64.c (aarch64_layout_frame): 9271 Update outgoing args size. 9272 (aarch64_stack_clash_protection_alloca_probe_range, 9273 TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New. 9274 92752018-10-01 Tamar Christina <tamar.christina@arm.com> 9276 9277 PR target/86486 9278 * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom 9279 probe ranges. 9280 * target.def (stack_clash_protection_alloca_probe_range): New. 9281 (stack_clash_protection_final_dynamic_probe): Remove. 9282 * targhooks.h (default_stack_clash_protection_alloca_probe_range) New. 9283 (default_stack_clash_protection_final_dynamic_probe): Remove. 9284 * targhooks.c: Likewise. 9285 * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New. 9286 (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove. 9287 * doc/tm.texi: Regenerate. 9288 92892018-10-01 Tamar Christina <tamar.christina@arm.com> 9290 9291 PR target/86486 9292 * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New. 9293 * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash, 9294 aarch64_clamp_to_uimm12_shift): New. 9295 (aarch64_allocate_and_probe_stack_space): Add SVE specific section. 9296 * config/aarch64/aarch64.md (probe_sve_stack_clash): New. 9297 92982018-10-01 Tamar Christina <tamar.christina@arm.com> 9299 9300 PR target/86486 9301 * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert. 9302 93032018-10-01 Jeff Law <law@redhat.com> 9304 Richard Sandiford <richard.sandiford@linaro.org> 9305 Tamar Christina <tamar.christina@arm.com> 9306 9307 PR target/86486 9308 * config/aarch64/aarch64.md 9309 (probe_stack_range): Add k (SP) constraint. 9310 * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD, 9311 STACK_CLASH_MAX_UNROLL_PAGES): New. 9312 * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit 9313 stack probes for stack clash. 9314 (aarch64_allocate_and_probe_stack_space): New. 9315 (aarch64_expand_prologue): Use it. 9316 (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria. 9317 (aarch64_sub_sp): Add emit_move_imm optional param. 9318 93192018-10-01 MCC CS <deswurstes@users.noreply.github.com> 9320 9321 PR tree-optimization/87261 9322 * match.pd: Remove trailing whitespace. 9323 Add (x & y) | ~(x | y) -> ~(x ^ y), 9324 (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y) 9325 93262018-10-01 Claudiu Zissulescu <claziss@synopsys.com> 9327 9328 * config/arc/arc.md (*add_n): Clean up pattern, update instruction 9329 constraints. 9330 (ashlsi3_insn): Update instruction constraints. 9331 (ashrsi3_insn): Likewise. 9332 (rotrsi3): Likewise. 9333 (add_shift): Likewise. 9334 * config/arc/constraints.md (Csz): New 32 bit constraint. It 9335 avoids placing in the limm field small constants which, otherwise, 9336 could end into a small instruction. 9337 93382018-10-01 Claudiu Zissulescu <claziss@synopsys.com> 9339 9340 * config/arc/arc.md (maddsidi4_split): Don't use dmac if the 9341 destination register is not odd-even. 9342 (umaddsidi4_split): Likewise. 9343 93442018-10-01 Richard Biener <rguenther@suse.de> 9345 9346 * tree-inline.c (expand_call_inline): Store origin of fn 9347 in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK. 9348 * tree.c (block_ultimate_origin): Simplify and do some 9349 checking. 9350 93512018-09-30 Uros Bizjak <ubizjak@gmail.com> 9352 9353 * config/i386/mmx.md (EMMS): New int iterator. 9354 (emms): New int attribute. 9355 (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using 9356 EMMS int iterator. Explicitly declare clobbers. 9357 (mmx_emms): Remove expander. 9358 (mmx_femms): Ditto. 9359 * config/i386/predicates.md (emms_operation): Remove predicate. 9360 (vzeroall_pattern): New predicate. 9361 (vzeroupper_pattern): Rename from vzeroupper_operation. 9362 * config/i386/i386.c (ix86_avx_u128_mode_after): Use 9363 vzeroupper_pattern and vzeroall_pattern predicates. 9364 93652018-09-30 Peter Bergner <bergner@linux.ibm.com> 9366 9367 PR rtl-optimization/86939 9368 * ira-lives.c (make_hard_regno_born): Rename from this... 9369 (make_hard_regno_live): ... to this. Remove update to conflict 9370 information. Update function comment. 9371 (make_hard_regno_dead): Add conflict information update. Update 9372 function comment. 9373 (make_object_born): Rename from this... 9374 (make_object_live): ... to this. Remove update to conflict information. 9375 Update function comment. 9376 (make_object_dead): Add conflict information update. Update function 9377 comment. 9378 (mark_pseudo_regno_live): Call make_object_live. 9379 (mark_pseudo_regno_subword_live): Likewise. 9380 (mark_hard_reg_dead): Update function comment. 9381 (mark_hard_reg_live): Call make_hard_regno_live. 9382 (process_bb_node_lives): Likewise. 9383 * lra-lives.c (make_hard_regno_born): Rename from this... 9384 (make_hard_regno_live): ... to this. Remove update to conflict 9385 information. Remove now uneeded check_pic_pseudo_p argument. 9386 Update function comment. 9387 (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update 9388 to conflict information. Update function comment. 9389 (mark_pseudo_live): Remove update to conflict information. Update 9390 function comment. 9391 (mark_pseudo_dead): Add conflict information update. 9392 (mark_regno_live): Call make_hard_regno_live. 9393 (mark_regno_dead): Call make_hard_regno_dead with new arguement. 9394 (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead. 9395 93962018-09-29 H.J. Lu <hongjiu.lu@intel.com> 9397 9398 PR target/87370 9399 * config/i386/i386.c (construct_container): Use TImode for 9400 BLKmode values in 2 integer registers. 9401 94022018-09-29 Jeff Law <law@redhat.com> 9403 9404 * builtins.c (unterminated_array): Pass in c_strlen_data * to 9405 c_strlen rather than just a tree *. 9406 (c_strlen): Change NONSTR argument to a c_strlen_data pointer. 9407 Update recursive calls appropriately. If caller did not provide a 9408 suitable data pointer, create a local one. When a non-terminated 9409 string is discovered, bubble up information about the string via the 9410 c_strlen_data object. 9411 * builtins.h (c_strlen): Update prototype. 9412 (c_strlen_data): New structure. 9413 * gimple-fold.c (get_range_strlen): Update calls to c_strlen. 9414 For a type 2 call, if c_strlen indicates a non-terminated string 9415 use the length of the non-terminated string. 9416 (gimple_fold_builtin_stpcpy): Update calls to c_strlen. 9417 94182018-09-29 Jakub Jelinek <jakub@redhat.com> 9419 9420 PR target/87467 9421 * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use 9422 __m512d type for __A argument rather than __m512. 9423 94242018-09-28 John David Anglin <danglin@gcc.gnu.org> 9425 9426 * match.pd (simple_comparison): Don't optimize if either operand is 9427 a function pointer when target needs function pointer canonicalization. 9428 94292018-09-28 Segher Boessenkool <segher@kernel.crashing.org> 9430 9431 * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for 9432 power5 .. power9 to remove indirection. 9433 * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC, 9434 ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC, 9435 ASM_CPU_476_SPEC): Delete. 9436 (ASM_CPU_SPEC): Adjust. 9437 (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7, 9438 asm_cpu_power8, asm_cpu_power9, asm_cpu_476. 9439 94402018-09-28 Segher Boessenkool <segher@kernel.crashing.org> 9441 9442 * config.in: Delete HAVE_AS_DCI. 9443 * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true. 9444 * config/rs6000/rs6000.h: Ditto. 9445 * configure.ac: Delete HAVE_AS_DCI. 9446 * configure: Regenerate. 9447 94482018-09-28 Segher Boessenkool <segher@kernel.crashing.org> 9449 9450 * config.in (HAVE_AS_LWSYNC): Delete. 9451 * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete. 9452 * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never 9453 do it as a .long . 9454 * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete. 9455 * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it 9456 as a .long . 9457 * configure.ac: Delete HAVE_AS_LWSYNC. 9458 * configure: Regenerate. 9459 94602018-09-28 Eric Botcazou <ebotcazou@adacore.com> 9461 Pierre-Marie de Rodat <derodat@adacore.com> 9462 9463 * calls.c (expand_call): Try to do a tail call for thunks at -O0 too. 9464 * cgraph.h (struct cgraph_thunk_info): Add indirect_offset. 9465 (cgraph_node::create_thunk): Add indirect_offset parameter. 9466 (thunk_adjust): Likewise. 9467 * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter 9468 and initialize the corresponding field with it. 9469 (cgraph_node::dump): Dump indirect_offset field. 9470 * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset. 9471 * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks. 9472 (thunk_adjust): Add indirect_offset parameter and deal with it. 9473 (cgraph_node::expand_thunk): Deal with the indirect_offset field and 9474 pass it to thunk_adjust. Do not call the target hook if it's non-zero 9475 or if the thunk is external or local. Fix formatting. Do not chain 9476 the RESULT_DECL to BLOCK_VARS. Pass the static chain to the target, 9477 if any, in the GIMPLE representation. 9478 * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset. 9479 * lto-cgraph.c (lto_output_node): Write indirect_offset field. 9480 (input_node): Read indirect_offset field. 9481 * tree-inline.c (expand_call_inline): Pass indirect_offset field in the 9482 call to thunk_adjust. 9483 * tree-nested.c (struct nesting_info): Add thunk_p field. 9484 (create_nesting_tree): Set it. 9485 (convert_all_function_calls): Copy static chain from targets to thunks. 9486 (finalize_nesting_tree_1): Return early for thunks. 9487 (unnest_nesting_tree_1): Do not finalize thunks. 9488 (gimplify_all_functions): Do not gimplify thunks. 9489 94902018-09-28 David Malcolm <dmalcolm@redhat.com> 9491 9492 * opt-suggestions.c (option_proposer::build_option_suggestions): 9493 Release "option_values". 9494 94952018-09-28 David Malcolm <dmalcolm@redhat.com> 9496 9497 * coverage.c (get_coverage_counts): Convert problem-reporting dump 9498 messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION. 9499 * dumpfile.c (kind_as_string): New function. 9500 (dump_loc): Rather than a hardcoded prefix of "note: ", use 9501 kind_as_string to vary the prefix based on dump_kind. 9502 (selftest::test_capture_of_dump_calls): Update for above. 9503 95042018-09-28 Uros Bizjak <ubizjak@gmail.com> 9505 9506 * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG. 9507 (GET_SSE_REGNO): Rename from SSE_REGNO. Update all uses for rename. 9508 95092018-09-28 Uros Bizjak <ubizjak@gmail.com> 9510 9511 * config/i386/i386.h (CC_REGNO): Remove FPSR_REG. 9512 * config/i386/i386.c (ix86_fixed_condition_code_regs): Use 9513 INVALID_REGNUM instead of FPSR_REG. 9514 (ix86_md_asm_adjust): Do not clobber FPSR_REG. 9515 * config/i386/i386.md: Update comment of FP compares. 9516 (fldenv): Do not clobber FPSR_REG. 9517 95182018-09-28 Richard Biener <rguenther@suse.de> 9519 9520 * tree.h (BLOCK_ORIGIN): New. 9521 * omp-expand.c (grid_expand_target_grid_body): Assign 9522 BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN. 9523 * tree-inline.c (remap_block): Likewise. 9524 * auto-profile.c (get_function_decl_from_block): Simplify 9525 by eliding the BLOCK_ABSTRACT_ORIGIN chasing. 9526 * langhooks.c (lhd_print_error_function): Likewise. 9527 * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to): 9528 Likewise. 9529 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise. 9530 * tree.c (block_nonartificial_location): Likewise. 9531 (block_ultimate_origin): Likewise. 9532 * tree-pretty-print.c (percent_K_format): Likewise. Remove 9533 no longer needed LTO case. 9534 95352018-09-28 Andrew Stubbs <ams@codesourcery.com> 9536 Jan Hubicka <jh@suse.cz> 9537 Martin Jambor <mjambor@suse.cz> 9538 9539 * simplify-rtx.c (simplify_merge_mask): New function. 9540 (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the 9541 same masks are used in op1 or op2. 9542 (test_vec_merge): New function. 9543 (test_vector_ops): Call test_vec_merge. 9544 95452018-09-28 Eric Botcazou <ebotcazou@adacore.com> 9546 9547 * config/sparc/sparc-protos.h (sparc_branch_cost): Declare. 9548 * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost. 9549 * config/sparc/sparc.c (struct processor_costs): Add branch_cost field. 9550 (cypress_costs): Set it. 9551 (supersparc_costs): Likewise. 9552 (hypersparc_costs): Likewise. 9553 (leon_cost): Likewise. 9554 (leon3_costs): Likewise. 9555 (sparclet_costs): Likewise. 9556 (ultrasparc_costs): Likewise. 9557 (ultrasparc_costs): Likewise. 9558 (niagara_costs): Likewise. 9559 (niagara2_costs): Likewise. 9560 (niagara3_costs): Likewise. 9561 (niagara4_costs): Likewise. 9562 (niagara7_costs): Likewise. 9563 (m8_costs): Likewise. 9564 (TARGET_CAN_FOLLOW_JUMP): Define. 9565 (pass_work_around_errata::gate): Minor tweak. 9566 (sparc_option_override): Remove MASK_FSMULD mask for V7 processors. 9567 Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor. 9568 Automaitcally clear MASK_FSMULD mask for V7 processors. 9569 (sparc_can_follow_jump): New static function. 9570 (output_ubranch): Deal with CROSSING_JUMP_P. 9571 (sparc_use_sched_lookahead): Rewrite using switch statement. 9572 (sparc_issue_rate): Reorder. 9573 (sparc_branch_cost): New function. 9574 95752018-09-27 Martin Sebor <msebor@redhat.com> 9576 9577 * tree.h (tree_to_shwi): Add attribute nonnull and pure. 9578 (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same. 9579 (int_fits_type_p): Same. 9580 95812018-09-27 Uros Bizjak <ubizjak@gmail.com> 9582 9583 * config/i386/i386.md (FPCR_REG): Remove. 9584 (UNSPEC_FLDCW): Remove. 9585 (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG. 9586 (x86_fldcw_1): Remove insn pattern. 9587 (fnstenv): Do not clobber FPCR_REG. 9588 (fldenv): Ditto. 9589 * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register. 9590 (CALL_USED_REGISTERS): Ditto. 9591 (REG_ALLOC_ORDER): Ditto. 9592 (REG_CLASS_CONTENTS): Ditto. 9593 (HI_REGISTER_NAMES): Ditto. 9594 (ADDITIONAL_REGISTER_NAMES): Use defines instead 9595 of numerical constants. 9596 * config/i386/i386.c (regclass_map): Remove fpsr register. 9597 (dbx_register_map): Ditto. 9598 (dbx64_register_map): Ditto. 9599 (svr4_dbx_register_map): Ditto. 9600 (print_reg): Do not handle FPCR_REG. 9601 96022018-09-27 Segher Boessenkool <segher@kernel.crashing.org> 9603 9604 PR target/87149 9605 * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR, 9606 HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9): 9607 Delete, always treat as true. 9608 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal): 9609 Ditto. Simplify remaining code. 9610 * config/powerpcspe/powerpcspe.h: Ditto. 9611 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto. 9612 Simplify remaining code. 9613 (rs6000_expand_builtin): Ditto. 9614 * config/rs6000/rs6000.h: Ditto. 9615 * configure.ac: Ditto. 9616 * configure: Regenerate. 9617 96182018-09-27 Martin Liska <mliska@suse.cz> 9619 9620 * coverage.c (get_coverage_counts): Revert the formatting 9621 of missing profile opt info. 9622 96232018-09-27 Richard Biener <rguenther@suse.de> 9624 9625 PR debug/37801 9626 PR debug/87440 9627 * dwarf2out.c (set_block_origin_self): Do not mark outermost 9628 block as we do not output that. 9629 (gen_inlined_subroutine_die): Elide the originally outermost 9630 block, matching what we do for concrete instances. 9631 (decls_for_scope): Add parameter specifying whether to recurse 9632 to subblocks. 9633 96342018-09-27 Andrew Stubbs <ams@codesourcery.com> 9635 Tom de Vries <tom@codesourcery.com> 9636 9637 PR 82089 9638 9639 * expmed.c (emit_cstore): Fix handling of result_mode == BImode and 9640 STORE_FLAG_VALUE == 1. 9641 96422018-09-27 Andreas Krebbel <krebbel@linux.ibm.com> 9643 9644 * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New 9645 constant definitions. 9646 ("tx_assist"): Replace magic number with PPA_TX_ABORT. 9647 ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm. 9648 ("speculation_barrier"): New expander definition. 9649 96502018-09-26 Indu Bhagat <indu.bhagat@oracle.com> 9651 9652 PR gcov-profile/86957 9653 * common.opt: New warning option -Wmissing-profile. 9654 * coverage.c (get_coverage_counts): Add warning for missing .gcda file. 9655 * doc/invoke.texi: Document -Wmissing-profile. 9656 96572018-09-26 Jim Wilson <jimw@sifive.com> 9658 9659 * config/riscv/riscv.md (subsi3_extended2): Add J constraint. 9660 (negdi2, negsi2, negsi2_extended, negsi2_extended2): New. 9661 96622018-09-26 Martin Sebor <msebor@redhat.com> 9663 9664 * tree.c (zerop): Change return type to bool. 9665 (integer_zerop, integer_onep, integer_each_onep): Same. 9666 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same. 9667 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same. 9668 (real_onep, real_minus_onep, chain_index): Same. 9669 (print_type_hash_statistics, type_list_equal): Same. 9670 * tree.h (zerop): Same. 9671 (zerop, integer_zerop, integer_onep, integer_each_onep): Same. 9672 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same. 9673 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same. 9674 (real_onep, real_minus_onep, chain_index): Same. 9675 (print_type_hash_statistics, type_list_equal): Same. 9676 96772018-09-26 Jim Wilson <jimw@sifive.com> 9678 9679 * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment. 9680 96812018-09-26 Jakub Jelinek <jakub@redhat.com> 9682 9683 PR target/87414 9684 * config/i386/i386.c: Include debug.h and dwarf2out.h. 9685 (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the 9686 call. 9687 96882018-09-25 Andrew Stubbs <ams@codesourcery.com> 9689 9690 * builtins.c (get_builtin_sync_mem): Force address mode conversion. 9691 96922018-09-26 Uros Bizjak <ubizjak@gmail.com> 9693 9694 * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS 9695 and FP_SECOND_SSE_REGS. 9696 (REG_CLASS_NAMES): Ditto. 9697 (REG_CLASS_CONTENTS): Ditto. 9698 * config/i386/i386.c (ix86_preferred_reload_class) Do not handle 9699 FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes. 9700 (ix86_preferred_output_reload_class): Ditto. 9701 * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f" 9702 clobber constraint to "=&f". 9703 (fix_truncdi_i387): Ditto. 9704 (lrintxfdi2): Ditto. 9705 (fistdi2_<rounding>): Ditto. 9706 (fpremxf4_i387): Change "=u" constraint to "=f". 9707 (fprem1xf4_i387): Ditto. 9708 (sincosxf3): Ditto. 9709 (fptanxf4_i387): Ditto. 9710 (fxtractxf3_i387): Ditto. 9711 (fscalexf4_i387): Ditto. 9712 (atan2xf3): Change "u" constraint to "f". 9713 (fyl2xxf3_i387): Ditto. 9714 (fyl2xp1xf3_i387): Ditto. 9715 97162018-09-26 Uros Bizjak <ubizjak@gmail.com> 9717 9718 PR target/87439 9719 * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update 9720 for removed I387_MASK_PM entity. 9721 9722 97232018-09-26 Jeff Law <law@redhat.com> 9724 Revert 9725 2018-09-26 Alexey Neyman <stilor@att.net> 9726 9727 * graphite.h: Include <isl/id.h> and <isl/space.h>; these 9728 headers are no longer pulled in by <isl/val.h>. 9729 97302018-09-26 Richard Biener <rguenther@suse.de> 9731 9732 PR debug/87443 9733 * dwarf2out.c (gen_lexical_block_die): Do not equate inline 9734 or concrete instance DIE to the tree. Create abstract origin 9735 attributes also for concrete instances. 9736 97372018-09-26 Alexey Neyman <stilor@att.net> 9738 9739 * graphite.h: Include <isl/id.h> and <isl/space.h>; these 9740 headers are no longer pulled in by <isl/val.h>. 9741 97422018-09-26 Matthew Malcomson <matthew.malcomson@arm.com> 9743 9744 * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op): 9745 Use new helper functions. 9746 * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>): 9747 Use new helper functions. 9748 * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire, 9749 aarch_mm_needs_release): New declarations. 9750 * config/arm/aarch-common.c (aarch_mm_needs_acquire, 9751 aarch_mm_needs_release): New. 9752 97532018-09-26 Eric Botcazou <ebotcazou@adacore.com> 9754 9755 * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks. 9756 (arm32_output_mi_thunk): Deal with long calls. 9757 97582018-09-26 Richard Biener <rguenther@suse.de> 9759 9760 PR debug/87428 9761 PR debug/87362 9762 * tree-inline.c (expand_call_inline): When the location 9763 of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION 9764 or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of 9765 the inserted BLOCK to make inlined_function_outer_scope_p 9766 recognize it. 9767 * dwarf2out.c (add_call_src_coords_attributes): Do not add 9768 coords for reserved locations. 9769 97702018-09-25 Segher Boessenkool <segher@kernel.crashing.org> 9771 9772 * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative. 9773 (*call_indirect_nonlocal_sysv<mode>): Ditto. 9774 (*call_value_indirect_nonlocal_sysv<mode>): Ditto. 9775 (*sibcall_nonlocal_sysv<mode>): Ditto. 9776 (*sibcall_value_nonlocal_sysv<mode>): Ditto. 9777 (<bd>_<mode>): Ditto. 9778 (<bd>tf_<mode>): Ditto. 9779 97802018-09-25 Segher Boessenkool <segher@kernel.crashing.org> 9781 9782 * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output 9783 control string as a list of templates instead of as C code. 9784 (*altivec_movti): Ditto. 9785 * config/rs6000/darwin.md (movdf_low_di): Ditto. 9786 97872018-09-25 Jim Wilson <jimw@sifive.com> 9788 9789 * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak 9790 when target symbol is weak. 9791 97922018-09-25 Bernd Edlinger <bernd.edlinger@hotmail.de> 9793 9794 PR c/87387 9795 * builtins.c (unterminated_array): Simplify. 9796 * expr.c (string_constant): Handle SSA_NAME. Add more exceptions 9797 where pointer arithmetic is safe. 9798 97992018-09-25 Segher Boessenkool <segher@kernel.crashing.org> 9800 9801 PR target/86987 9802 * config/rs6000/altivec.md (altivec_vspltb): Use 9803 const_0_to_15_operand instead of u5bit_cint_operand. 9804 (*altivec_vspltb_internal): Ditto. 9805 (altivec_vspltb_direct): Ditto. 9806 (altivec_vsplth): Use const_0_to_7_operand instead of 9807 u5bit_cint_operand. 9808 (*altivec_vsplth_internal): Ditto. 9809 (altivec_vsplth_direct): Ditto. 9810 (altivec_vspltw): Use const_0_to_3_operand instead of 9811 u5bit_cint_operand. 9812 (*altivec_vspltw_internal): Ditto. 9813 (altivec_vspltw_direct): Ditto. 9814 (altivec_vspltsf): Ditto. 9815 (*altivec_vspltsf_internal): Ditto. 9816 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the 9817 various splats with the proper size immediate. Reorder the various 9818 cases by ascending size of immediate, and put all such together. 9819 98202018-09-25 Richard Biener <rguenther@suse.de> 9821 9822 PR debug/83941 9823 * dwarf2out.c (add_AT_external_die_ref): Remove now redundant 9824 GC-ification. 9825 (maybe_create_die_with_external_ref): Do not create 9826 DW_TAG_imported_unit here. 9827 (add_abstract_origin_attribute): Handle external BLOCK refs. 9828 (dwarf2out_abstract_function): Simplify LTO case. 9829 (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely 9830 rather than using maybe_create_die_with_external_ref. 9831 98322018-09-25 Uros Bizjak <ubizjak@gmail.com> 9833 9834 PR target/71278 9835 * config/i386/i386.md (frndintxf2_mask_pm): Remove. 9836 (frndintxf2_mask_pm_i387): Ditto. 9837 (nearbyintxf2): Rewrite expander pattern to match rintxf2. 9838 Enable for !flag_trapping_math. 9839 (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math. 9840 Enable SSE modes for TARGET_SSE4_1 and expand them with round insn. 9841 Change operand 1 predicate to nonimmediate_operand. 9842 (attr "i387_cw"): Remove mask_pm. 9843 * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM. 9844 (enum ix86_entity): Remove I387_MASK_PM. 9845 * config/i386/i386.c (ix86_i387_mode_needed): Do not 9846 handle I387_MASK_PM. 9847 (ix86_mode_needed): Ditto. 9848 (ix86_mode_after): Ditto. 9849 (ix86_mode_entry): Ditto. 9850 (ix86_mode_exit): Ditto. 9851 (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM. 9852 98532018-09-25 Jakub Jelinek <jakub@redhat.com> 9854 9855 * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and 9856 to_update_switch_stmts to vNULL instead of calling create on them 9857 immediately. 9858 98592018-09-25 Richard Biener <rguenther@suse.de> 9860 9861 PR tree-optimization/87402 9862 * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function. 9863 (visit_phi): Re-instantiate handling of supposed to be VARYING 9864 but non-VARYING backedge value. 9865 98662018-09-25 Richard Biener <rguenther@suse.de> 9867 9868 PR debug/83941 9869 * dwarf2out.c (struct sym_off_pair): New. 9870 (external_die_map): New global. 9871 (lookup_decl_die): When in LTO create DIEs lazily from the 9872 external_die_map. 9873 (lookup_block_die): New function, create DIEs lazily in LTO. 9874 (equate_block_to_die): New function. 9875 (dwarf2out_die_ref_for_decl): During WPA get the association 9876 from the external DIE map. 9877 (dwarf2out_register_external_die): Record mapping into the 9878 external DIE map. 9879 (maybe_create_die_with_external_ref): New function split out from 9880 DIE generation part of old dwarf2out_register_external_die. 9881 (add_abstract_origin_attribute): Do not return the DIE. When 9882 in LTO reference externals directly. 9883 (dwarf2out_abstract_function): When in LTO ignore calls for 9884 decls with external DIEs (already present abstract instances). 9885 (gen_call_site_die): Adjust. 9886 (add_high_low_attributes): Likewise. 9887 (gen_lexical_block_die): Likewise. 9888 (gen_inlined_subroutine_die): Likewie. 9889 (gen_block_die): Likewise. 9890 (dwarf2out_inline_entry): Likewise. 9891 (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit 9892 DIEs. 9893 98942018-09-25 Martin Liska <mliska@suse.cz> 9895 9896 * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two 9897 integers and not by a float value. 9898 98992018-09-25 Martin Liska <mliska@suse.cz> 9900 9901 PR fortran/87394 9902 * dbgcnt.c (dbg_cnt_process_single_pair): Return false 9903 instead of NULL. 9904 * dumpfile.c (dump_enable_all): Remove extra parenthesis. 9905 * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN. 9906 * godump.c (go_format_type): Remove extra parenthesis. 9907 99082018-09-25 Martin Liska <mliska@suse.cz> 9909 9910 * alias.c (set_dest_equal_p): Remove unused function. 9911 * config/i386/i386.c (def_builtin_pure2): Likewise. 9912 * diagnostic-show-locus.c (class layout): Remove 9913 unused field. 9914 (layout::layout): Likewise here. 9915 * dump-context.h (class temp_dump_context): Likewise. 9916 * dwarf2out.c (add_AT_fde_ref): Remove unused function. 9917 (add_AT_loclistsptr): Likewise. 9918 (add_AT_offset): Likewise. 9919 (get_AT_hi_pc): Likewise. 9920 (is_comdat_die): Likewise. 9921 (type_is_enum): Likewise. 9922 (ceiling): Likewise. 9923 (add_AT_vms_delta): Likewise. 9924 (is_class_die): Likewise. 9925 * edit-context.c (class line_event): Remove unused field. 9926 * graphite-sese-to-poly.c (tree_int_to_gmp): Remove 9927 unused function. 9928 * ipa-cp.c (ipa_get_vr_lat): Likewise. 9929 * lra-constraints.c (ok_for_index_p_nonstrict): Likewise. 9930 (ok_for_base_p_nonstrict): Likewise. 9931 * tree-chrec.c (is_not_constant_evolution): Likewise. 9932 (chrec_fold_poly_cst): Likewise. 9933 * tree-if-conv.c (has_pred_critical_p): Likewise. 9934 * tree-ssa-coalesce.c (print_exprs): Likewise. 9935 * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise. 9936 * tree-ssa-uninit.c (is_and_or_or_p): Likewise. 9937 * tree-vrp.c (value_ranges_intersect_p): Likewise. 9938 (value_range_nonnegative_p): Likewise. 9939 99402018-09-25 Martin Liska <mliska@suse.cz> 9941 9942 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue): 9943 Do not handle "GNU Pascal". 9944 * config/rs6000/rs6000.c (rs6000_output_function_epilogue): 9945 Likewise. 9946 * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal 9947 from documentation. Likewise. 9948 * dbxout.c (dbxout_range_type): Likewise. 9949 * doc/cpp.texi: Likewise. 9950 * doc/extend.texi: Likewise. 9951 * doc/frontends.texi: Likewise. 9952 * doc/invoke.texi: Remove Pascal entry. 9953 * tree.def (CLEANUP_POINT_EXPR): Likewise. 9954 * doc/rtl.texi (MODE_FUNCTION): Remove not used entry. 9955 99562018-09-25 Martin Liska <mliska@suse.cz> 9957 9958 PR middle-end/86078 9959 * doc/invoke.texi: Document all parameters and remove default 9960 of the parameters. 9961 99622018-09-25 Ilya Leoshkevich <iii@linux.ibm.com> 9963 9964 PR bootstrap/87417 9965 * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT 9966 contains HOST_WIDE_INTs when computing its size. 9967 99682018-09-24 Jim Wilson <jimw@sifive.com> 9969 9970 PR target/87391 9971 * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E 9972 not TARGET_RVE. 9973 (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise. 9974 99752018-09-24 Andrew Pinski <apinski@marvell.com> 9976 9977 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't 9978 access prev before checking it for NULLness in the 9979 AARCH64_FUSE_CMP_BRANCH case. 9980 99812018-09-24 H.J. Lu <hongjiu.lu@intel.com> 9982 9983 PR target/82699 9984 * config/i386/i386.c (rest_of_insert_endbranch): Set 9985 endbr_queued_at_entrance to true and don't insert ENDBR if 9986 x86_function_profiler will be called. 9987 (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance 9988 is true. 9989 * config/i386/i386.h (machine_function): Add 9990 endbr_queued_at_entrance. 9991 99922018-09-24 Ilya Leoshkevich <iii@linux.ibm.com> 9993 9994 * genattrtab.c (mk_attr_alt): Use alternative_mask. 9995 (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field 9996 types. 9997 (check_attr_test): Use alternative_mask. 9998 (get_attr_value): Likewise. 9999 (compute_alternative_mask): Use alternative_mask and XWINT. 10000 (make_alternative_compare): Use alternative_mask. 10001 (attr_alt_subset_p): Use XWINT. 10002 (attr_alt_subset_of_compl_p): Likewise. 10003 (attr_alt_intersection): Use alternative_mask and XWINT. 10004 (attr_alt_union): Likewise. 10005 (attr_alt_complement): Use HOST_WIDE_INT and XWINT. 10006 (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT. 10007 (simplify_test_exp): Use alternative_mask and XWINT. 10008 (write_test_expr): Use alternative_mask and XWINT, adjust bit 10009 number calculation to support 64 bits. Generate code that 10010 checks 64-bit masks. 10011 (main): Use alternative_mask. 10012 * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww. 10013 100142018-09-24 Ilya Leoshkevich <iii@linux.ibm.com> 10015 10016 PR target/80080 10017 * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL 10018 RETURN+USE when returning via %r14. 10019 100202018-09-24 Martin Liska <mliska@suse.cz> 10021 10022 * gcov.c (output_lines): Print colorization legend 10023 for both flag_use_colors and flag_use_hotness_colors. 10024 Reword the help. 10025 100262018-09-24 Martin Liska <mliska@suse.cz> 10027 10028 * coverage.c (get_coverage_counts): Use warning_at 10029 with current_function_decl location. Use %qD in warning 10030 message. 10031 100322018-09-24 Martin Liska <mliska@suse.cz> 10033 10034 * memory-block.h (memory_block_pool::release): Annotate with 10035 valgrind that the memory is not accessible. 10036 100372018-09-24 Martin Liska <mliska@suse.cz> 10038 10039 PR sanitizer/85774 10040 * asan.c: Make asan_handled_variables extern. 10041 * asan.h: Likewise. 10042 * cfgexpand.c (expand_stack_vars): Make sure 10043 a representative is unpoison if another 10044 variable in the partition is handled by 10045 use-after-scope sanitization. 10046 100472018-09-24 Richard Biener <rguenther@suse.de> 10048 10049 PR tree-optimization/63155 10050 * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to 10051 the worklist when the edge of the respective argument isn't 10052 executable. 10053 100542018-09-23 Uros Bizjak <ubizjak@gmail.com> 10055 10056 * config/i386/i386.h (enum reg_class): Rename MASK_REGS to 10057 ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS. 10058 (MASK_CLASS_P): Update for rename. 10059 (MAYBE_MASK_CLASS_P): Ditto. 10060 (REG_CLASS_NAMES): Update. 10061 (REG_CLASS_CONTENT): Update. 10062 * config/i386/i386.c (regclass_map): Update for MASK_REG 10063 and ALL_MASK_REGS rename. 10064 * config/i386/constraints.md (Yk): Update for rename. 10065 (k): Ditto. 10066 100672018-09-23 Uros Bizjak <ubizjak@gmail.com> 10068 10069 * config/i386/i386.h (enum reg_class): Remove 10070 EVEX_SSE_REGS and MOD4_SSE_REGS. 10071 (REG_CLASS_NAMES): Update. 10072 (REG_CLASS_CONTENT): Update. 10073 * config/i386/i386.c (regclass_map): Declare AVX-512 SSE 10074 registers as ALL_SSE_REGS. 10075 (ix86_additional_allocno_class_p): Remove. 10076 (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove. 10077 (ix86_register_priority): Lower priority of EVEX SSE registers. 10078 Use IN_RANGE macro where appropriate. 10079 (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and 10080 AVX-5124VNNIW checks. 10081 (ix86_modes_tieable_p): Tie 512-bit SSE modes. 10082 * config/i386/sse.md (avx5124fmaddps_4fmaddps) 10083 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz) 10084 (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask) 10085 (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps) 10086 (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz) 10087 (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask) 10088 (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd) 10089 (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz) 10090 (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask) 10091 (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint. 10092 * config/i386/constraints.md (Yh): Remove. 10093 100942018-09-23 Uros Bizjak <ubizjak@gmail.com> 10095 10096 * config/i386/i386.c (regclass_map): Declare integer REX registers 10097 as GENERAL_REGS. 10098 100992018-09-23 Gerald Pfeifer <gerald@pfeifer.com> 10100 10101 * doc/service.texi (Service): Switch the fsf.org link to https. 10102 101032018-09-22 Chung-Ju Wu <jasonwucj@gmail.com> 10104 10105 PR target/86798 10106 * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): 10107 Define to speculation_safe_value_not_needed. 10108 101092018-09-21 Florian Weimer <fweimer@redhat.com> 10110 10111 PR middle-end/81035 10112 * doc/extend.texi (Common Function Attributes): Mention that 10113 noreturn suppresses tail call optimization. 10114 101152018-09-21 Jeff Law <law@redhat.com> 10116 10117 * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call 10118 vr_values::cleanup_edges_and_switches. 10119 * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into 10120 vr_values class. 10121 (identify_jump_threads): Remove EDGE_IGNORE handling. 10122 (execute_vrp): Move handling of to_remove_edges and 10123 to_update_switch_stmts into vr_values class member functions. 10124 * tree-vrp.h (switch_update, to_remove_edges): Remove declarations. 10125 (to_update_switch_stmts): Likewise. 10126 * vr-values.c: Include cfghooks.h. 10127 (vr_values::vr_values): Initialize to_remove_edges and 10128 to_update_switch_stmts. 10129 (vr_values::~vr_values): Verify to_remove_edges and 10130 to_update_switch_stmts are empty. 10131 (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed. 10132 (vr_values::cleanup_edges_and_switches): New member function. 10133 * vr-values.h (vr_values): Add cleanup_edges_and_switches member 10134 function. Add new data members. 10135 101362018-09-21 David Malcolm <dmalcolm@redhat.com> 10137 10138 PR tree-optimization/87309 10139 * dumpfile.c (dump_context::begin_scope): Filter the dump_loc 10140 calls with pflags and alt_flags. 10141 (selftest::test_capture_of_dump_calls): Add test of interaction of 10142 MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE. 10143 101442018-09-21 Olivier Hainque <hainque@adacore.com> 10145 10146 * config.gcc: Factorize and comment inclusion of vxworks-dummy.h. 10147 101482018-09-21 Olivier Hainque <hainque@adacore.com> 10149 10150 * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1. 10151 101522018-09-21 Olivier Hainque <hainque@adacore.com> 10153 10154 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal. 10155 Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then: 10156 (VXWORKS_LIBS_RTP): Minor reordering. 10157 101582018-09-21 Olivier Hainque <hainque@adacore.com> 10159 10160 * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define. 10161 (VXWORKS_LIBS_DIR_RTP): Remove definition and use. 10162 101632018-09-21 Olivier Hainque <hainque@adacore.com> 10164 10165 * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64. 10166 (PTRDIFF_TYPE): Likewise. 10167 101682018-09-21 Olivier Hainque <hainque@adacore.com> 10169 10170 * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from 10171 triplet, similar to support for VxWorks7. 10172 * config/vxworks-dummy.h: Provide a default definition 10173 of TARGET_VXWORKS64 to 0. 10174 101752018-09-21 Olivier Hainque <hainque@adacore.com> 10176 10177 * config/vxworks.h (TARGET_VXWORKS7): Move default definition ... 10178 * config/vxworks-dummy.h: here. 10179 101802018-09-21 Olivier Hainque <hainque@adacore.com> 10181 10182 * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc* 10183 101842018-09-21 Shaokun Zhang <zhangshaokun@hisilicon.com> 10185 Bo Zhou <zbo.zhou@hisilicon.com> 10186 10187 * config/aarch64/aarch64-cores.def (tsv110): New CPU. 10188 * config/aarch64/aarch64-tune.md: Regenerated. 10189 * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110". 10190 * config/aarch64/aarch64.c (tsv110_tunings): New tuning table. 10191 * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs. 10192 101932018-09-21 Andrew Stubbs <ams@codesourcery.com> 10194 Julian Brown <julian@codesourcery.com> 10195 10196 * builtins.c (get_builtin_sync_mem): Handle address spaces. 10197 101982018-09-21 Eric Botcazou <ebotcazou@adacore.com> 10199 10200 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false 10201 if the call takes a static chain. 10202 102032018-09-21 Martin Liska <mliska@suse.cz> 10204 10205 * auto-profile.c (autofdo_source_profile::read): Do not 10206 set sum_all. 10207 (read_profile): Do not add working sets. 10208 (read_autofdo_file): Remove sum_all. 10209 (afdo_callsite_hot_enough_for_early_inline): Remove const 10210 qualifier. 10211 * coverage.c (struct counts_entry): Remove gcov_summary. 10212 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY, 10213 do not support GCOV_TAG_PROGRAM_SUMMARY. 10214 (get_coverage_counts): Remove summary and expected 10215 arguments. 10216 * coverage.h (get_coverage_counts): Likewise. 10217 * doc/gcov-dump.texi: Remove -w option. 10218 * gcov-dump.c (dump_working_sets): Remove. 10219 (main): Do not support '-w' option. 10220 (print_usage): Likewise. 10221 (tag_summary): Likewise. 10222 * gcov-io.c (gcov_write_summary): Do not dump 10223 histogram. 10224 (gcov_read_summary): Likewise. 10225 (gcov_histo_index): Remove. 10226 (gcov_histogram_merge): Likewise. 10227 (compute_working_sets): Likewise. 10228 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark 10229 it not obsolete. 10230 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete. 10231 (GCOV_TAG_SUMMARY_LENGTH): Adjust. 10232 (GCOV_HISTOGRAM_SIZE): Remove. 10233 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise. 10234 (struct gcov_summary): Simplify rapidly just 10235 to runs and sum_max fields. 10236 (gcov_histo_index): Remove. 10237 (NUM_GCOV_WORKING_SETS): Likewise. 10238 (compute_working_sets): Likewise. 10239 * gcov-tool.c (print_overlap_usage_message): Remove 10240 trailing empty line. 10241 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY. 10242 (output_lines): Remove program related line. 10243 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram. 10244 * lto-cgraph.c (output_profile_summary): Do not stream GCOV 10245 histogram. 10246 (input_profile_summary): Do not read it. 10247 (merge_profile_summaries): And do not merge it. 10248 (input_symtab): Do not call removed function. 10249 * modulo-sched.c (sms_schedule): Do not print sum_max. 10250 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was 10251 removed when histogram method was invented. 10252 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO 10253 mode. 10254 * postreload-gcse.c (eliminate_partially_redundant_load): Fix 10255 GCOV coding style. 10256 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION 10257 and dump selected value. 10258 * profile.c (add_working_set): Remove. 10259 (get_working_sets): Likewise. 10260 (find_working_set): Likewise. 10261 (get_exec_counts): Do not work with working sets. 10262 (read_profile_edge_counts): Do not inform as sum_max is removed. 10263 (compute_branch_probabilities): Likewise. 10264 (compute_value_histograms): Remove argument for call of 10265 get_coverage_counts. 10266 * profile.h: Do not make gcov_summary const. 10267 102682018-09-21 Monk Chiang <sh.chiang04@gmail.com> 10269 10270 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0. 10271 102722018-09-21 Eric Botcazou <ebotcazou@adacore.com> 10273 10274 PR tree-optimization/86990 10275 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate): 10276 Check that the entire merged store group is made of constants only for 10277 overlapping stores. 10278 102792018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io> 10280 10281 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib. 10282 (VTABLE_VERIFICATION_SPEC): Likewise. 10283 (SANITIZER_EARLY_SPEC): Likewise. 10284 (SANITIZER_SPEC): Likewise. 10285 * config/darwin.h (LINK_COMMAND_SPEC): Likewise. 10286 * doc/invoke.texi (Link Options): Document -r. 10287 102882018-09-20 Richard Biener <rguenther@suse.de> 10289 10290 PR middle-end/87054 10291 * gimplify.c (gimplify_expr): Retain alignment of 10292 addressable lvalue in dereference. 10293 102942018-09-20 Alexandre Oliva <aoliva@redhat.com> 10295 10296 PR bootstrap/87013 10297 * configure.ac: Check for .loc is_stmt support. 10298 * configure, config.in: Rebuilt. 10299 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt 10300 if not supported. 10301 103022018-09-20 Segher Boessenkool <segher@kernel.crashing.org> 10303 10304 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete. 10305 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and 10306 -misel=no. 10307 103082018-09-20 Segher Boessenkool <segher@kernel.crashing.org> 10309 10310 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete 10311 VECTOR_OTHER. 10312 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete 10313 case VECTOR_OTHER. 10314 103152018-09-20 Marek Polacek <polacek@redhat.com> 10316 10317 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options. 10318 103192018-09-20 Richard Sandiford <richard.sandiford@arm.com> 10320 10321 PR tree-optimization/87288 10322 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS 10323 into account when determining PEELING_FOR_NITERS. 10324 103252018-09-20 Richard Sandiford <richard.sandiford@arm.com> 10326 10327 PR tree-optimization/86877 10328 * tree-vect-loop.c (vect_analyze_loop_2): Call 10329 vect_verify_datarefs_alignment. 10330 103312018-09-19 Marek Polacek <polacek@redhat.com> 10332 10333 * doc/invoke.texi: Document -Wclass-conversion. 10334 103352018-09-19 John David Anglin <danglin@gcc.gnu.org> 10336 10337 * config/pa/pa.c (pa_adjust_priority): Delete. 10338 (TARGET_SCHED_ADJUST_PRIORITY): Delete define. 10339 10340 * config/pa/pa.md (atomic_storeqi): Restore deleted expander. 10341 (atomic_storehi): Likewise. 10342 (atomic_storesi): Likewise. 10343 (atomic_loaddi): Restore compare and swap exchange loop code. 10344 103452018-09-19 Segher Boessenkool <segher@kernel.crashing.org> 10346 10347 PR rtl-optimization/86902 10348 * combine.c (try_combine): When changing the CC mode used, don't change 10349 an unrelated mode in other_insn to that new CC mode. 10350 103512018-09-19 David Malcolm <dmalcolm@redhat.com> 10352 10353 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing 10354 with %T in place of calls to dump_generic_expr. 10355 (prune_runtime_alias_test_list): Likewise. 10356 (create_runtime_alias_checks): Likewise. 10357 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise. 10358 (vect_analyze_data_ref_dependence): Likewise. 10359 (vect_slp_analyze_data_ref_dependence): Likewise. 10360 (vect_record_base_alignment): Likewise. Use %G in place of call 10361 to dump_gimple_stmt. 10362 (vect_compute_data_ref_alignment): Likewise. 10363 (verify_data_ref_alignment): Likewise. 10364 (vect_find_same_alignment_drs): Likewise. 10365 (vect_analyze_group_access_1): Likewise. 10366 (vect_analyze_data_ref_accesses): Likewise. 10367 (dependence_distance_ge_vf): Likewise. 10368 (dump_lower_bound): Likewise. 10369 (vect_prune_runtime_alias_test_list): Likewise. 10370 (vect_find_stmt_data_reference): Likewise. 10371 (vect_analyze_data_refs): Likewise. 10372 (vect_create_addr_base_for_vector_ref): Likewise. 10373 (vect_create_data_ref_ptr): Likewise. 10374 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise. 10375 (vect_can_advance_ivs_p): Likewise. 10376 (vect_update_ivs_after_vectorizer): Likewise. 10377 (vect_gen_prolog_loop_niters): Likewise. 10378 (vect_prepare_for_masked_peels): Likewise. 10379 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise. 10380 (vect_determine_vectorization_factor): Likewise. 10381 (vect_is_simple_iv_evolution): Likewise. 10382 (vect_analyze_scalar_cycles_1): Likewise. 10383 (vect_analyze_loop_operations): Likewise. 10384 (report_vect_op): Likewise. 10385 (vect_is_slp_reduction): Likewise. 10386 (check_reduction_path): Likewise. 10387 (vect_is_simple_reduction): Likewise. 10388 (vect_create_epilog_for_reduction): Likewise. 10389 (vect_finalize_reduction:): Likewise. 10390 (vectorizable_induction): Likewise. 10391 (vect_transform_loop_stmt): Likewise. 10392 (vect_transform_loop): Likewise. 10393 (optimize_mask_stores): Likewise. 10394 * tree-vect-patterns.c (vect_pattern_detected): Likewise. 10395 (vect_split_statement): Likewise. 10396 (vect_recog_over_widening_pattern): Likewise. 10397 (vect_recog_average_pattern): Likewise. 10398 (vect_determine_min_output_precision_1): Likewise. 10399 (vect_determine_precisions_from_range): Likewise. 10400 (vect_determine_precisions_from_users): Likewise. 10401 (vect_mark_pattern_stmts): Likewise. 10402 (vect_pattern_recog_1): Likewise. 10403 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise. 10404 (vect_record_max_nunits): Likewise. 10405 (vect_build_slp_tree_1): Likewise. 10406 (vect_build_slp_tree_2): Likewise. 10407 (vect_print_slp_tree): Likewise. 10408 (vect_analyze_slp_instance): Likewise. 10409 (vect_detect_hybrid_slp_stmts): Likewise. 10410 (vect_detect_hybrid_slp_1): Likewise. 10411 (vect_slp_analyze_operations): Likewise. 10412 (vect_slp_analyze_bb_1): Likewise. 10413 (vect_transform_slp_perm_load): Likewise. 10414 (vect_schedule_slp_instance): Likewise. 10415 * tree-vect-stmts.c (vect_mark_relevant): Likewise. 10416 (vect_mark_stmts_to_be_vectorized): Likewise. 10417 (vect_init_vector_1): Likewise. 10418 (vect_get_vec_def_for_operand): Likewise. 10419 (vect_finish_stmt_generation_1): Likewise. 10420 (vect_check_load_store_mask): Likewise. 10421 (vectorizable_call): Likewise. 10422 (vectorizable_conversion): Likewise. 10423 (vectorizable_operation): Likewise. 10424 (vectorizable_load): Likewise. 10425 (vect_analyze_stmt): Likewise. 10426 (vect_is_simple_use): Likewise. 10427 (vect_get_vector_types_for_stmt): Likewise. 10428 (vect_get_mask_type_for_stmt): Likewise. 10429 * tree-vectorizer.c (increase_alignment): Likewise. 10430 104312018-09-19 Andrew Stubbs <ams@codesourcery.com> 10432 10433 * doc/rtl.texi: Adjust vec_select description. 10434 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use 10435 non-constant selectors. 10436 104372018-09-19 Matthew Malcomson <matthew.malcomson@arm.com> 10438 10439 * config/aarch64/aarch64-protos.h 10440 (aarch64_offset_9bit_signed_unscaled_p): New declaration. 10441 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value. 10442 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch". 10443 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield. 10444 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4. 10445 (AARCH64_FL_PROFILE): Move index so flags are ordered. 10446 (AARCH64_ISA_RCPC8_4): New flag. 10447 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed 10448 to aarch64_offset_9bit_signed_unscaled_p. 10449 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset 10450 and use stlur. 10451 * config/aarch64/constraints.md (Ust): New constraint. 10452 * config/aarch64/predicates.md. 10453 (aarch64_9bit_offset_memory_operand): New predicate. 10454 (aarch64_rcpc_memory_operand): New predicate. 10455 104562018-09-19 Eric Botcazou <ebotcazou@adacore.com> 10457 10458 PR rtl-optimization/87361 10459 * rtlanal.c (nonzero_bits1): Revert accidental change. 10460 104612018-09-19 Richard Biener <rguenther@suse.de> 10462 10463 PR tree-optimization/87349 10464 PR tree-optimization/87342 10465 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation. 10466 104672018-09-18 Marek Polacek <polacek@redhat.com> 10468 10469 P1064R0 - Allowing Virtual Function Calls in Constant Expressions 10470 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert. 10471 104722018-09-18 Segher Boessenkool <segher@kernel.crashing.org> 10473 10474 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment. 10475 104762018-09-18 Segher Boessenkool <segher@kernel.crashing.org> 10477 10478 PR rtl-optimization/86882 10479 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER. 10480 104812018-09-18 Uros Bizjak <ubizjak@gmail.com> 10482 10483 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from 10484 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator. 10485 104862018-09-18 Jonathan Wakely <jwakely@redhat.com> 10487 10488 PR other/87353 10489 * doc/invoke.texi (Link Options): Fix formatting and grammar. 10490 104912018-09-18 Richard Biener <rguenther@suse.de> 10492 10493 PR middle-end/63155 10494 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove. 10495 (compute_samebase_partition_bases): Likewise. 10496 (coalesce_ssa_name): Always use compute_optimized_partition_bases. 10497 (gimple_can_coalesce_p): Simplify. 10498 104992018-09-18 Hans-Peter Nilsson <hp@bitrange.com> 10500 10501 Handle a library implementation of ffs calling __builtin_ffs. 10502 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with... 10503 (mmix_init_libfuncs): New function: make __builtin_ffs expand 10504 to __ffsdi2. 10505 105062018-09-17 David Malcolm <dmalcolm@redhat.com> 10507 10508 * diagnostic-show-locus.c (class layout_range): Add field 10509 "m_original_idx". 10510 (layout_range::layout_range): Add "original_idx" param and use it 10511 to initialize new field. 10512 (make_range): Use 0 for original_idx. 10513 (layout::layout): Pass in index to calls to 10514 maybe_add_location_range. 10515 (layout::maybe_add_location_range): Add param "original_idx" and 10516 pass it on to layout_range. 10517 (layout::print_any_labels): Pass on range->m_original_idx to 10518 get_text call. 10519 (gcc_rich_location::add_location_if_nearby): Use 0 for 10520 original_idx. 10521 * gcc-rich-location.h (text_range_label::get_text): Update for new 10522 param. 10523 (range_label_for_type_mismatch::get_text): Likewise. 10524 105252018-09-17 Uros Bizjak <ubizjak@gmail.com> 10526 10527 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier. 10528 105292018-09-17 David Malcolm <dmalcolm@redhat.com> 10530 10531 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of 10532 format_string_diagnostic_t. 10533 (fmtwarn_n): Likewise. 10534 * substring-locations.c 10535 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor. 10536 (format_warning_n_va): Convert to... 10537 (format_string_diagnostic_t::emit_warning_n_va): ...this. 10538 (format_warning_va): Convert to... 10539 (format_string_diagnostic_t::emit_warning_va): ...this. 10540 (format_warning_at_substring): Convert to... 10541 (format_string_diagnostic_t::emit_warning): ...this. 10542 (format_warning_at_substring_n): Convert to... 10543 (format_string_diagnostic_t::emit_warning_n): ...this. 10544 * substring-locations.h (class format_string_diagnostic_t): New 10545 class. 10546 (format_warning_va): Convert to 10547 format_string_diagnostic_t::emit_warning_va. 10548 (format_warning_n_va): Convert to 10549 format_string_diagnostic_t::emit_warning_n_va. 10550 (format_warning_at_substring): Convert to 10551 format_string_diagnostic_t::emit_warning. 10552 (format_warning_at_substring_n): Convert to 10553 format_string_diagnostic_t::emit_warning_n. 10554 105552018-09-17 Cesar Philippidis <cesar@codesourcery.com> 10556 Bernd Schmidt <bernds_cb1@t-online.de> 10557 10558 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with 10559 SImode args. 10560 105612018-09-17 Uros Bizjak <ubizjak@gmail.com> 10562 10563 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change 10564 operand 0 predicate to nonimmediate operand. 10565 (rint<mode>2_frndint): Remove insn pattern. 10566 (rint<mode>2): Change operand 1 predicate to general_operand. 10567 Extend operand 1 to XFmode and generate rintxf2 insn. 10568 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>. 10569 Do not use X87MODEF mode macro. 10570 (frndintxf2_<rounding>_i387): Rename from 10571 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro. 10572 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1 10573 to XFmode and generate significandxf3 insn. 10574 105752018-09-17 Richard Biener <rguenther@suse.de> 10576 10577 PR tree-optimization/87328 10578 * tree-ssa-sccvn.c (process_bb): Remove assertion about not 10579 visiting unexecutable backedges when not iterating. 10580 (do_rpo_vn): Mark all edges not executable even when not 10581 iterating. 10582 105832018-09-17 Martin Jambor <mjambor@suse.cz> 10584 10585 PR c/63886 10586 * doc/invoke.texi (Warning Options): Likewise. 10587 105882018-09-17 Richard Biener <rguenther@suse.de> 10589 10590 PR tree-optimization/87301 10591 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly 10592 clean EH info from leftover copy assignments. 10593 105942018-09-17 Martin Liska <mliska@suse.cz> 10595 10596 PR gcov-profile/85871 10597 * gcov.c (output_intermediate_file): Fix out of bounds 10598 access. 10599 106002018-09-17 Vineet Gupta <vgupta@synopsys.com> 10601 10602 * config/arc/arc.c: Object attributes for core4 not reflected 10603 correctly. 10604 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes 10605 core3). 10606 106072018-09-17 Alexey Brodkin <abrodkin@synopsys.com> 10608 10609 * config/arc/linux.h (LINK_EH_SPEC): Add missing space. 10610 106112018-09-17 Martin Liska <mliska@suse.cz> 10612 10613 * doc/gcov.texi: Document new option --use-hotness-colors. 10614 * gcov.c (struct source_info): Declare new field. 10615 (source_info::source_info): Set default for maximum_count. 10616 (print_usage): Add new -q option. 10617 (process_args): Process it. 10618 (accumulate_line_info): Save src->maximum_count. 10619 (output_line_beginning): Make color line number if 10620 flag_use_hotness_colors is set. 10621 (output_line_details): Pass default argument value. 10622 (output_lines): Pass src->maximum_count. 10623 106242018-09-17 Martin Liska <mliska@suse.cz> 10625 10626 * common/config/i386/i386-common.c (ix86_get_valid_option_values): 10627 Use processor_names table. 10628 * config/i386/i386.c (ix86_default_align): Use 10629 processor_cost_table for alignment values. 10630 (ix86_option_override_internal): Use processor_names. 10631 (ix86_function_specific_print): Likewise. 10632 * config/i386/i386.h (struct processor_costs): 10633 Add alignment values. 10634 (struct ptt): Remove and replace with const char *. 10635 * config/i386/x86-tune-costs.h (struct processor_costs): 10636 Declare default alignments for all costs. 10637 106382018-09-17 Aldy Hernandez <aldyh@redhat.com> 10639 10640 * tree-vrp.c (extract_range_from_unary_expr): Do not special case 10641 symbolics or VR_VARYING ranges for ABS_EXPR. 10642 * wide-int-range.cc (wide_int_range_abs): Return positive numbers 10643 when range will wrap. 10644 106452018-09-15 Eric Botcazou <ebotcazou@adacore.com> 10646 10647 PR middle-end/86864 10648 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER 10649 before and after a JUMP_TABLE_DATA. 10650 106512018-09-14 John David Anglin <danglin@gcc.gnu.org> 10652 10653 PR middle-end/87188 10654 * dojump.c (do_compare_and_jump): Canonicalize function pointers 10655 when one operand is a function pointer. Use POINTER_TYPE_P and 10656 FUNC_OR_METHOD_TYPE_P. 10657 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P. 10658 * fold-const.c (build_range_check): Likewise. 10659 * match.pd (simple_comparison): Likewise. 10660 106612018-09-14 David Malcolm <dmalcolm@redhat.com> 10662 10663 PR c/82967 10664 * spellcheck.c (get_edit_distance_cutoff): New function. 10665 (selftest::test_edit_distance_unit_test_oneway): Rename to... 10666 (selftest::test_get_edit_distance_one_way): ...this. 10667 (selftest::test_get_edit_distance_unit): Rename to... 10668 (selftest::test_get_edit_distance_both_ways): ...this. 10669 (selftest::test_edit_distances): Move tests to this new function, 10670 and test some more pairs of strings. Update for above renaming. 10671 (selftest::get_old_cutoff): New function. 10672 (selftest::test_get_edit_distance_cutoff): New function. 10673 (selftest::assert_suggested_for): New function. 10674 (ASSERT_SUGGESTED_FOR): New macro. 10675 (selftest::assert_not_suggested_for): New function. 10676 (ASSERT_NOT_SUGGESTED_FOR): New macro. 10677 (selftest::test_suggestions): New function. 10678 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit 10679 tests to selftest::test_edit_distances and call it. Add calls to 10680 selftest::test_get_edit_distance_cutoff and 10681 selftest::test_suggestions. 10682 * spellcheck.h (get_edit_distance_cutoff): New function declaration. 10683 (best_match::consider): Replace hard-coded cutoff calculation with 10684 a call to... 10685 (best_match::get_cutoff): New declaration. 10686 (best_match::get_best_meaningful_candidate): Likewise. 10687 106882018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de> 10689 10690 * builtins.c (fold_builtin_strlen): Remove TODO comment. 10691 106922018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de> 10693 10694 revert: 10695 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de> 10696 10697 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL 10698 terminated string literal. 10699 107002018-09-14 Martin Sebor <msebor@redhat.com> 10701 10702 * builtins.c (unterminated_array): Handle ARRAY_REF. 10703 (expand_builtin_stpcpy_1): Detect unterminated char arrays. 10704 * builtins.h (unterminated_array): Declare extern. 10705 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated 10706 arrays. 10707 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed 10708 calls. 10709 107102018-09-14 Martin Sebor <msebor@redhat.com> 10711 Jeff Law <law@redhat.com> 10712 10713 * builtins.c (unterminated_array): New. 10714 (expand_builtin_strcpy): Adjust. 10715 (expand_builtin_strcpy_args): Detect unterminated arrays. 10716 * gimple-fold.c (get_maxval_strlen): Add argument. Detect 10717 unterminated arrays. 10718 * gimple-fold.h (get_maxval_strlen): Add argument. 10719 (gimple_fold_builtin_strcpy): Detec unterminated arrays. 10720 10721 * gimple-fold.c (get_range_strlen): Add argument. 10722 (get_maxval_strlen): Adjust. 10723 * gimple-fold.h (get_range_strlen): Add argument. 10724 107252018-09-14 Wei Xiao <wei3.xiao@intel.com> 10726 10727 * config/i386/movdirintrin.h: Fix copyright year. 10728 107292018-09-14 Uros Bizjak <ubizjak@gmail.com> 10730 10731 * reg-stack.c: Include regs.h. 10732 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT. 10733 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of 10734 FIRST_STACK_REG, not DFmode. 10735 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of 10736 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX. 10737 (change stack): Default register mode to the reg_raw_mode of 10738 FIRST_STACK_REG, not DFmode. 10739 * config/i386/i386.md (*swap<mode>): Remove insn pattern. 10740 (*swapxf): Rename from swapxf. 10741 107422018-09-14 Carl Love <cel@us.ibm.com> 10743 10744 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2. 10745 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE. 10746 107472018-09-14 Segher Boessenkool <segher@kernel.crashing.org> 10748 10749 PR target/87224 10750 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y 10751 alternatives. 10752 107532018-09-14 Sam Tebbs <sam.tebbs@arm.com> 10754 10755 PR target/85628 10756 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define. 10757 107582018-09-14 Jason Merrill <jason@redhat.com> 10759 10760 Fix --enable-gather-detailed-mem-stats. 10761 * hash-table.c (hash_table_usage): Change from variable to function. 10762 * hash-table.h: Adjust. 10763 * Makefile.in: Add missing dependencies on hash-table.h. 10764 107652018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 10766 10767 PR tree-optimization/87259 10768 PR lto/87283 10769 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after 10770 execute_cse_reciprocals_1 has tried transforming. 10771 107722018-09-14 Aldy Hernandez <aldyh@redhat.com> 10773 10774 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize 10775 VR_VARYING for PLUS/MINUS_EXPR. 10776 107772018-09-14 Ilya Leoshkevich <iii@linux.ibm.com> 10778 10779 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve 10780 formatting. 10781 107822018-09-14 Richard Biener <rguenther@suse.de> 10783 10784 PR middle-end/63155 10785 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict 10786 bits for the merged partition. 10787 107882018-09-13 Martin Sebor <msebor@redhat.com> 10789 Bernd Edlinger <bernd.edlinger@hotmail.de> 10790 10791 * builtins.h (c_srlen): Add argument. 10792 * builtins.c (warn_string_no_nul): New function. 10793 (c_strlen): Add argument and use it. Update recursive calls. 10794 Pass DECL argument to string_constant to get info on non 10795 terminated strings. Update *NONSTR as needed. 10796 (fold_builtin_strlen): Add argument to calls to c_strlen. 10797 Warn for unterminated arrays. 10798 (warn_string_no_null): Add prototype. 10799 * expr.c (string_constant): Update arguments. Update recursive 10800 calls appropriately. Detect missing NUL terminator and outermost 10801 declaration its missing in. 10802 Improve checks for arrays with nonzero lower bound or elements 10803 that are not a single byte. Simplify offset computation. 10804 Simplify checks for non-NUL terminated strings. 10805 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call. 10806 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code. 10807 108082018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de> 10809 10810 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs 10811 correctly. 10812 * fold-const.c (c_getstr): Fix function comment. Remove unused third 10813 argument. Fix range checks. 10814 * fold-const.h (c_getstr): Adjust protoype. 10815 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when 10816 string is constant but contains no NUL byte. 10817 10818 * expr.c (string_constant): Adjust function comment. 10819 Remove bogus check for zero termination. 10820 10821 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE. 10822 10823 * varasm.c (compare_constant): Compare type size of STRING_CSTs. 10824 (get_constant_size): Don't make STRING_CSTs larger than they are. 10825 (check_string_literal): New check function for STRING_CSTs. 10826 (output_constant): Use it. 10827 108282018-09-13 Eric Botcazou <ebotcazou@adacore.com> 10829 10830 PR target/86812 10831 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define. 10832 108332018-09-13 Richard Biener <rguenther@suse.de> 10834 10835 PR tree-optimization/87263 10836 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes. 10837 (struct unwind_state): Add max_rpo field. 10838 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating. 10839 Compute max_rpo, the max RPO number a block can be backwards reached 10840 from. Re-write non-iterating mode to a RPO ordered worklist approach, 10841 separating it from the iterating mode. 10842 108432018-09-13 Vlad Lazar <vlad.lazar@arm.com> 10844 10845 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST. 10846 (rfs_decision): New scheduling decision. 10847 108482018-09-13 Richard Biener <rguenther@suse.de> 10849 10850 PR bootstrap/87134 10851 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert. 10852 (vn_nary_op_insert_pieces_predicated): Do not write useless 10853 valid_dominated_by_p entry outside of the allocated storage. 10854 108552018-09-13 Omar Sandoval <osandov@osandov.com> 10856 Tom de Vries <tdevries@suse.de> 10857 10858 PR debug/86985 10859 * dwarf2out.c (is_c): New function. 10860 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays. 10861 108622018-09-13 Sam Tebbs <sam.tebbs@arm.com> 10863 10864 PR target/85628 10865 * config/aarch64/aarch64.md (*aarch64_bfxil): 10866 Define. 10867 * config/aarch64/constraints.md (Ulc): Define. 10868 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p): 10869 Define. 10870 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p): 10871 New function. 10872 108732018-09-13 Vlad Lazar <vlad.lazar@arm.com> 10874 10875 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define. 10876 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove 10877 aarch64_layout_frame call. 10878 (aarch64_expand_epilogue): Likewise. 10879 (aarch64_initial_elimination_offset): Likewise. 10880 (aarch64_get_separate_components): Likewise. 10881 (aarch64_use_return_insn_p): Likewise. 10882 (aarch64_layout_frame): Remove unneeded check. 10883 108842018-09-13 Jakub Jelinek <jakub@redhat.com> 10885 10886 * configure.ac: Only append 10887 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to 10888 gcc_config_arguments if it was never reconfigured or last reconfigure 10889 was with different arguments. 10890 * configure: Regenerated. 10891 108922018-09-13 Jakub Jelinek <jakub@redhat.com> 10893 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 10894 10895 PR middle-end/87290 10896 * expr.c (maybe_optimize_pow2p_mod_cmp): New function. 10897 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1. 10898 108992018-09-13 Jakub Jelinek <jakub@redhat.com> 10900 10901 PR tree-optimization/87287 10902 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo 10903 X % C == 0 to X % (unsigned) C == 0 optimization to ... 10904 * match.pd (X % C == 0): ... here. New optimization. 10905 109062018-09-12 Jakub Jelinek <jakub@redhat.com> 10907 10908 PR middle-end/82853 10909 * expr.h (maybe_optimize_mod_cmp): Declare. 10910 * expr.c (mod_inv): New function. 10911 (maybe_optimize_mod_cmp): New function. 10912 (do_store_flag): Use it. 10913 * cfgexpand.c (expand_gimple_cond): Likewise. 10914 109152018-09-09 Cesar Philippidis <cesar@codesourcery.com> 10916 Julian Brown <julian@codesourcery.com> 10917 10918 PR middle-end/86336 10919 * gimplify.c (gimplify_scan_omp_clauses): Set 10920 target_firstprivatize_array_bases in OpenACC parallel and kernels 10921 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from 10922 OpenACC data regions. 10923 109242018-09-12 Uros Bizjak <ubizjak@gmail.com> 10925 10926 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove. 10927 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn. 10928 109292018-09-12 Richard Biener <rguenther@suse.de> 10930 10931 PR tree-optimization/87280 10932 * tree-ssa-sccvn.c (process_bb): Handle the case of executable 10933 edge but unreachable target. 10934 (do_rpo_vn): For conservatively handling a PHI only mark 10935 the backedge executable but not the block reachable. 10936 109372018-09-12 Richard Biener <rguenther@suse.de> 10938 10939 PR tree-optimization/87266 10940 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet 10941 visited blocks. 10942 109432018-09-12 Andreas Krebbel <krebbel@linux.ibm.com> 10944 10945 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New 10946 constants. 10947 ("trunc<BFP:mode><DFP_ALL:mode>2") 10948 ("trunc<DFP_ALL:mode><BFP:mode>2") 10949 ("extend<BFP:mode><DFP_ALL:mode>2") 10950 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode 10951 according to the target operand type. 10952 109532018-09-12 Jakub Jelinek <jakub@redhat.com> 10954 Andreas Krebbel <krebbel@linux.ibm.com> 10955 10956 PR tree-optimization/86844 10957 * gimple-ssa-store-merging.c 10958 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if 10959 there are any overlapping stores in between them, make sure they are 10960 also coalesced or we give up completely. 10961 109622018-09-12 Jakub Jelinek <jakub@redhat.com> 10963 10964 PR middle-end/87248 10965 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that 10966 BIT_AND_EXPR's second operand is a power of two. Formatting fix. 10967 109682018-09-12 Tom de Vries <tdevries@suse.de> 10969 10970 * common.opt (gdescribe-dies): Add option. 10971 * dwarf2out.c (add_name_and_src_coords_attributes): Add description 10972 attribute for artifical and nameless decls. 10973 (dwarf2out_register_external_die): Add description attribute to 10974 external reference die. 10975 (add_desc_attribute): New functions. 10976 (gen_subprogram_die): Add description attribute to 10977 DW_TAG_call_site_parameter. 10978 * tree-pretty-print.c (print_generic_expr_to_str): New function. 10979 * tree-pretty-print.h (print_generic_expr_to_str): Declare. 10980 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and 10981 -gno-describe-dies. 10982 (@item -gdescribe-dies): Add. 10983 109842018-09-12 Aldy Hernandez <aldyh@redhat.com> 10985 10986 * tree-vrp.c (vrp_shift_undefined_p): Remove. 10987 (extract_range_from_binary_expr_1: Call 10988 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p. 10989 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not 10990 depend on sign. 10991 109922018-09-12 Aldy Hernandez <aldyh@redhat.com> 10993 10994 * gimple-ssa-warn-alloca.c 10995 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit 10996 field for ALLOCA_BOUND_*_LARGE. 10997 109982018-09-11 Nathan Sidwell <nathan@acm.org> 10999 11000 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'. 11001 110022018-09-11 Uros Bizjak <ubizjak@gmail.com> 11003 11004 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also 11005 for clobbers. Remove obsolete comment. 11006 110072018-09-11 Uros Bizjak <ubizjak@gmail.com> 11008 11009 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk, 11010 mpxchk, mpxld and mpxst types. 11011 (define_attr length_immediate): Remove all processing of mpx types. 11012 (define_attr prefix_0f): Ditto. 11013 (define_attr memory): Ditto. 11014 110152018-09-11 Uros Bizjak <ubizjak@gmail.com> 11016 11017 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove. 11018 (log<mode>2): Change operand 1 predicate to general_operand. 11019 Extend operand 1 to XFmode and generate logxf3 insn. 11020 (log10<mode>2): Change operand 1 predicate to general_operand. 11021 Extend operand 1 to XFmode and generate log10xf3 insn. 11022 (log2<mode>2): Change operand 1 predicate to general_operand. 11023 Extend operand 1 to XFmode and generate log2xf3 insn. 11024 (fyl2xp1_extend<mode>xf3_i387): Remove. 11025 (log1p<mode>2): Change operand 1 predicate to general_operand. 11026 Extend operand 1 to XFmode and generate log1pxf3 insn. 11027 (fxtract_extend<mode>xf3_i387): Remove. 11028 (logb<mode>2): Change operand 1 predicate to general_operand. 11029 Extend operand 1 to XFmode and generate logbxf3 insn. 11030 (ilogb<mode>2): Change operand 1 predicate to general_operand. 11031 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn. 11032 (significand<mode>2): Change operand 1 predicate to general_operand. 11033 Extend operand 1 to XFmode and generate significandxf3 insn. 11034 110352018-09-11 Nathan Sidwell <nathan@acm.org> 11036 11037 * gcc.c (perror_with_name, pfatal_with_name): Delete. 11038 (load_specs): Use fatal_error. 11039 (DELETE_IF_ORDINARY, process_command): Use error. 11040 (execute, run_attempt): Use fatal_error. 11041 110422018-09-11 Andrew Stubbs <ams@codesourcery.com> 11043 11044 * diagnostic-core.h (sorry_at): New prototype. 11045 * diagnostic.c (sorry_at): New function. 11046 110472018-09-11 Aldy Hernandez <aldyh@redhat.com> 11048 11049 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions 11050 by zero as VR_UNDEFINED. 11051 110522018-09-10 Uros Bizjak <ubizjak@gmail.com> 11053 11054 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387. 11055 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern. 11056 (<sincos>mode2): New expander. 11057 (sincos_extend<mode>xf3_i387): Remove insn pattern. 11058 (sincos -> sin, cos splitters): Remove splitter patterns. 11059 (sincos<mode>3): Change operand 2 predicate to general_operand. 11060 Extend operand 2 to XFmode and generate sincosxf3 insn. 11061 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode. 11062 Change operand 3 predicate to const1_operand. 11063 (fptan_extend<mode>xf4_i387): Remove insn pattern. 11064 (tanxf2): Update operands in the call to fptanxf4_i387. 11065 (tan<mode>2): Change operand 1 predicate to general_operand. 11066 Extend operand 1 to XFmode and generate tanxf3 insn. 11067 (atan2xf3): Rename from *fpatanxf3_i387. 11068 (fpatan_extend<mode>xf3_i387): Remove insn pattern. 11069 (atan2xf3): Remove expander. 11070 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand. 11071 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn. 11072 (atan<mode>2): Change operand 1 predicate to general_operand. 11073 Extend operand 1 to XFmode and generate atanxf3 insn. 11074 110752018-09-10 Uros Bizjak <ubizjak@gmail.com> 11076 11077 * config/i386/i386.md (x87/SSE constant load splitter): Use 11078 memory_operand instead of nonimmediate_operand for input operand 11079 predicate. 11080 110812018-09-09 Uros Bizjak <ubizjak@gmail.com> 11082 11083 * config/i386/i386.md (float partial SSE register stall splitter): Move 11084 splitter near its instruction pattern. 11085 (float_extend partial SSE register stall splitter): Ditto. 11086 (float_truncate partial SSE register stall splitter): Ditto. 11087 110882018-09-09 Hans-Peter Nilsson <hp@bitrange.com> 11089 11090 PR target/86794 11091 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine 11092 to speculation_safe_value_not_needed. 11093 11094 PR target/85666 11095 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size 11096 non-CONST_INT rtx:es using assemble_integer_with_op ".byte". 11097 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call 11098 leaf_function_p, instead use has_hard_reg_initial_val. 11099 111002018-09-09 Nathan Sidwell <nathan@acm.org> 11101 11102 * gcc.h (pfatal_with_name): Don't declare here. 11103 * gcc.c (pfatal_with_name): Make static. 11104 111052018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com> 11106 11107 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add 11108 earlyclobber. 11109 111102018-09-08 John David Anglin <danglin@gcc.gnu.org> 11111 11112 PR rtl-optimization/85458 11113 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust 11114 priority hook to reduce the priority of EXPR. 11115 111162018-09-07 Uros Bizjak <ubizjak@gmail.com> 11117 11118 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable 11119 DImode for x87 on 32bit targets. Conditionally disable x87 modes 11120 with X87_ENABLE_FLOAT. Remove preparation code. 11121 (*float<SWI48:mode><MODEF:mode>2): Rename from 11122 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed 11123 math using "enabled" attribute. 11124 (*floatdi<MODEF:mode>2_i387): Rename from 11125 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and 11126 enable for 32bit targets only. 11127 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New 11128 splitter. 11129 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand 11130 as operand 1 predicate. Rewrite as define_insn_and_split. 11131 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove. 11132 111332018-09-06 Uros Bizjak <ubizjak@gmail.com> 11134 11135 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder 11136 to fallthru to FLOAT case. 11137 111382018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com> 11139 11140 PR target/86731 11141 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic 11142 around folding of vec_sl to handle out of range shift values. 11143 111442018-09-06 Uros Bizjak <ubizjak@gmail.com> 11145 11146 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove. 11147 Update callers to gen_fix_trunc<mode>_i387_fisttp 11148 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to 11149 nonimmediate_operand. 11150 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern 11151 and corresponding splitters. 11152 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387. 11153 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand. 11154 (fix_truncdi_i387_with_temp): Remove insn pattern 11155 and corresponding splitters. 11156 (fix_trunc<mode>_i387): Change operand 0 predicate to 11157 nonimmediate_operand. 11158 (fix_trunc<mode>_i387_with_temp): Remove insn pattern 11159 and corresponding splitters. 11160 (*fistdi2_1): Remove. 11161 (fistdi2): Ditto. 11162 (fistdi2_with_temp): Remove insn pattern and corresponding splitters. 11163 (lrintxfdi2): Remove expander. Reimplement as define_insn. 11164 (*fist<mode>2_1): Remove. 11165 (fist<mode>2): Ditto. 11166 (fist<mode>2_with_temp): Remove insn pattern and corresponding 11167 splitters. 11168 (lrintxf<mode>2): Remove expander. Reimplement as define_insn. 11169 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>. 11170 (fistdi2_<rounding>): Change operand 0 predicate to 11171 nonimmediate_operand. 11172 (fistdi2_<rounding>_with_temp): Remove insn pattern 11173 and corresponding splitters. 11174 (fist<mode>2_<rounding>): Change operand 0 predicate to 11175 nonimmediate_operand. 11176 (fist<mode>2_<rounding>_with_temp): Remove insn pattern 11177 and corresponding splitters. 11178 11179 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH. 11180 111812018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de> 11182 11183 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of 11184 the init value. 11185 111862018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com> 11187 11188 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for 11189 early gimple folding of vec_splat(). 11190 * tree-vect-generic.c: Remove static from tree_vec_extract() definition. 11191 * gimple-fold.h: Add an extern define for tree_vec_extract(). 11192 111932018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com> 11194 11195 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p 11196 wrappers around TREE_TYPE comparisons. 11197 111982018-09-06 Ilya Leoshkevich <iii@linux.ibm.com> 11199 11200 PR target/80080 11201 * config/s390/predicates.md: Add nonsym_memory_operand. 11202 * config/s390/s390.c (s390_legitimize_cs_operand): If operand 11203 contains a SYMBOL_REF, load it into an intermediate pseudo. 11204 (s390_emit_compare_and_swap): Legitimize operand. 11205 * config/s390/s390.md: Use the new nonsym_memory_operand 11206 with UNSPECV_CAS patterns. 11207 112082018-09-06 Ilya Leoshkevich <iii@linux.ibm.com> 11209 11210 PR target/80080 11211 * config/s390/s390-passes.def: New file. 11212 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward 11213 declaration. 11214 (make_pass_s390_early_mach): Add declaration. 11215 * config/s390/s390.c (make_pass_s390_early_mach): 11216 (s390_option_override): Remove dynamic registration. 11217 * config/s390/t-s390: Add s390-passes.def. 11218 112192018-09-06 Ilya Leoshkevich <iii@linux.ibm.com> 11220 11221 * config/s390/s390.c (s390_decompose_constant_pool_ref): 11222 Remove UNSPEC_LTREL_BASE check. 11223 (annotate_constant_pool_refs): Likewise. 11224 (find_constant_pool_ref): Likewise. 11225 (find_ltrel_base): Removed. 11226 (replace_ltrel_base): Removed. 11227 (s390_mainpool_finish): Remove replace_ltrel_base call. 11228 (s390_chunkify_start): Remove pending LTREL_BASE logic. 11229 (s390_chunkify_finish): Remove replace_ltrel_base call. 11230 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE. 11231 112322018-09-06 Hans-Peter Nilsson <hp@axis.com> 11233 11234 PR target/86779 11235 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine 11236 to speculation_safe_value_not_needed. 11237 112382018-09-05 Cesar Philippidis <cesar@codesourcery.com> 11239 Bernd Schmidt <bernds_cb1@t-online.de> 11240 11241 * config/nvptx/nvptx-opts.h: New file. 11242 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target. 11243 * config/nvptx/nvptx.h: Include "nvptx-opts.h". 11244 (ASM_SPEC): Define. 11245 (TARGET_SM35): New macro. 11246 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the 11247 correct predicate. 11248 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its 11249 values. 11250 (misa=): New option. 11251 * doc/invoke.texi (Nvidia PTX Options): Document -misa. 11252 112532018-09-05 Uros Bizjak <ubizjak@gmail.com> 11254 11255 * config/i386/i386.md (truncdfsf2): Remove expander. 11256 (truncdfsf2_with_temp): Ditto. 11257 (truncxf<mode>2): Ditto. 11258 (*truncdfsf_fast_mixed): Remove insn pattern. 11259 (*truncdfsf_fast_i387): Ditto. 11260 (*truncdfsf_mixed): Ditto. 11261 (*truncdfsf_i387): Ditto. 11262 (*truncdfsf2_i387_1): Ditto. 11263 (*truncxfsf2_mixed): Ditto. 11264 (*truncxfdf2_mixed): Ditto. 11265 (*truncxf<mode>2_i387_noop): Ditto. Update callers 11266 to call gen_truncxf<mode>2 instead. 11267 (*truncxf<mode>2_i387): Remove. 11268 (reg->reg splitters): Remove splitter pattern. 11269 (reg->mem splitters): Ditto. 11270 11271 (truncdfsf2): New insn pattern. 11272 (truncxf<mode>2): Ditto. 11273 112742018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 11275 11276 * tree-ssa-math-opts.c (is_mult_by): New function. 11277 (is_square_of): Use the above. 11278 (optimize_recip_sqrt): New function. 11279 (pass_cse_reciprocals::execute): Use the above. 11280 112812018-09-05 Richard Biener <rguenther@suse.de> 11282 11283 PR bootstrap/87134 11284 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure 11285 to zero-init the emplaced vec. 11286 112872018-09-05 Martin Liska <mliska@suse.cz> 11288 11289 PR tree-optimization/87205 11290 * tree-switch-conversion.c (pass_lower_switch::execute): 11291 Group cases for switch statements. 11292 112932018-09-05 Richard Biener <rguenther@suse.de> 11294 11295 PR tree-optimization/87217 11296 * tree-ssa-sccvn.c (vuse_valueize): New. 11297 (vn_reference_lookup_pieces): Use it. 11298 (vn_reference_lookup): Likewise. 11299 113002018-09-05 Nathan Sidwell <nathan@acm.org> 11301 11302 PR c++/87137 11303 * stor-layout.c (place_field): Scan forwards to check last 11304 bitfield when ms_bitfield_placement is in effect. 11305 113062018-09-05 Richard Biener <rguenther@suse.de> 11307 11308 PR bootstrap/87225 11309 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus 11310 return. 11311 113122018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org> 11313 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org> 11314 11315 * config/aarch64/falkor-tag-collision-avoidance.c: New file. 11316 * config.gcc (extra_objs): Build it. 11317 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o): 11318 Likewise. 11319 * config/aarch64/aarch64-passes.def 11320 (pass_tag_collision_avoidance): New pass. 11321 * config/aarch64/aarch64.c (qdf24xx_tunings): Add 11322 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags. 11323 (aarch64_classify_address): Remove static qualifier. 11324 (aarch64_address_info, aarch64_address_type): Move to... 11325 * config/aarch64/aarch64-protos.h: ... here. 11326 (make_pass_tag_collision_avoidance): New function. 11327 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs): 11328 New tuning flag. 11329 113302018-09-05 Martin Liska <mliska@suse.cz> 11331 11332 * doc/gcov.texi: Update documentation of humar 11333 readable mode. 11334 * gcov.c (format_count): Print one decimal place, it provides 11335 more fine number of situations like '1G' vs. '1.4G'. 11336 113372018-09-05 Martin Liska <mliska@suse.cz> 11338 11339 PR target/87164 11340 * config/rs6000/rs6000.opt: Mark the option as Deprecated. 11341 * optc-gen.awk: Allow 'Var' for Deprecated options in order 11342 to generate a MASK value. 11343 113442018-09-04 H.J. Lu <hongjiu.lu@intel.com> 11345 11346 PR debug/86593 11347 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even 11348 if frame pointer isn't used. 11349 (compute_frame_pointer_to_fb_displacement): Likewise. 11350 113512018-09-04 Jakub Jelinek <jakub@redhat.com> 11352 11353 PR target/87198 11354 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET, 11355 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use 11356 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE. 11357 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET 11358 and OPTION_MASK_ISA_XSAVEC_UNSET. 11359 113602018-09-04 Max Filippov <jcmvbkbc@gmail.com> 11361 11362 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and 11363 XOR operations in NAND case. 11364 113652018-09-04 Aldy Hernandez <aldyh@redhat.com> 11366 11367 * wide-int-range.cc (wide_int_range_convert): New. 11368 * wide-int-range.h (wide_int_range_convert): New. 11369 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int 11370 code into wide_int_range_convert. 11371 (extract_range_into_wide_ints): Do not munge anti range constants 11372 into the entire domain. Just return the range back. 11373 113742018-09-04 Martin Liska <mliska@suse.cz> 11375 11376 * genmatch.c (output_line_directive): Add new argument 11377 fnargs. 11378 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect. 11379 113802018-09-04 Jonathan Wakely <jwakely@redhat.com> 11381 11382 * doc/invoke.texi (Option Summary): Add whitespace. 11383 11384 * doc/invoke.texi (Option Summary): Add -Waligned-new. 11385 113862018-09-04 Richard Biener <rguenther@suse.de> 11387 11388 PR tree-optimization/87211 11389 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a 11390 backedge value we're supposed to treat as VARYING also number 11391 the PHI to VARYING in case it got a different value-number already. 11392 113932018-09-04 Aldy Hernandez <aldyh@redhat.com> 11394 11395 * tree-vrp.c (vrp_can_optimize_bit_op): Remove. 11396 (extract_range_from_binary_expr_1): Do not call 11397 vrp_can_optimize_bit_op. 11398 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make 11399 static. 11400 (wide_int_range_get_mask_and_bounds): New. 11401 (wide_int_range_optimize_bit_op): New. 11402 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op. 11403 (wide_int_range_bit_and): Same. 11404 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove. 11405 (wide_int_range_optimize_bit_op): New. 11406 (wide_int_range_get_mask_and_bounds): New. 11407 114082018-09-04 Richard Biener <rguenther@suse.de> 11409 11410 PR tree-optimization/87176 11411 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame 11412 variable. When value-numbering a virtual PHI node make sure 11413 to not value-number to the backedge value. 11414 114152018-09-04 Jonathan Wakely <jwakely@redhat.com> 11416 11417 * doc/extend.texi (Long Long, Hex Floats): Document support for 11418 long long and hex floats in more recent versions of ISO C++. 11419 114202018-09-03 Richard Biener <rguenther@suse.de> 11421 11422 PR tree-optimization/87177 11423 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep 11424 cleanup. 11425 114262018-09-03 Alexander Monakov <amonakov@ispras.ru> 11427 11428 * bb-reorder.c (edge_order): Convert to C-qsort-style 11429 tri-state comparator. 11430 (reorder_basic_blocks_simple): Change std::stable_sort to 11431 gcc_stablesort. 11432 114332018-09-03 Alexander Monakov <amonakov@ispras.ru> 11434 11435 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style 11436 tri-state comparator. 11437 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort. 11438 114392018-09-03 Alexander Monakov <amonakov@ispras.ru> 11440 11441 * sort.cc (struct sort_ctx): New field 'nlim'. Use it... 11442 (mergesort): ... here as maximum count for using netsort. 11443 (gcc_qsort): Set nlim to 3 if stable sort is requested. 11444 (gcc_stablesort): New. 11445 * system.h (gcc_stablesort): Declare. 11446 114472018-09-03 Alexander Monakov <amonakov@ispras.ru> 11448 11449 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk. 11450 * system.h (qsort): Always redirect to gcc_qsort. Update comment. 11451 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment. 11452 114532018-09-03 Segher Boessenkool <segher@kernel.crashing.org> 11454 11455 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the 11456 lxsdx and stxsdx alternatives. 11457 (*mov<mode>_hardfloat64): Ditto. 11458 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto. 11459 114602018-09-03 Richard Biener <rguenther@suse.de> 11461 11462 PR tree-optimization/87200 11463 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a 11464 simplify result. 11465 114662018-09-03 Martin Liska <mliska@suse.cz> 11467 11468 PR tree-optimization/87201 11469 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes): 11470 Fix parenthesis in an expression. 11471 114722018-09-03 Richard Biener <rguenther@suse.de> 11473 11474 PR tree-optimization/87197 11475 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def 11476 visited. CSE the VN_INFO hashtable lookup. 11477 11478 PR tree-optimization/87169 11479 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not 11480 iterating make sure there's no extra backedges from irreducible 11481 regions feeding the header. Mark the destination block 11482 executable. 11483 114842018-09-03 Martin Liska <mliska@suse.cz> 11485 11486 PR driver/83193 11487 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES. 11488 * common/common-targhooks.c (default_get_valid_option_values): 11489 New function. 11490 * common/common-targhooks.h (default_get_valid_option_values): 11491 Likewise. 11492 * common/config/i386/i386-common.c: Move processor_target_table 11493 from i386.c. 11494 (ix86_get_valid_option_values): New function. 11495 (TARGET_GET_VALID_OPTION_VALUES): New macro. 11496 * config/i386/i386.c (struct ptt): Move to i386-common.c. 11497 (PTA_*): Move all defined masks into i386-common.c. 11498 (ix86_function_specific_restore): Use new processor_cost_table. 11499 * config/i386/i386.h (struct ptt): Moved from i386.c. 11500 (struct pta): Likewise. 11501 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES. 11502 * doc/tm.texi.in: Likewise. 11503 * opt-suggestions.c (option_proposer::suggest_option): 11504 Pass prefix to build_option_suggestions. 11505 (option_proposer::get_completions): Likewise. 11506 (option_proposer::build_option_suggestions): Use the new target 11507 hook. 11508 * opts.c (struct option_help_tuple): New struct. 11509 (print_filtered_help): Use the new target hook. 11510 115112018-09-03 Martin Liska <mliska@suse.cz> 11512 11513 PR middle-end/59521 11514 * predict.c (set_even_probabilities): Add likely_edges 11515 argument and handle cases where we have precisely one 11516 likely edge. 11517 (combine_predictions_for_bb): Catch also likely_edges. 11518 (tree_predict_by_opcode): Handle gswitch statements. 11519 * tree-cfg.h (find_case_label_for_value): New declaration. 11520 (find_taken_edge_switch_expr): Likewise. 11521 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes): 11522 Find pivot in decision tree based on probabily, not by number of 11523 nodes. 11524 115252018-09-02 Gerald Pfeifer <gerald@pfeifer.com> 11526 11527 * doc/standards.texi (Standards): Update Objective-C reference. 11528 115292018-09-01 Gerald Pfeifer <gerald@pfeifer.com> 11530 11531 * doc/install.texi (Prerequisites): Update link for MPC. 11532 115332018-09-01 Michael Matz <matz@suse.de> 11534 11535 PR tree-optimization/87074 11536 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit 11537 PHIs for outer-loop uses. 11538 115392018-09-01 Gerald Pfeifer <gerald@pfeifer.com> 11540 11541 * doc/generic.texi (OpenMP): Adjust link to openmp.org. 11542 * doc/invoke.texi (C Dialect Options): Ditto. 11543 115442018-09-01 Gerald Pfeifer <gerald@pfeifer.com> 11545 11546 * doc/install.texi (Prerequisites): Adjust link mpfr.org. 11547 115482018-08-31 Richard Biener <rguenther@suse.de> 11549 11550 PR tree-optimization/87168 11551 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter. 11552 (rpo_elim::eliminate_avail): When OP was not visited it must 11553 be available. 11554 115552018-08-31 David Malcolm <dmalcolm@redhat.com> 11556 11557 * tree-vrp.c (copy_value_range): Convert param "from" from 11558 "value_range *" to "const value_range *". 11559 (range_is_null): Likewise for param "vr". 11560 (range_int_cst_p): Likewise. 11561 (range_int_cst_singleton_p): Likewise. 11562 (symbolic_range_p): Likewise. 11563 (value_ranges_intersect_p): Likewise for both params. 11564 (value_range_nonnegative_p): Likewise for param "vr". 11565 (value_range_constant_singleton): Likewise. 11566 (vrp_set_zero_nonzero_bits): Likewise for param "ar". 11567 (extract_range_into_wide_ints): Likewise for param "vr". 11568 (extract_range_from_multiplicative_op): Likewise for params "vr0" 11569 and "vr1". 11570 (vrp_can_optimize_bit_op): Likewise. 11571 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and 11572 "vr1_". 11573 (extract_range_from_unary_expr): Likewise. 11574 (debug_value_range): Likewise for param "vr". 11575 (value_range::dump): Add "const" qualifier. 11576 (vrp_prop::check_array_ref): Convert local "vr" from 11577 "value_range *" to "const value_range *". 11578 (vrp_prop::check_mem_ref): Likewise. 11579 (vrp_prop::visit_stmt): Likewise for local "old_vr". 11580 (vrp_intersect_ranges_1): Likewise for param "vr_1". 11581 (vrp_intersect_ranges): Likewise. 11582 (simplify_stmt_for_jump_threading): Likewise for local "vr". 11583 (vrp_prop::vrp_finalize): Likewise. 11584 * tree-vrp.h (value_range::dump): Add "const" qualifier. 11585 (vrp_intersect_ranges): Add "const" qualifier to params as above. 11586 (extract_range_from_unary_expr): Likewise. 11587 (value_range_constant_singleton): Likewise. 11588 (symbolic_range_p): Likewise. 11589 (copy_value_range): Likewise. 11590 (extract_range_from_binary_expr_1): Likewise. 11591 (range_int_cst_p): Likewise. 11592 (vrp_set_zero_nonzero_bits): Likewise. 11593 (range_int_cst_singleton_p): Likewise. 11594 115952018-08-31 Vlad Lazar <vlad.lazar@arm.com> 11596 11597 * config/aarch64/arm_neon.h (vabsd_s64): New. 11598 (vnegd_s64): Likewise. 11599 116002018-08-31 Martin Jambor <mjambor@suse.cz> 11601 11602 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN. 11603 116042018-08-31 Martin Liska <mliska@suse.cz> 11605 11606 * ipa-icf.c (sem_item::add_type): Use 11607 sem_item::m_type_hash_cache. 11608 * ipa-icf.h: Move the cache from sem_item_optimizer 11609 to sem_item. 11610 116112018-08-31 Nathan Sidwell <nathan@acm.org> 11612 11613 * doc/extend.texi (Backwards Compatibility): Remove implicit 11614 extern C leeway of () being (...). 11615 116162018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 11617 11618 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment. 11619 116202018-08-31 Segher Boessenkool <segher@kernel.crashing.org> 11621 11622 PR target/86684 11623 PR target/87149 11624 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND. 11625 116262018-08-31 Jakub Jelinek <jakub@redhat.com> 11627 11628 PR middle-end/87138 11629 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of 11630 gen_int_mode. Formatting fixes. 11631 116322018-08-30 Sandra Loosemore <sandra@codesourcery.com> 11633 11634 * target.def (custom_function_descriptors): Improve documentation. 11635 * doc/tm.texi.in (Trampolines): Expand discussion of function 11636 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the 11637 beginning of the section. 11638 * doc/tm.texi: Regenerated. 11639 116402018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com> 11641 11642 * cfg.h (class auto_edge_flag): Spell out the template-id of the 11643 base class in the initializer list. This is a workaround for 11644 building with older GCC. 11645 (class auto_bb_flag): Likewise. 11646 116472018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com> 11648 11649 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star. 11650 (altivec_vcmpequ<VI_char>_p): Remove star. 11651 * config/rs6000/rs6000-string.c (do_load_for_compare): Support 11652 vector load modes. 11653 (expand_strncmp_vec_sequence): New function. 11654 (emit_final_str_compare_vec): New function. 11655 (expand_strn_compare): Add support for vector strncmp. 11656 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change 11657 length specification to bytes. 11658 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star. 11659 (vcmpnezb_p): New pattern. 11660 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation 11661 for option -mstring-compare-inline-limit. 11662 116632018-08-30 Thiago Macieira <thiago.macieira@intel.com> 11664 11665 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES. 11666 (PTA_SKYLAKE): Add PTA_AES. 11667 (PTA_GOLDMONT): Likewise. 11668 116692018-08-29 Jan Hubicka <jh@suse.cz> 11670 11671 PR lto/86517 11672 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode. 11673 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE. 11674 116752018-08-29 Jan Hubicka <jh@suse.cz> 11676 11677 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow 11678 TYPE_STUB_DECL. 11679 (hash_tree): Do not visit TYPE_STUB_DECL. 11680 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not 11681 stream TYPE_STUB_DECL. 11682 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise. 11683 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL 11684 after free_lang_data. 11685 (type_in_anonymous_namespace_p): Likewise. 11686 116872018-08-29 Jan Hubicka <jh@suse.cz> 11688 11689 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary 11690 comment that it has to be even number. 11691 (class sreal): Change m_sig type to int32_t. 11692 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use 11693 int64_t for temporary calculations. 11694 (sreal_verify_basics): Drop one bit from minimum and maximum. 11695 116962018-08-30 Richard Biener <rguenther@suse.de> 11697 11698 PR tree-optimization/87147 11699 * tree-ssa-sccvn.c (SSA_VISITED): New function. 11700 (visit_phi): When the degenerate result is from the backedge and 11701 we didn't visit its definition yet drop to VARYING. 11702 (do_rpo_vn): Properly mark blocks with incoming backedges as executable. 11703 117042018-08-29 Jan Hubicka <jh@suse.cz> 11705 11706 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk 11707 DECL_VINDEX. 11708 (hash_tree): Likewise. 11709 117102018-08-29 Jan Hubicka <jh@suse.cz> 11711 11712 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO 11713 and TYPE_NEXT_REF_TO. 11714 117152018-08-29 Jan Hubicka <jh@suse.cz> 11716 11717 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary 11718 comment that it has to be even number. 11719 (class sreal): Change m_sig type to int32_t. 11720 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use 11721 int64_t for temporary calculations. 11722 (sreal_verify_basics): Drop one bit from minimum and maximum. 11723 117242018-08-30 Tamar Christina <tamar.christina@arm.com> 11725 11726 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max. 11727 117282018-08-30 Vlad Lazar <vlad.lazar@arm.com> 11729 11730 PR middle-end/86995 11731 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add 11732 if to_add is negative. 11733 117342018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de> 11735 11736 PR middle-end/87053 11737 * builtins.c (c_strlen): Improve range checks. 11738 117392018-08-29 Martin Sebor <msebor@redhat.com> 11740 Jeff Law <law@redhat.com> 11741 11742 PR tree-optimization/86714 11743 PR tree-optimization/86711 11744 * builtins.c (c_strlen): Add arguments to call to string_constant. 11745 * expr.c (string_constant): Add argument. Detect missing nul 11746 terminator and outermost declaration it's missing in. 11747 * expr.h (string_constant): Add argument. 11748 * fold-const.c (read_from_constant_string): Add arguments to call to 11749 string_constant. 11750 (c_getstr): Likewise. 11751 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise. 11752 to string_constant. 11753 * tree-ssa-strlen.c (get_stridx): Likewise. 11754 117552018-08-29 Jan Hubicka <jh@suse.cz> 11756 11757 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers): 11758 Do not stream DECL_VINDEX. 11759 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise. 11760 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX. 11761 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX. 11762 117632018-08-29 Richard Biener <rguenther@suse.de> 11764 11765 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited 11766 virtual operands that are not default defs to honor region 11767 boundaries. 11768 (rpo_vn_valueize): Remove ineffective code here. 11769 117702018-08-29 Richard Biener <rguenther@suse.de> 11771 11772 PR tree-optimization/87132 11773 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate 11774 when skipping defs reachable over backedges. 11775 117762018-08-29 Richard Biener <rguenther@suse.de> 11777 11778 * tree-core.h: Document use of deprecated_flag in SSA_NAME. 11779 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define. 11780 * tree-into-ssa.c (pass_build_ssa::execute): Initialize 11781 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec. 11782 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove. 11783 (vn_reference_lookup_3): Remove use of const_parms. 11784 (free_rpo_vn): Do not free const_parms. 11785 (do_rpo_vn): Do not call init_const_parms. 11786 * tree-ssa-alias.c (refs_may_alias_p_1): Honor 11787 SSA_NAME_POINTS_TO_READONLY_MEMORY. 11788 (call_may_clobber_ref_p_1): Likewise. 11789 117902018-08-29 Alexander Monakov <amonakov@ispras.ru> 11791 11792 PR other/86726 11793 * invoke.texi (Optimization Options): List -ftree-scev-cprop. 11794 (-O): Ditto. 11795 (-ftree-scev-cprop): Document. 11796 117972018-08-29 Jan Hubicka <jh@suse.cz> 11798 11799 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp 11800 parameters. 11801 (sreal constructor): Update. 11802 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*, 11803 sreal:operator/): Update. 11804 118052018-08-29 Martin Liska <mliska@suse.cz> 11806 11807 * tree-switch-conversion.c (switch_conversion::expand): 11808 Strenghten assumption about gswitch statements. 11809 118102018-08-29 Richard Biener <rguenther@suse.de> 11811 11812 PR tree-optimization/87117 11813 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only 11814 re-value-number released SSA VDEFs. 11815 118162018-08-29 Richard Biener <rguenther@suse.de> 11817 11818 PR tree-optimization/87126 11819 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert. 11820 118212018-08-28 Jim Wilson <jimw@sifive.com> 11822 11823 * config/riscv/pic.md: Rewrite. 11824 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for 11825 invalid address. 11826 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete. 11827 (SOFTF, default_load, softload, softstore): New. 11828 118292018-08-28 Jeff Law <law@redhat.com> 11830 11831 * fold-const.c (fold_binary_loc): Remove recently added assert. 11832 118332018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com> 11834 11835 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED 11836 to OP parmeter of generated function. 11837 118382018-08-28 MCC CS <deswurstes@users.noreply.github.com> 11839 11840 PR tree-optimization/87009 11841 * match.pd: Add boolean optimizations. 11842 118432018-08-28 Martin Sebor <msebor@redhat.com> 11844 11845 PR middle-end/86631 11846 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special. 11847 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function. 11848 (pass_walloca::gate): Use it. 11849 (alloca_call_type): Same. 11850 (pass_walloca::execute): Same. 11851 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special. 11852 118532018-08-28 David Malcolm <dmalcolm@redhat.com> 11854 11855 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on 11856 GCC_VERSION for usage of "__gcc_dump_printf__" format from 11857 >= 3005 to >= 9000. 11858 118592018-08-28 Richard Biener <rguenther@suse.de> 11860 11861 PR tree-optimization/87124 11862 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against 11863 constants before looking up avail. 11864 118652018-08-28 Jakub Jelinek <jakub@redhat.com> 11866 11867 PR middle-end/87099 11868 * calls.c (maybe_warn_nonstring_arg): Punt early if 11869 warn_stringop_overflow is zero. Don't call get_range_strlen 11870 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST. 11871 Swap comparison operands to have constants on rhs. Only use 11872 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly 11873 increment lenrng[0]. 11874 118752018-08-28 Richard Sandiford <richard.sandiford@arm.com> 11876 11877 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded 11878 use of tree_to_shwi. Remove duplicated test for the size being 11879 a whole number of bytes. 11880 118812018-08-28 Richard Biener <rguenther@suse.de> 11882 11883 PR tree-optimization/87117 11884 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): 11885 Handle removed stmt without LHS (GIMPLE_NOP). 11886 118872018-08-28 Richard Biener <rguenther@suse.de> 11888 11889 PR tree-optimization/87117 11890 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude 11891 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P. 11892 118932018-08-28 Richard Biener <rguenther@suse.de> 11894 11895 PR tree-optimization/87117 11896 * tree-ssa-pre.c (compute_avail): Do not make expressions 11897 with predicated values available. 11898 (get_expr_value_id): Assert we do not run into predicated value 11899 expressions. 11900 119012018-08-28 Richard Biener <rguenther@suse.de> 11902 11903 PR tree-optimization/87117 11904 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may 11905 get virtual operands. 11906 (get_expr_operands): Handle STRING_CST like other decls. 11907 119082018-08-28 Martin Liska <mliska@suse.cz> 11909 11910 * tree.h: Update documentation of fndecl_built_in_p 11911 functions. 11912 11913 119142018-08-27 Jeff Law <law@redhat.com> 11915 PR tree-optimization/87110 11916 * tree-ssa-dse.c (compute_trims): Handle non-constant 11917 TYPE_SIZE_UNIT. 11918 119192018-08-27 Martin Sebor <msebor@redhat.com> 11920 11921 PR tree-optimization/86914 11922 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF. 11923 119242018-08-27 Martin Sebor <msebor@redhat.com> 11925 11926 PR tree-optimization/87112 11927 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to 11928 the type of the bound argument. 11929 119302018-08-27 Jeff Law <law@redhat.com> 11931 11932 * tree-ssa-dse.c (compute_trims): Handle case where the reference's 11933 type does not have a TYPE_SIZE_UNIT. 11934 119352018-08-27 Steve Ellcey <sellcey@cavium.com> 11936 11937 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h 11938 with include of backend.h. 11939 119402018-08-27 Richard Biener <rguenther@suse.de> 11941 11942 PR tree-optimization/86927 11943 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly 11944 use const cond reduction code. 11945 119462018-08-27 Alexander Monakov <amonakov@ispras.ru> 11947 11948 PR tree-optimization/85758 11949 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression. 11950 119512018-08-27 David Malcolm <dmalcolm@redhat.com> 11952 11953 PR c++/87091 11954 * diagnostic-show-locus.c (class layout_range): Update for 11955 conversion of show_caret_p to a tri-state. 11956 (layout_range::layout_range): Likewise. 11957 (make_range): Likewise. 11958 (layout::maybe_add_location_range): Likewise. 11959 (layout::should_print_annotation_line_p): Don't show annotation 11960 lines for ranges that are SHOW_LINES_WITHOUT_RANGE. 11961 (layout::get_state_at_point): Update for conversion of 11962 show_caret_p to a tri-state. Bail out early for 11963 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect 11964 underlining or source colorization. 11965 (gcc_rich_location::add_location_if_nearby): Update for conversion 11966 of show_caret_p to a tri-state. 11967 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise. 11968 (selftest::test_one_liner_fixit_replace_equal_secondary_range): 11969 Likewise. 11970 (selftest::test_one_liner_labels): Likewise. 11971 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for 11972 conversion of show_caret_p to a tri-state. 11973 * pretty-print.c (text_info::set_location): Likewise. 11974 * pretty-print.h (text_info::set_location): Likewise. 11975 * substring-locations.c (format_warning_n_va): Likewise. 11976 * tree-diagnostic.c (default_tree_printer): Likewise. 11977 * tree-pretty-print.c (newline_and_indent): Likewise. 11978 119792018-08-27 David Malcolm <dmalcolm@redhat.com> 11980 11981 PR c++/87091 11982 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the 11983 line above for line-insertion fix-it hints. 11984 (selftest::test_fixit_insert_containing_newline): Update the 11985 expected results, and add a test with line-numbering enabled. 11986 119872018-08-27 Martin Liska <mliska@suse.cz> 11988 11989 PR sanitizer/86962 11990 * sanopt.c (sanitize_rewrite_addressable_params): Ignore 11991 params with DECL_HAS_VALUE_EXPR_P. 11992 119932018-08-27 Martin Liska <mliska@suse.cz> 11994 11995 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump 11996 selected expansion strategy. 11997 119982018-08-27 Martin Liska <mliska@suse.cz> 11999 12000 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p. 12001 * builtins.c (is_builtin_fn): Likewise. 12002 * attribs.c (diag_attr_exclusions): Use new function 12003 fndecl_built_in_p and remove check for FUNCTION_DECL if 12004 possible. 12005 (builtin_mathfn_code): Likewise. 12006 (fold_builtin_expect): Likewise. 12007 (fold_call_expr): Likewise. 12008 (fold_builtin_call_array): Likewise. 12009 (fold_call_stmt): Likewise. 12010 (set_builtin_user_assembler_name): Likewise. 12011 (is_simple_builtin): Likewise. 12012 * calls.c (gimple_alloca_call_p): Likewise. 12013 (maybe_warn_nonstring_arg): Likewise. 12014 * cfgexpand.c (expand_call_stmt): Likewise. 12015 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise. 12016 (cgraph_edge::verify_corresponds_to_fndecl): Likewise. 12017 (cgraph_node::verify_node): Likewise. 12018 * cgraphclones.c (build_function_decl_skip_args): Likewise. 12019 (cgraph_node::create_clone): Likewise. 12020 * config/arm/arm.c (arm_insert_attributes): Likewise. 12021 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise. 12022 * dse.c (scan_insn): Likewise. 12023 * expr.c (expand_expr_real_1): Likewise. 12024 * fold-const.c (operand_equal_p): Likewise. 12025 (fold_binary_loc): Likewise. 12026 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise. 12027 * gimple-low.c (lower_stmt): Likewise. 12028 * gimple-pretty-print.c (dump_gimple_call): Likewise. 12029 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): 12030 Likewise. 12031 * gimple.c (gimple_build_call_from_tree): Likewise. 12032 (gimple_call_builtin_p): Likewise. 12033 (gimple_call_combined_fn): Likewise. 12034 * gimplify.c (gimplify_call_expr): Likewise. 12035 (gimple_boolify): Likewise. 12036 (gimplify_modify_expr): Likewise. 12037 (gimplify_addr_expr): Likewise. 12038 * hsa-gen.c (gen_hsa_insns_for_call): Likewise. 12039 * ipa-cp.c (determine_versionability): Likewise. 12040 * ipa-fnsummary.c (compute_fn_summary): Likewise. 12041 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise. 12042 * ipa-split.c (visit_bb): Likewise. 12043 (split_function): Likewise. 12044 * ipa-visibility.c (cgraph_externally_visible_p): Likewise. 12045 * lto-cgraph.c (input_node): Likewise. 12046 * lto-streamer-out.c (write_symbol): Likewise. 12047 * omp-low.c (setjmp_or_longjmp_p): Likewise. 12048 (lower_omp_1): Likewise. 12049 * predict.c (strip_predict_hints): Likewise. 12050 * print-tree.c (print_node): Likewise. 12051 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise. 12052 * trans-mem.c (is_tm_irrevocable): Likewise. 12053 (is_tm_load): Likewise. 12054 (is_tm_simple_load): Likewise. 12055 (is_tm_store): Likewise. 12056 (is_tm_simple_store): Likewise. 12057 (is_tm_abort): Likewise. 12058 (tm_region_init_1): Likewise. 12059 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise. 12060 * tree-cfg.c (verify_gimple_call): Likewise. 12061 (move_stmt_r): Likewise. 12062 (stmt_can_terminate_bb_p): Likewise. 12063 * tree-eh.c (lower_eh_constructs_2): Likewise. 12064 * tree-if-conv.c (if_convertible_stmt_p): Likewise. 12065 * tree-inline.c (remap_gimple_stmt): Likewise. 12066 (copy_bb): Likewise. 12067 (estimate_num_insns): Likewise. 12068 (fold_marked_statements): Likewise. 12069 * tree-sra.c (scan_function): Likewise. 12070 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise. 12071 (optimize_stack_restore): Likewise. 12072 (pass_fold_builtins::execute): Likewise. 12073 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise. 12074 (mark_all_reaching_defs_necessary_1): Likewise. 12075 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise. 12076 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise. 12077 (pass_forwprop::execute): Likewise. 12078 * tree-ssa-loop-im.c (stmt_cost): Likewise. 12079 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise. 12080 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise. 12081 * tree-ssa-strlen.c (get_string_length): Likewise. 12082 * tree-ssa-structalias.c (handle_lhs_call): Likewise. 12083 (find_func_aliases_for_call): Likewise. 12084 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise. 12085 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise. 12086 * tree-tailcall.c (find_tail_calls): Likewise. 12087 * tree.c (need_assembler_name_p): Likewise. 12088 (free_lang_data_in_decl): Likewise. 12089 (get_call_combined_fn): Likewise. 12090 * ubsan.c (is_ubsan_builtin_p): Likewise. 12091 * varasm.c (incorporeal_function_p): Likewise. 12092 * tree.h (DECL_BUILT_IN): Remove and replace with 12093 fndecl_built_in_p. 12094 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p. 12095 (fndecl_built_in_p): New. 12096 120972018-08-27 Martin Liska <mliska@suse.cz> 12098 12099 PR tree-optimization/86847 12100 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes): 12101 Dump also subtree probability. 12102 (switch_decision_tree::do_jump_if_equal): New function. 12103 (switch_decision_tree::emit_case_nodes): Handle special 12104 situations in balanced tree that can be emitted much simpler. 12105 Fix calculation of probabilities that happen in tree expansion. 12106 * tree-switch-conversion.h (struct cluster): Add 12107 is_single_value_p. 12108 (struct simple_cluster): Likewise. 12109 (struct case_tree_node): Add new function has_child. 12110 (do_jump_if_equal): New. 12111 121122018-08-27 Martin Liska <mliska@suse.cz> 12113 12114 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests): 12115 Add new argument to bit_test_cluster constructor. 12116 (bit_test_cluster::emit): Set bits really number of values 12117 handlel by a test. 12118 (bit_test_cluster::hoist_edge_and_branch_if_true): Add 12119 probability argument. 12120 * tree-switch-conversion.h (struct bit_test_cluster): 12121 Add m_handles_entire_switch. 12122 121232018-08-27 Martin Liska <mliska@suse.cz> 12124 12125 PR tree-optimization/86702 12126 * tree-switch-conversion.c (jump_table_cluster::emit): 12127 Make probabilities even for values in jump table 12128 according to number of cases handled. 12129 (switch_decision_tree::compute_cases_per_edge): Pass 12130 argument to reset_out_edges_aux function. 12131 (switch_decision_tree::analyze_switch_statement): Likewise. 12132 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux): 12133 Make it static. 12134 121352018-08-27 Martin Liska <mliska@suse.cz> 12136 12137 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass 12138 cfun argument explicitly. 12139 * gimple-pretty-print.c (dump_gimple_switch): Likewise. 12140 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new 12141 function gimple_switch_default_bb. 12142 (convert_switch_statements): 12143 (expand_builtins): 12144 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): 12145 * stmt.c (label_to_block_fn): Use label_to_block and pass 12146 cfun argument explicitly and use gimple_switch_label_bb. 12147 (expand_case): Likewise. 12148 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass 12149 cfun argument explicitly. Likewise. 12150 (make_edges_bb): Likewise. 12151 (make_cond_expr_edges): Likewise. 12152 (get_cases_for_edge): Likewise. 12153 (make_gimple_switch_edges): Likewise. 12154 (label_to_block_fn): Likewise. 12155 (label_to_block): Likewise. 12156 (make_goto_expr_edges): Likewise. 12157 (make_gimple_asm_edges): Likewise. 12158 (main_block_label): Likewise. 12159 (group_case_labels_stmt): Likewise. 12160 (find_taken_edge_computed_goto): Likewise. 12161 (find_taken_edge_switch_expr): Likewise. 12162 (gimple_verify_flow_info): Likewise. 12163 (gimple_redirect_edge_and_branch): Likewise. 12164 (gimple_switch_label_bb): New function. 12165 (gimple_switch_default_bb): Likewise. 12166 (gimple_switch_edge): Likewise. 12167 (gimple_switch_default_edge): Likewise. 12168 * tree-cfg.h (label_to_block_fn): Remove and replace ... 12169 (label_to_block): ... with this. 12170 (gimple_switch_label_bb): New. 12171 (gimple_switch_default_bb): Likewise. 12172 (gimple_switch_edge): Likewise. 12173 (gimple_switch_default_edge): Likewise. 12174 * tree-cfgcleanup.c (convert_single_case_switch): Use 12175 new gimple functions and pass new argument to label_to_block. 12176 (cleanup_control_flow_bb): 12177 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass 12178 cfun argument explicitly. 12179 (make_eh_edges): Likewise. 12180 (redirect_eh_dispatch_edge): Likewise. 12181 (lower_resx): Likewise. 12182 (lower_eh_dispatch): Likewise. 12183 (maybe_remove_unreachable_handlers): Likewise. 12184 (unsplit_eh): Likewise. 12185 (cleanup_empty_eh): Likewise. 12186 (verify_eh_edges): Likewise. 12187 (verify_eh_dispatch_edge): Likewise. 12188 * tree-ssa-dom.c (record_edge_info): Likewise. 12189 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise. 12190 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise. 12191 (thread_through_normal_block): Likewise. 12192 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise. 12193 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): 12194 * tree-switch-conversion.c (switch_conversion::collect): Use new 12195 gimple functions. 12196 (switch_conversion::check_final_bb): Likewise. 12197 (switch_conversion::gather_default_values): Pass new argument 12198 to label_to_block. 12199 (switch_conversion::build_constructors): Likewise. 12200 (switch_decision_tree::compute_cases_per_edge): Use new 12201 gimple_switch_edge function. 12202 (switch_decision_tree::analyze_switch_statement): Pass new argument 12203 to label_to_block. 12204 (switch_decision_tree::try_switch_expansion): Use 12205 gimple_switch_default_edge. 12206 * tree-vrp.c (find_switch_asserts): Pass new argument 12207 to label_to_block. 12208 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise. 12209 (vr_values::simplify_switch_using_ranges): Likewise. 12210 122112018-08-27 Richard Biener <rguenther@suse.de> 12212 12213 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare. 12214 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function. 12215 12216 * tree-ssa-sccvn.h (struct vn_pval): New structure. 12217 (struct vn_nary_op_s): Add unwind_to member. Add 12218 predicated_values flag and put result into a union together 12219 with a linked list of vn_pval. 12220 (struct vn_ssa_aux): Add name member to make maintaining 12221 a map of SSA name to vn_ssa_aux possible. Remove no longer 12222 needed info, dfsnum, low, visited, on_sccstack, use_processed 12223 and range_info_anti_range_p members. 12224 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove. 12225 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn): 12226 New functions. 12227 (vn_valueize): New global. 12228 (vn_context_bb): Likewise. 12229 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE, 12230 VN_INFO_PTR_INFO): Remove. 12231 * tree-ssa-sccvn.c: ... (rewrite) 12232 (pass_fre::execute): For -O2+ initialize loops and run 12233 RPO VN in optimistic mode (iterating). For -O1 and -Og 12234 run RPO VN in non-optimistic mode. 12235 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove. 12236 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add. 12237 * doc/invoke.texi (sccvn-max-scc-size): Remove. 12238 (rpo-vn-max-loop-depth): Document. 12239 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking 12240 when valuezing the VUSE signals we walked out of the region. 12241 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values. 12242 (phi_translate): Set VN context block to use for availability 12243 lookup. 12244 (compute_avail): Likewise. 12245 (pre_valueize): New function. 12246 (pass_pre::execute): Adjust to the RPO VN API. 12247 12248 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h. 12249 (propagate_constants_for_unrolling): Remove. 12250 (tree_unroll_loops_completely): Perform value-numbering 12251 on the unrolled bodies loop parent. 12252 122532018-08-27 Richard Biener <rguenther@suse.de> 12254 12255 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder 12256 for partial antic compute. 12257 122582018-08-27 Jakub Jelinek <jakub@redhat.com> 12259 12260 PR rtl-optimization/87065 12261 * combine.c (simplify_if_then_else): Formatting fix. 12262 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P 12263 check. 12264 (known_cond): Don't return const_true_rtx for vector modes. Use 12265 CONST0_RTX instead of const0_rtx. Formatting fixes. 12266 122672018-08-27 Martin Liska <mliska@suse.cz> 12268 12269 PR gcov-profile/87069 12270 * gcov.c (process_file): Record files already processed 12271 and warn about a file being processed multiple times. 12272 122732018-08-27 Martin Liska <mliska@suse.cz> 12274 12275 PR driver/83193 12276 * config/aarch64/aarch64.c (aarch64_override_options_internal): 12277 Set default values for x_aarch64_*_string strings. 12278 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}== 12279 prefix. For -mabi do not print '=ABI' in help and use 12280 <option_value> format for -msve-vector-bits and -moverride 12281 options. 12282 122832018-08-26 Jeff Law <law@redhat.com> 12284 12285 * config/mips/frame-header-opt.c: Include "backend.h" rather than 12286 "cfg.h" 12287 122882018-08-26 Marek Polacek <polacek@redhat.com> 12289 12290 PR c++/87029, Implement -Wredundant-move. 12291 * doc/invoke.texi: Document -Wredundant-move. 12292 122932018-08-25 Martin Sebor <msebor@redhat.com> 12294 12295 PR tree-optimization/87059 12296 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand 12297 to the same type as the other. 12298 * fold-const.c (fold_binary_loc): Assert expectation. 12299 123002018-08-25 Iain Sandoe <iain@sandoe.co.uk> 12301 12302 * config/darwin.c (machopic_legitimize_pic_address): Clean up 12303 extraneous parentheses, dead code section and formatting. 12304 123052018-08-24 David Malcolm <dmalcolm@redhat.com> 12306 12307 PR c++/87091 12308 * diagnostic-show-locus.c (layout::layout): Ensure the margin is 12309 wide enough for jumps in the line-numbering to be visible. 12310 (layout::print_gap_in_line_numbering): New member function. 12311 (layout::calculate_line_spans): When using line numbering, merge 12312 line spans that are only 1 line apart. 12313 (diagnostic_show_locus): When printing line numbers, show gaps in 12314 line numbering directly, rather than printing headers. 12315 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of 12316 line-numbering with multiple line spans. 12317 (selftest::test_fixit_insert_containing_newline_2): Add test of 12318 line-numbering, in which the spans are close enough to be merged. 12319 123202018-08-24 Aldy Hernandez <aldyh@redhat.com> 12321 12322 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range 12323 to range_includes_zero_p. Do not special case VR_ANTI_RANGE. 12324 * tree-vrp.c (range_is_nonnull): Remove. 12325 (range_includes_zero_p): Accept value_range instead of min/max. 12326 (extract_range_from_binary_expr_1): Do not early bail on 12327 POINTER_PLUS_EXPR. 12328 Use range_includes_zero_p instead of range_is_nonnull. 12329 (extract_range_from_unary_expr): Use range_includes_zero_p instead 12330 of range_is_nonnull. 12331 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not 12332 special case VR_ANTI_RANGE. 12333 (vrp_finalize): Same. 12334 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument 12335 instead of min/max. 12336 (range_is_nonnull): Remove. 12337 * vr-values.c (vrp_stmt_computes_nonzero): Use 12338 range_includes_zero_p instead of range_is_nonnull. 12339 (extract_range_basic): Pass value_range to range_includes_zero_p 12340 instead of range_is_nonnull. 12341 123422018-08-24 Uros Bizjak <ubizjak@gmail.com> 12343 12344 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes. 12345 * emit-rtl.h (rtl_data): Remove return_bnd. 12346 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P. 12347 * function.c (diddle_return_value): Do not handle crtl->return_bnd. 12348 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS. 12349 (POINTER_BOUNDS_MODE): Remove definition. 12350 (make_pointer_bounds_mode): Remove. 12351 (get_mode_class): Do not handle MODE_POINTER_BOUNDS. 12352 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition. 12353 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS. 12354 * mode-classes.def: Do not define MODE_POINTER_BOUNDS. 12355 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS. 12356 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE. 12357 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS. 12358 12359 * config/i386/i386-modes.def (BND32, BND64): Remove. 12360 * config/i386/i386.c (dbx_register_map): Remove bound registers. 12361 (dbx64_register_map): Ditto. 12362 (svr4_dbx_register_map): Ditto. 12363 (indirect_thunk_bnd_needed): Remove. 12364 (indirect_thunks_bnd_used): Ditto. 12365 (indirect_return_bnd_needed): Ditto. 12366 (indirect_return_via_cx_bnd): Ditto. 12367 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd. 12368 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd. 12369 (output_indirect_thunk): Ditto. Remove need_prefix argument. 12370 (output_indirect_thunk_function): Remove handling of 12371 indirect_return_bnd_needed, indirect_return_via_cx_bnd, 12372 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables. 12373 (ix86_save_reg): Remove handling of crtl->return_bnd. 12374 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P. 12375 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR 12376 and UNSPEC_BNDLX_ADDR. 12377 (ix86_output_indirect_branch_via_reg): Remove handling of 12378 indirect_thunk_prefix_bnd. 12379 (ix86_output_indirect_branch_via_push): Ditto. 12380 (ix86_output_function_return): Ditto. 12381 (ix86_output_indirect_function_return): Ditto. 12382 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX. 12383 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers. 12384 (CALL_USED_REGISTERS): Ditto. 12385 (REG_ALLOC_ORDER): Update for removal of bound registers. 12386 (HI_REGISTER_NAMES): Ditto. 12387 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX) 12388 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU) 12389 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove. 12390 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove 12391 (FIRST_PSEUDO_REG): Update. 12392 (BND): Remove mode iterator. 12393 * config/i386/predicates.md (bnd_mem_operator): Remove. 12394 123952018-08-24 Richard Sandiford <richard.sandiford@arm.com> 12396 12397 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length 12398 vectors. 12399 124002018-08-24 Richard Sandiford <richard.sandiford@arm.com> 12401 12402 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out 12403 the case in which the permute needs only a single element and 12404 repeats for every vector of the result. Extend that case to 12405 handle variable-length vectors. 12406 * tree-vect-stmts.c (vectorizable_load): Update accordingly. 12407 124082018-08-24 H.J. Lu <hongjiu.lu@intel.com> 12409 12410 PR debug/79342 12411 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string 12412 on DW_MACINFO_start_file for -gsplit-dwarf -g3. 12413 124142018-08-24 Richard Biener <rguenther@suse.de> 12415 12416 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and 12417 bb_flags_allocated members. 12418 (auto_flag): New RAII class for allocating flags. 12419 (auto_edge_flag): New RAII class for allocating edge flags. 12420 (auto_bb_flag): New RAII class for allocating bb flags. 12421 * cfgloop.c (verify_loop_structure): Allocate temporary edge 12422 flag dynamically. 12423 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap 12424 in favor of temporarily allocated BB flag. 12425 * hsa-brig.c: Re-order includes. 12426 * hsa-dump.c: Likewise. 12427 * hsa-regalloc.c: Likewise. 12428 * print-rtl.c: Likewise. 12429 * profile-count.c: Likewise. 12430 124312018-08-24 Segher Boessenkool <segher@kernel.crashing.org> 12432 12433 PR target/86989 12434 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is 12435 the TOC register. 12436 124372018-08-24 Aldy Hernandez <aldyh@redhat.com> 12438 12439 PR 87073/bootstrap 12440 * wide-int-range.cc (wide_int_range_div): Do not ignore result 12441 from wide_int_range_multiplicative_op. 12442 124432018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org> 12444 12445 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo 12446 "permutaion". 12447 124482018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br> 12449 12450 * genmatch.c (parser::parse_operation): Fix typo 'exapnded' 12451 to 'expanded'. 12452 124532018-08-23 Alexander Monakov <amonakov@ispras.ru> 12454 12455 * tree-scalar-evolution.c (final_value_replacement_loop): Dump 12456 full GENERIC expression used for replacement. 12457 124582018-08-23 Aldy Hernandez <aldyh@redhat.com> 12459 12460 * tree-vrp.c (abs_extent_range): Remove. 12461 (extract_range_into_wide_ints): Pass wide ints by reference. 12462 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code. 12463 Pass wide ints by reference in all calls to 12464 extract_range_into_wide_ints. 12465 * wide-int-range.cc (wide_int_range_div): New. 12466 * wide-int-range.h (wide_int_range_div): New. 12467 (wide_int_range_includes_zero_p): New. 12468 (wide_int_range_zero_p): New. 12469 124702018-08-23 Matthew Malcomson <matthew.malcomson@arm.com> 12471 12472 * config/aarch64/aarch64.md (arches): New enum. 12473 (arch): New enum attr. 12474 (arch_enabled): New attr. 12475 (enabled): Now uses arch_enabled only. 12476 (simd, sve, fp16): Removed attribute. 12477 (fp): Attr now defined in terms of 'arch'. 12478 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64, 12479 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2, 12480 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3, 12481 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd' 12482 attributes into 'arch'. 12483 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64, 12484 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2, 12485 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3, 12486 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3, 12487 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of 12488 'simd' attribute into 'arch'. 12489 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>, 12490 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>): 12491 Convert use of 'fp' attribute to 'arch'. 12492 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>, 12493 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd' 12494 into 'arch'. 12495 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>): 12496 (different modes) Merge 'fp' and 'simd' into 'arch'. 12497 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and 12498 'simd' into 'arch'. 12499 125002018-08-23 Segher Boessenkool <segher@kernel.crashing.org> 12501 12502 PR rtl-optimization/87026 12503 * expmed.c (canonicalize_comparison): If we can no longer create 12504 pseudoregisters, don't. 12505 125062018-08-23 Richard Earnshaw <rearnsha@arm.com> 12507 12508 PR target/86951 12509 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New 12510 prototype. 12511 * config/arm/arm.c (speculation_barrier_libfunc): New static 12512 variable. 12513 (arm_init_libfuncs): Initialize it. 12514 (arm_emit_speculation_barrier): New function. 12515 * config/arm/arm.md (speculation_barrier): Call 12516 arm_emit_speculation_barrier for architectures that do not have 12517 DSB or ISB. 12518 (speculation_barrier_insn): Only match on Armv7 or later. 12519 125202018-08-23 Richard Biener <rguenther@suse.de> 12521 12522 PR middle-end/87024 12523 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len 12524 calls. 12525 125262018-08-23 Richard Sandiford <richard.sandiford@arm.com> 12527 12528 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling 12529 of single-vector TBLs. 12530 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only 12531 one input is given. 12532 125332018-08-23 Richard Sandiford <richard.sandiford@arm.com> 12534 12535 PR target/85910 12536 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix 12537 aarch64_evpc_tbl guard. 12538 125392018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de> 12540 12541 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined 12542 behaviour. 12543 125442018-08-22 Martin Sebor <msebor@redhat.com> 12545 12546 PR middle-end/87052 12547 * tree-pretty-print.c (pretty_print_string): Add argument. 12548 (dump_generic_node): Call to pretty_print_string with string size. 12549 125502018-08-22 Segher Boessenkool <segher@kernel.crashing.org> 12551 12552 PR rtl-optimization/86771 12553 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL 12554 of two SETs into those two SETs, one to be placed at i2, if that SETs 12555 destination is modified between i2 and i3. 12556 125572018-08-22 Richard Sandiford <richard.sandiford@arm.com> 12558 12559 PR tree-optimization/86725 12560 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New 12561 function. 12562 (vect_analyze_scalar_cycles_1): Check it. 12563 125642018-08-22 Richard Sandiford <richard.sandiford@arm.com> 12565 12566 PR tree-optimization/86725 12567 * tree-vect-loop.c (vect_is_simple_reduction): When treating 12568 an outer loop phi as a double reduction, make sure that the 12569 single user of the phi result is an inner loop phi. 12570 125712018-08-22 Richard Sandiford <richard.sandiford@arm.com> 12572 12573 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert 12574 grouped stores with gaps to a strided group. 12575 125762018-08-22 Richard Sandiford <richard.sandiford@arm.com> 12577 12578 * tree-vect-stmts.c (get_group_load_store_type) 12579 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the 12580 first statement in a group. 12581 125822018-08-22 Iain Sandoe <iain@sandoe.co.uk> 12583 12584 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with 12585 the sequence used in gcc/gcc.c. 12586 125872018-08-22 Iain Sandoe <iain@sandoe.co.uk> 12588 12589 PR other/704 12590 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not 12591 building it. 12592 125932018-08-22 Iain Sandoe <iain@sandoe.co.uk> 12594 12595 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the 12596 Darwin10-specific unwinder-shim. 12597 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove. 12598 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC): 12599 New to cater for Darwin10 Rosetta. 12600 126012018-08-22 Iain Sandoe <iain@sandoe.co.uk> 12602 12603 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O 12604 specific test before the one for HAVE_AS_GOTOFF_IN_DATA. 12605 126062018-08-22 Iain Sandoe <iain@sandoe.co.uk> 12607 12608 PR bootstrap/81033 12609 PR target/81733 12610 PR target/52795 12611 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New. 12612 (dwarf2out_switch_text_section): Generate a local label for the 12613 second function sub-section and apply it as the second FDE start 12614 label. 12615 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the 12616 second sub-section start. 12617 126182018-08-22 Richard Biener <rguenther@suse.de> 12619 12620 PR tree-optimization/86988 12621 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs. 12622 126232018-08-22 Richard Biener <rguenther@suse.de> 12624 12625 PR tree-optimization/86945 12626 * tree-cfg.c (generate_range_test): Use unsigned arithmetic. 12627 126282018-08-22 Alexandre Oliva <oliva@adacore.com> 12629 12630 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add 12631 a comment about how uses of r2 for .sdata2 come about. 12632 126332018-08-22 Alexandre Oliva <aoliva@redhat.com> 12634 12635 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto. 12636 126372018-08-21 Marek Polacek <polacek@redhat.com> 12638 12639 PR c++/86981, Implement -Wpessimizing-move. 12640 * doc/invoke.texi: Document -Wpessimizing-move. 12641 126422018-08-21 Jan Hubicka <jh@suse.cz> 12643 12644 * tree.c (find_decls_types_r): Do not check for redundant typedefs. 12645 * tree.h (is_redundant_typedef): Remove. 12646 * dwarf2out.c (is_redundant_typedef): Turn into static function. 12647 126482018-08-21 Jan Hubicka <jh@suse.cz> 12649 12650 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT 12651 when possible. 12652 126532018-08-21 Tamar Christina <tamar.christina@arm.com> 12654 12655 * expmed.c (extract_low_bits): Reject invalid subregs early. 12656 126572018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de> 12658 12659 PR middle-end/86121 12660 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined 12661 behaviour. 12662 126632018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk> 12664 12665 * config/vxworks.h: Guard vxworks_asm_out_constructor and 12666 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT 12667 * config/vxworks.c: Likewise. 12668 126692018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk> 12670 12671 * config/vxworks.c: Set targetm.have_ctors_dtors 12672 if HAVE_INITFINI_ARRAY_SUPPORT. 12673 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY 12674 if HAVE_INITFINI_ARRAY_SUPPORT. 12675 126762018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk> 12677 12678 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to 12679 default search path for VxWorks < 7. 12680 126812018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de> 12682 12683 * gimple-ssa-sprintf.c (decl_constant_value): Remove. 12684 (get_format_string): Refer to c_getstr. 12685 126862018-08-21 Tom de Vries <tdevries@suse.de> 12687 12688 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini) 12689 (debuginfo_start, debuginfo_stop, debuginfo_early_start) 12690 (debuginfo_early_stop): Declare. 12691 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini) 12692 (debuginfo_start, debuginfo_stop, debuginfo_early_start) 12693 (debuginfo_early_stop): New function. 12694 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start 12695 and debuginfo_early_stop. 12696 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf. 12697 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop. 12698 (general_init): Call debuginfo_early_init. 12699 (finalize): Call debuginfo_fini. 12700 (do_compile): Call debuginfo_init. 12701 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and 12702 -fdump-early-debug. 12703 (@item -fdump-debug, @item -fdump-earlydebug): Add. 12704 127052018-08-21 Tom de Vries <tdevries@suse.de> 12706 12707 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle 12708 flag_dump_noaddr and flag_dump_unnumbered. 12709 127102018-08-21 Aldy Hernandez <aldyh@redhat.com> 12711 12712 * wide-int-range.cc (wide_int_range_abs): New. 12713 (wide_int_range_order_set): Rename from wide_int_range_min_max. 12714 * wide-int-range.h (wide_int_range_abs): New. 12715 (wide_int_range_min_max): New. 12716 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR 12717 case to call wide_int_range_abs. 12718 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max. 12719 (extract_range_from_abs_expr): Delete. 12720 127212018-08-20 Michael Meissner <meissner@linux.ibm.com> 12722 12723 PR target/87033 12724 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints 12725 from 'Y' to 'YZ' to enable the LWAX instruction to be generated 12726 for indexed loads. 12727 127282018-08-20 Nathan Sidwell <nathan@acm.org> 12729 Jeff Law <law@redhat.com> 12730 12731 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p. 12732 * config/spu/spu-c.c (spu_macro_to_expand): Likewise. 12733 127342018-08-20 David Malcolm <dmalcolm@redhat.com> 12735 12736 PR other/84889 12737 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance. 12738 (decl_attributes): Likewise. 12739 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group 12740 instance. 12741 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise. 12742 * diagnostic-core.h (class auto_diagnostic_group): New class. 12743 * diagnostic.c (diagnostic_initialize): Initialize the new fields. 12744 (diagnostic_report_diagnostic): Handle the first diagnostics within 12745 a group. 12746 (emit_diagnostic): Add auto_diagnostic_group instance. 12747 (inform): Likewise. 12748 (inform_n): Likewise. 12749 (warning): Likewise. 12750 (warning_at): Likewise. 12751 (warning_n): Likewise. 12752 (pedwarn): Likewise. 12753 (permerror): Likewise. 12754 (error): Likewise. 12755 (error_n): Likewise. 12756 (error_at): Likewise. 12757 (sorry): Likewise. 12758 (fatal_error): Likewise. 12759 (internal_error): Likewise. 12760 (internal_error_no_backtrace): Likewise. 12761 (auto_diagnostic_group::auto_diagnostic_group): New ctor. 12762 (auto_diagnostic_group::~auto_diagnostic_group): New dtor. 12763 * diagnostic.h (struct diagnostic_context): Add fields 12764 "diagnostic_group_nesting_depth", 12765 "diagnostic_group_emission_count", "begin_group_cb", 12766 "end_group_cb". 12767 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): 12768 Add auto_diagnostic_group instance(s). 12769 (find_explicit_erroneous_behavior): Likewise. 12770 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise. 12771 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise. 12772 * gimplify.c (warn_implicit_fallthrough_r): Likewise. 12773 (gimplify_va_arg_expr): Likewise. 12774 * hsa-gen.c (HSA_SORRY_ATV): Likewise. 12775 (HSA_SORRY_AT): Likewise. 12776 * ipa-devirt.c (compare_virtual_tables): Likewise. 12777 (warn_odr): Likewise. 12778 * multiple_target.c (expand_target_clones): Likewise. 12779 * opts-common.c (cmdline_handle_error): Likewise. 12780 * reginfo.c (globalize_reg): Likewise. 12781 * substring-locations.c (format_warning_n_va): Likewise. 12782 * tree-inline.c (expand_call_inline): Likewise. 12783 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise. 12784 * tree-ssa-loop-niter.c 12785 (do_warn_aggressive_loop_optimizations): Likewise. 12786 * tree-ssa-uninit.c (warn_uninit): Likewise. 12787 * tree.c (warn_deprecated_use): Likewise. 12788 127892018-08-20 H.J. Lu <hongjiu.lu@intel.com> 12790 12791 PR target/87014 12792 * config/i386/i386.md (eh_return): Always update EH return 12793 address in word_mode. 12794 127952018-08-20 Chung-Lin Tang <cltang@codesourcery.com> 12796 12797 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when 12798 TARGET_SPLIT_COMPLEX_ARG is defined. 12799 128002018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de> 12801 12802 * expr.c (store_field): Change gcc_assert to gcc_checking_assert. 12803 128042018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de> 12805 12806 PR target/86984 12807 * expr.c (expand_assignment): Assert that bitpos is positive. 12808 (store_field): Likewise 12809 (expand_expr_real_1): Make sure that bitpos is positive. 12810 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed 12811 integer overflow. 12812 128132018-08-20 Nathan Sidwell <nathan@acm.org> 12814 12815 * Makefile.in (CPP_ID_DATA_H): Delete. 12816 (CPP_INTERNAL_H): Don't add it. 12817 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H. 12818 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h 12819 128202018-08-20 Richard Biener <rguenther@suse.de> 12821 12822 PR tree-optimization/78655 12823 * tree-vrp.c (extract_range_from_binary_expr_1): Make 12824 pointer + offset nonnull if either operand is nonnull work. 12825 128262018-08-20 Tom de Vries <tdevries@suse.de> 12827 12828 * dwarf2out.c (add_scalar_info): Don't add reference to existing die 12829 unless the referenced die describes the added property using 12830 DW_AT_location or DW_AT_const_value. Fall back to exprloc case. 12831 Otherwise, add a DW_AT_location to the referenced die. 12832 128332018-08-19 Uros Bizjak <ubizjak@gmail.com> 12834 12835 PR target/86994 12836 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for 12837 register_operand when calling ix86_set_reg_reg_cost. 12838 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]: 12839 Set *total to 0 for operands that satisfy x86_64_immediate_operand 12840 predicate and to 1 otherwise. 12841 128422018-08-18 Iain Sandoe <iain@sandoe.co.uk> 12843 12844 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set, 12845 emit a diagnostic that it is not supported and reset the option. 12846 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not 12847 supported and consume the option. (ASM_FINAL_SPEC): New. 12848 128492018-08-17 Segher Boessenkool <segher@kernel.crashing.org> 12850 12851 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of 12852 a sentence. 12853 128542018-08-17 Sandra Loosemore <sandra@codesourcery.com> 12855 12856 C-SKY port: Documentation 12857 12858 * doc/extend.texi (C-SKY Function Attributes): New section. 12859 * doc/invoke.texi (Option Summary): Add C-SKY options. 12860 (C-SKY Options): New section. 12861 * doc/md.texi (Machine Constraints): Document C-SKY constraints. 12862 128632018-08-17 Jojo <jijie_rong@c-sky.com> 12864 Huibin Wang <huibin_wang@c-sky.com> 12865 Sandra Loosemore <sandra@codesourcery.com> 12866 Chung-Lin Tang <cltang@codesourcery.com> 12867 12868 C-SKY port: Backend implementation 12869 12870 * config/csky/*: New. 12871 * common/config/csky/*: New. 12872 128732018-08-17 Jojo <jijie_rong@c-sky.com> 12874 Huibin Wang <huibin_wang@c-sky.com> 12875 Sandra Loosemore <sandra@codesourcery.com> 12876 Chung-Lin Tang <cltang@codesourcery.com> 12877 Andrew Jenner <andrew@codesourcery.com> 12878 12879 C-SKY port: Configury 12880 12881 * config.gcc (csky-*-*): New. 12882 * configure.ac: Add csky to targets for dwarf2 debug_line support. 12883 * configure: Regenerated. 12884 128852018-08-17 David Malcolm <dmalcolm@redhat.com> 12886 12887 * dump-context.h: Include "dumpfile.h". 12888 (dump_context::dump_printf_va): Convert final param from va_list 12889 to va_list *. Convert from ATTRIBUTE_PRINTF to 12890 ATTRIBUTE_GCC_DUMP_PRINTF. 12891 (dump_context::dump_printf_loc_va): Likewise. 12892 * dumpfile.c: Include "stringpool.h". 12893 (make_item_for_dump_printf_va): Delete. 12894 (make_item_for_dump_printf): Delete. 12895 (class dump_pretty_printer): New class. 12896 (dump_pretty_printer::dump_pretty_printer): New ctor. 12897 (dump_pretty_printer::emit_items): New member function. 12898 (dump_pretty_printer::emit_any_pending_textual_chunks): New member 12899 function. 12900 (dump_pretty_printer::emit_item): New member function. 12901 (dump_pretty_printer::stash_item): New member function. 12902 (dump_pretty_printer::format_decoder_cb): New member function. 12903 (dump_pretty_printer::decode_format): New member function. 12904 (dump_context::dump_printf_va): Reimplement in terms of 12905 dump_pretty_printer. 12906 (dump_context::dump_printf_loc_va): Convert final param from va_list 12907 to va_list *. 12908 (dump_context::begin_scope): Reimplement call to 12909 make_item_for_dump_printf. 12910 (dump_printf): Update for change to dump_printf_va. 12911 (dump_printf_loc): Likewise. 12912 (selftest::test_capture_of_dump_calls): Convert "stmt" from 12913 greturn * to gimple *. Add a test_decl. Add tests of dump_printf 12914 with %T, %E, and %G. 12915 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro. 12916 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with 12917 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3). 12918 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with 12919 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0). 12920 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert 12921 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count" 12922 within a dump_printf_loc call to "%wu". 12923 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls, 12924 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a 12925 missing space after "=". 12926 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf 12927 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd". 12928 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call, 12929 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu". 12930 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove 12931 duplicate "vectorized" from message. 12932 129332018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com> 12934 12935 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear 12936 polyNxK_t element's TYPE_STRING_FLAG. 12937 129382018-08-17 Segher Boessenkool <segher@kernel.crashing.org> 12939 12940 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns 12941 (they were unnamed before). Fix comments. 12942 129432018-08-17 Nathan Sidwell <nathan@acm.org> 12944 12945 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h". 12946 129472018-08-17 Richard Biener <rguenther@suse.de> 12948 12949 PR tree-optimization/86841 12950 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi. 12951 129522018-08-17 Martin Liska <mliska@suse.cz> 12953 12954 * common.opt: Remove Warn, Init and Report for options with 12955 Ignore/Deprecated flag. Warning is done automatically for 12956 Deprecated flags. 12957 * config/i386/i386.opt: Likewise. 12958 * config/ia64/ia64.opt: Likewise. 12959 * config/rs6000/rs6000.opt: Likewise. 12960 * cppbuiltin.c (define_builtin_macros_for_compilation_flags): 12961 Remove usage of flag_check_pointer_bounds. 12962 * lto-wrapper.c (merge_and_complain): Do not handle 12963 OPT_fcheck_pointer_bounds. 12964 (append_compiler_options): Likewise. 12965 * opt-functions.awk: Do not handle Deprecated. 12966 * optc-gen.awk: Check that Var, Report and Init are not 12967 used for an option with Ignore/Deprecated flag. 12968 * opts-common.c (decode_cmdline_option): Do not report 12969 CL_ERR_DEPRECATED. 12970 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated 12971 options. 12972 * opts.h (struct cl_option): Remove cl_deprecated flag. 12973 (CL_ERR_DEPRECATED): Remove error enum value. 12974 129752018-08-17 Richard Biener <rguenther@suse.de> 12976 12977 PR middle-end/86505 12978 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len () 12979 across a va-arg-pack using call adjust its return value accordingly. 12980 129812018-08-16 Martin Sebor <msebor@redhat.com> 12982 12983 PR tree-optimization/86853 12984 * gimple-ssa-sprintf.c (struct format_result): Rename member. 12985 (struct fmtresult): Add member and initialize it in ctors. 12986 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL. 12987 (format_string): Handle %S the same as %ls. Set MAYFAIL. 12988 (format_directive): Set POSUNDER4K when MAYFAIL is set. 12989 (parse_directive): Handle %C same as %c. 12990 (sprintf_dom_walker::compute_format_length): Adjust. 12991 (is_call_safe): Adjust. 12992 129932018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de> 12994 12995 * builtins.c (c_strlen): Add new parameter eltsize. Use it 12996 for determining how to count the elements. 12997 * builtins.h (c_strlen): Adjust prototype. 12998 * expr.c (string_constant): Add new parameter mem_size. 12999 Set *mem_size appropriately. 13000 * expr.h (string_constant): Adjust protoype. 13001 * gimple-fold.c (get_range_strlen): Add new parameter eltsize. 13002 * gimple-fold.h (get_range_strlen): Adjust prototype. 13003 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize. 13004 (format_string): Call get_string_length with eltsize. 13005 130062018-08-16 David Malcolm <dmalcolm@redhat.com> 13007 13008 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string 13009 to emit the span, rather than setting it as the prefix. 13010 130112018-08-16 David Malcolm <dmalcolm@redhat.com> 13012 13013 * diagnostic-show-locus.c (layout::start_annotation_line): Add 13014 "margin_char" parameter, defaulting to space. Use it in place 13015 of pp_space for the initial part of the margin. 13016 (layout::print_leading_fixits): Use '+' when filling the margin 13017 of line-insertion fix-it hints. 13018 130192018-08-16 Segher Boessenkool <segher@kernel.crashing.org> 13020 13021 * config/rs6000/rs6000.md (two unnamed define_insn and define_split): 13022 Delete. 13023 130242018-08-16 Segher Boessenkool <segher@kernel.crashing.org> 13025 13026 * config/rs6000/altivec.md: Don't set length attribute to the default 13027 value. 13028 * config/rs6000/darwin.md: Ditto. 13029 * config/rs6000/dfp.md: Ditto. 13030 * config/rs6000/htm.md: Ditto. 13031 * config/rs6000/rs6000.md: Ditto. 13032 * config/rs6000/sync.md: Ditto. 13033 * config/rs6000/vsx.md: Ditto. 13034 130352018-08-16 Segher Boessenkool <segher@kernel.crashing.org> 13036 13037 * config/rs6000/altivec.md: Don't set length attribute to the default 13038 value, for branch instructions. 13039 * config/rs6000/darwin.md: Ditto. 13040 * config/rs6000/rs6000.md: Ditto. 13041 130422018-08-16 Segher Boessenkool <segher@kernel.crashing.org> 13043 13044 * config/rs6000/rs6000.md (length): Always define as const_int 4. 13045 (unnamed conditional branch define_insn): Set length to 4 or 8 13046 depending on offset. 13047 (<bd>_<mode>): Similar, for alternative 0. 13048 (<bd>tf_<mode>): Ditto. 13049 130502018-08-16 Tamar Christina <tamar.christina@arm.com> 13051 13052 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe. 13053 130542018-08-16 Matthew Malcomson <matthew.malcomson@arm.com> 13055 13056 * doc/rtl.texi: Replace old RTX class names with new names. 13057 13058 130592018-08-16 Vlad Lazar <vlad.lazar@arm.com> 13060 13061 * expmed.h (canonicalize_comparison): New declaration. 13062 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function. 13063 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison. 13064 * optabs.c (prepare_cmp_insn): Likewise. 13065 * rtl.h (unsigned_condition_p): New function which checks if a 13066 comparison operator is unsigned. 13067 130682018-08-16 Nathan Sidwell <nathan@acm.org> 13069 13070 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p. 13071 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise. 13072 130732018-08-16 Tamar Christina <tamar.christina@arm.com> 13074 13075 PR target/84711 13076 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg. 13077 * config/arm/neon.md (movv4hf, movv8hf): Refactored to.. 13078 (mov<mov>): ..this and enable unconditionally. 13079 130802018-08-16 Tamar Christina <tamar.christina@arm.com> 13081 13082 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative. 13083 130842018-08-16 Sam Tebbs <sam.tebbs@arm.com> 13085 13086 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt) 13087 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace 13088 "Common" with "Target". 13089 130902018-08-15 Uros Bizjak <ubizjak@gmail.com> 13091 13092 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated. 13093 * doc/invoke.texi (mmitigate-rop): Remove. 13094 * config/i386/i386.c: Do not include "regrename.h". 13095 (ix86_rop_should_change_byte_p, reg_encoded_number) 13096 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop): 13097 Remove. 13098 (ix86_reorg): Remove call to ix86_mitigate_rop. 13099 * config/i386/i386.md (attr "modrm_class"): Remove. 13100 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor) 13101 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se) 13102 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override. 13103 131042018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com> 13105 13106 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to 13107 allow folding of mergeh() and mergel() for the float and double types. 13108 (fold_mergehl_helper): Rework to handle building a permute tree 13109 for float vectors. 13110 131112018-08-15 Uros Bizjak <ubizjak@gmail.com> 13112 13113 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode 13114 for TARGET_SSE. 13115 131162018-08-15 David Malcolm <dmalcolm@redhat.com> 13117 13118 * common.opt (fdiagnostics-show-labels): New option. 13119 * diagnostic-show-locus.c (class layout_range): Add field 13120 "m_label". 13121 (class layout): Add field "m_show_labels_p". 13122 (layout_range::layout_range): Add param "label" and use it to 13123 initialize m_label. 13124 (make_range): Pass in NULL for new "label" param of layout_range's 13125 ctor. 13126 (layout::layout): Initialize m_show_labels_p. 13127 (layout::maybe_add_location_range): Pass in loc_range->m_label 13128 when constructing layout_range instances. 13129 (struct line_label): New struct. 13130 (layout::print_any_labels): New member function. 13131 (layout::print_line): Call it if label-printing is enabled. 13132 (selftest::test_one_liner_labels): New test. 13133 (selftest::test_diagnostic_show_locus_one_liner): Call it. 13134 * diagnostic.c (diagnostic_initialize): Initialize 13135 context->show_labels_p. 13136 * diagnostic.h (struct diagnostic_context): Add field 13137 "show_labels_p". 13138 * doc/invoke.texi (Diagnostic Message Formatting Options): Add 13139 -fno-diagnostics-show-labels. 13140 * dwarf2out.c (gen_producer_string): Add 13141 OPT_fdiagnostics_show_labels to the ignored options. 13142 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label" 13143 param. 13144 (gcc_rich_location::maybe_add_expr): Likewise. 13145 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add 13146 label" param, defaulting to NULL. 13147 (gcc_rich_location::add_expr): Add "label" param. 13148 (gcc_rich_location::maybe_add_expr): Likewise. 13149 (class text_range_label): New class. 13150 (class range_label_for_type_mismatch): New class. 13151 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params 13152 of format_warning_va. 13153 (fmtwarn_n): Likewise for new params of format_warning_n_va. 13154 * lto-wrapper.c (merge_and_complain): Add 13155 OPT_fdiagnostics_show_labels to the "pick one setting" options. 13156 (append_compiler_options): Likewise to the dropped options. 13157 (append_diag_options): Likewise to the passed-on options. 13158 * opts.c (common_handle_option): Handle the new option. 13159 * selftest-diagnostic.c 13160 (test_diagnostic_context::test_diagnostic_context): Enable 13161 show_labels_p. 13162 * substring-locations.c: Include "gcc-rich-location.h". 13163 (format_warning_n_va): Add "fmt_label" and "param_label" params 13164 and use them as appropriate. 13165 (format_warning_va): Add "fmt_label" and "param_label" params, 13166 passing them on to format_warning_n_va. 13167 (format_warning_at_substring): Likewise. 13168 (format_warning_at_substring_n): Likewise. 13169 * substring-locations.h (format_warning_va): Add "fmt_label" and 13170 "param_label" params. 13171 (format_warning_n_va): Likewise. 13172 (format_warning_at_substring): Likewise. 13173 (format_warning_at_substring_n): Likewise. 13174 * toplev.c (general_init): Initialize global_dc->show_labels_p. 13175 131762018-08-15 Qing Zhao <qing.zhao@oracle.com> 13177 13178 PR testsuite/86519 13179 * builtins.c (expand_builtin_memcmp): Do not expand the call 13180 when overflow is detected. 13181 131822018-08-15 Martin Sebor <msebor@redhat.com> 13183 13184 PR tree-optimization/71625 13185 * config/aarch64/aarch64-builtins.c 13186 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG. 13187 131882018-08-15 Ilya Leoshkevich <iii@linux.ibm.com> 13189 13190 * config/s390/s390.c (s390_reorg): Remove loop. 13191 131922018-08-15 Iain Sandoe <iain@sandoe.co.uk> 13193 13194 * config/darwin.c 13195 (darwin_function_switched_text_sections): Delete. 13196 * gcc/config/darwin.h 13197 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise. 13198 131992018-08-15 Iain Sandoe <iain@sandoe.co.uk> 13200 13201 PR target/81685 13202 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION, 13203 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION, 13204 DEBUG_PUBTYPES_SECTION) update to include GNU variant. 13205 132062018-08-15 Martin Liska <mliska@suse.cz> 13207 13208 PR tree-optimization/86925 13209 * predict.c (expr_expected_value_1): When taking 13210 later predictor, assign also probability. 13211 Use fold_build2_initializer_loc in order to fold 13212 the expression in -frounding-math. 13213 132142018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io> 13215 13216 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs 13217 patterns. 13218 (expand_vec_perm_1): Try the new method. 13219 132202018-08-14 Ilya Leoshkevich <iii@linux.ibm.com> 13221 13222 PR target/86547 13223 * lra-lives.c (remove_some_program_points_and_update_live_ranges): 13224 Check whether lra_live_max_point is 0 before dividing. 13225 132262018-08-14 Martin Sebor <msebor@redhat.com> 13227 13228 PR tree-optimization/86650 13229 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message. 13230 (vrp_prop::check_mem_ref): Same. 13231 132322018-08-13 Liu Hao <lh_mouse@126.com> 13233 13234 * pretty-print.c (eat_esc_sequence): Swap the foreground and 13235 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set, 13236 and clear it thereafter, as it only works for DBCS. 13237 132382018-08-13 Liu Hao <lh_mouse@126.com> 13239 13240 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the 13241 handle returned by _get_osf_handle(). 13242 132432018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com> 13244 13245 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support 13246 for folding vec_perm. 13247 132482018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com> 13249 13250 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): 13251 Add support for gimple-folding of vec_pack() and vec_unpack() 13252 intrinsics. 13253 132542018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com> 13255 13256 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add 13257 vec_xst variants to the list. 13258 (rs6000_gimple_fold_builtin): Add support for folding unaligned 13259 vector loads and stores. 13260 132612018-08-13 David Edelsohn <dje.gcc@gmail.com> 13262 13263 * config.gcc (rs6000-ibm-aix4.x): Delete. 13264 (rs6000-ibm-aix5.1): Delete. 13265 (rs6000-ibm-aix5.2): Delete. 13266 (rs6000-ibm-aix5.3): Delete. 13267 * config/rs6000/aix43.h: Delete. 13268 * config/rs6000/aix51.h: Delete. 13269 * config/rs6000/aix52.h: Delete. 13270 * config/rs6000/t-aix43: Delete. 13271 132722018-08-13 Ilya Leoshkevich <iii@linux.ibm.com> 13273 13274 * config/s390/s390.c (s390_decompose_constant_pool_ref): 13275 New function. 13276 (s390_decompose_address): Factor out constant pool ref 13277 decomposition. 13278 132792018-08-12 Chung-Ju Wu <jasonwucj@gmail.com> 13280 13281 * config/nds32/nds32-predicates.c 13282 (nds32_can_use_bclr_p): Change return type as bool. 13283 (nds32_can_use_bset_p): Ditto. 13284 (nds32_can_use_btgl_p): Ditto. 13285 (nds32_can_use_bitci_p): Ditto. 13286 * config/nds32/nds32-protos.h 13287 (nds32_can_use_bclr_p): Change declaration. 13288 (nds32_can_use_bset_p): Ditto. 13289 (nds32_can_use_btgl_p): Ditto. 13290 (nds32_can_use_bitci_p): Ditto. 13291 132922018-08-12 Chung-Ju Wu <jasonwucj@gmail.com> 13293 13294 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue): 13295 Support -msched-prolog-epilog option. 13296 * config/nds32/nds32.opt (msched-prolog-epilog): New option. 13297 132982018-08-12 Chung-Ju Wu <jasonwucj@gmail.com> 13299 13300 * common/config/nds32/nds32-common.c 13301 (nds32_option_optimization_table): Enalbe -malways-align. 13302 133032018-08-12 Chung-Ju Wu <jasonwucj@gmail.com> 13304 13305 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in 13306 extra_headers. 13307 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle 13308 OPT_misr_secure_ case. 13309 * config/nds32/nds32-isr.c: Implementation of backward compatibility. 13310 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New. 13311 * config/nds32/nds32.c (nds32_attribute_table): Add critical and 13312 secure attribute. 13313 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL. 13314 (nds32_isr_info): New field security_level. 13315 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro. 13316 * config/nds32/nds32.md (return_internal): Consider critical attribute. 13317 * config/nds32/nds32.opt (misr-secure): New option. 13318 * config/nds32/nds32_init.inc: New file. 13319 * config/nds32/nds32_isr.h: New file. 13320 133212018-08-11 John David Anglin <danglin@gcc.gnu.org> 13322 13323 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum. 13324 Update comment for atomic instructions. 13325 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf, 13326 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1): 13327 Remove. 13328 (atomic_loaddi): Revise fence expansion to only emit fence prior to 13329 load for __ATOMIC_SEQ_CST model. 13330 (atomic_loaddi_1): Remove float register target. 13331 (atomic_storedi): Handle CONST_INT values. 13332 (atomic_storedi_1): Remove float register source. Add special case 13333 for zero value. 13334 (memory_barrier): New expander and insn. 13335 133362018-08-11 Jakub Jelinek <jakub@redhat.com> 13337 13338 PR tree-optimization/86835 13339 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting 13340 new_stmt after def_gsi, make sure to insert new_square_stmt after 13341 that stmt, not 2 stmts before it. 13342 133432018-08-10 Alexander Monakov <amonakov@ispras.ru> 13344 13345 PR target/82418 13346 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator 13347 instead of SWI48. 13348 133492018-08-10 Martin Liska <mliska@suse.cz> 13350 13351 PR target/83610 13352 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new 13353 function type. 13354 * builtins.c (expand_builtin_expect_with_probability): 13355 New function. 13356 (expand_builtin_expect_with_probability): New function. 13357 (build_builtin_expect_predicate): Add new argumnet probability 13358 for BUILT_IN_EXPECT_WITH_PROBABILITY. 13359 (fold_builtin_expect): 13360 (fold_builtin_2): 13361 (fold_builtin_3): 13362 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY): 13363 * builtins.h (fold_builtin_expect): Set new argument. 13364 * doc/extend.texi: Document __builtin_expect_with_probability. 13365 * doc/invoke.texi: Likewise. 13366 * gimple-fold.c (gimple_fold_call): Pass new argument. 13367 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle 13368 also BUILT_IN_EXPECT_WITH_PROBABILITY. 13369 * predict.c (get_predictor_value): New function. 13370 (expr_expected_value): Add new argument probability. Assume 13371 that predictor and probability are always non-null. 13372 (expr_expected_value_1): Likewise. For __builtin_expect and 13373 __builtin_expect_with_probability set probability. Handle 13374 combination in binary expressions. 13375 (tree_predict_by_opcode): Simplify code by simply calling 13376 get_predictor_value. 13377 (pass_strip_predict_hints::execute): Add handling of 13378 BUILT_IN_EXPECT_WITH_PROBABILITY. 13379 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add 13380 new predictor. 13381 * tree.h (DECL_BUILT_IN_P): New function. 13382 133832018-08-10 Martin Liska <mliska@suse.cz> 13384 13385 PR tree-optimization/85799 13386 * passes.def: Add argument for pass_strip_predict_hints. 13387 * predict.c (class pass_strip_predict_hints): Add new argument 13388 early_p. 13389 (strip_predictor_early): New function. 13390 (pass_strip_predict_hints::execute): Call the function to 13391 strip predictors. 13392 (strip_predict_hints): New function. 13393 * predict.def: Fix comment. 13394 133952018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org> 13396 13397 * Makefile.in: Clarify which tm.texi to copy over to assert the 13398 right to grant a GFDL license for all. 13399 134002018-08-09 Jeff Law <law@redhat.com> 13401 13402 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove 13403 unused variable. 13404 134052018-08-09 Andreas Schwab <schwab@linux-m68k.org> 13406 13407 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove 13408 prototype. 13409 134102018-08-09 Richard Sandiford <richard.sandiford@arm.com> 13411 13412 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop 13413 reductions for variable-length vectors. 13414 134152018-08-09 David Malcolm <dmalcolm@redhat.com> 13416 13417 PR other/84889 13418 * common.opt (fdiagnostics-show-line-numbers): New option. 13419 * diagnostic-show-locus.c (class layout): Add fields 13420 "m_show_line_numbers_p" and "m_linenum_width"; 13421 (num_digits): New function. 13422 (test_num_digits): New function. 13423 (layout::layout): Initialize new fields. Update m_x_offset 13424 logic to handle any left margin. 13425 (layout::print_source_line): Print line number when requested. 13426 (layout::start_annotation_line): New member function. 13427 (layout::print_annotation_line): Call it. 13428 (layout::print_leading_fixits): Likewise. 13429 (layout::print_trailing_fixits): Likewise. Update calls to 13430 move_to_column for new parameter. 13431 (layout::get_x_bound_for_row): Add "add_left_margin" param and use 13432 it to potentially call start_annotation_line. 13433 (layout::show_ruler): Call start_annotation_line. 13434 (selftest::test_line_numbers_multiline_range): New selftest. 13435 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits 13436 and selftest::test_line_numbers_multiline_range. 13437 * diagnostic.c (diagnostic_initialize): Initialize 13438 show_line_numbers_p. 13439 * diagnostic.h (struct diagnostic_context): Add field 13440 "show_line_numbers_p". 13441 * doc/invoke.texi (Diagnostic Message Formatting Options): Add 13442 -fno-diagnostics-show-line-numbers. 13443 * dwarf2out.c (gen_producer_string): Add 13444 OPT_fdiagnostics_show_line_numbers to the ignored options. 13445 * lto-wrapper.c (merge_and_complain): Likewise to the "pick 13446 one setting" options. 13447 (append_compiler_options): Likewise to the dropped options. 13448 (append_diag_options): Likewise to the passed-on options. 13449 * opts.c (common_handle_option): Handle the new option. 13450 * toplev.c (general_init): Set up global_dc->show_line_numbers_p. 13451 134522018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org> 13453 13454 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on 13455 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add 13456 third argument of type "const signed char" to descriptions of 13457 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq, 13458 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub, 13459 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt, 13460 __builtin_bcdsub_ov functions. 13461 134622018-08-09 Richard Sandiford <richard.sandiford@arm.com> 13463 13464 PR tree-optimization/86858 13465 * tree-vect-loop.c (vect_is_simple_reduction): Restore 13466 flow_bb_inside_loop_p calls. 13467 134682018-08-09 Richard Sandiford <richard.sandiford@arm.com> 13469 13470 PR tree-optimization/86871 13471 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs 13472 instead of gimple_assign_lhs. 13473 134742018-08-09 Richard Earnshaw <rearnsha@arm.com> 13475 13476 PR target/86887 13477 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing 13478 register constraint to operand 0. 13479 (add<mode>3_carryinC): Likewise. 13480 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise. 13481 134822018-08-09 Martin Liska <mliska@suse.cz> 13483 13484 PR c/86895 13485 * common.opt: Remove extra line. 13486 134872018-08-09 Martin Liska <mliska@suse.cz> 13488 13489 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots 13490 at the end of a line, make first letter capital and end up 13491 a sentence with a dot. 13492 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise. 13493 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise. 13494 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise. 13495 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise. 13496 (PARAM_MAX_ISL_OPERATIONS): Likewise. 13497 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise. 13498 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise. 13499 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise. 13500 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise. 13501 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise. 13502 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise. 13503 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise. 13504 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise. 13505 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise. 13506 (PARAM_TREE_REASSOC_WIDTH): Likewise. 13507 (PARAM_HSA_GEN_DEBUG_STORES): Likewise. 13508 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise. 13509 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise. 13510 135112018-08-09 Andreas Krebbel <krebbel@linux.ibm.com> 13512 13513 PR target/84332 13514 * config/s390/s390.c (s390_option_override_internal): Reduce the 13515 stack-clash-protection-probe-interval param if it would be too big 13516 for z900. 13517 135182018-08-08 Andreas Schwab <schwab@linux-m68k.org> 13519 13520 PR target/46179 13521 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define. 13522 * config/m68k/m68k.c (handle_move_double): Don't call 13523 m68k_final_prescan_insn. 13524 (m68k_adjust_decorated_operand): Renamed from 13525 m68k_final_prescan_insn, remove first and third operand and 13526 simplify. 13527 (print_operand): Call it. 13528 (print_operand_address): Call it. 13529 135302018-08-08 Nathan Sidwell <nathan@acm.org> 13531 13532 * diagnostic.c (diagnostic_report_current_module): Use 13533 linemap_included_from & linemap_included_from_linemap. 13534 135352018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org> 13536 13537 * config/aarch64/aarch64-cores.def: Add phecda core. 13538 * config/aarch64/aarch64-tune.md: Regenerate. 13539 * doc/invoke.texi: Add phecda core. 13540 135412018-08-08 Andreas Krebbel <krebbel@linux.ibm.com> 13542 13543 PR target/85295 13544 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint 13545 definitions. 13546 * config/s390/s390.md ("movti"): Add more alternatives for 13547 constant to GPR copies. 13548 135492018-08-08 Andreas Krebbel <krebbel@linux.ibm.com> 13550 13551 * config/s390/s390.c: Fix whitespace damage throughout the file. 13552 * config/s390/s390.h: Likewise. 13553 * config/s390/tpf.h: Likewise. 13554 135552018-08-08 Ilya Leoshkevich <iii@linux.ibm.com> 13556 13557 * config/s390/s390.c (s390_loadrelative_operand_p): 13558 Remove TARGET_CPU_ZARCH usages. 13559 (s390_rtx_costs): Likewise. 13560 (s390_legitimate_constant_p): Likewise. 13561 (s390_cannot_force_const_mem): Likewise. 13562 (legitimate_reload_constant_p): Likewise. 13563 (s390_preferred_reload_class): Likewise. 13564 (legitimize_pic_address): Likewise. 13565 (legitimize_tls_address): Likewise. 13566 (s390_split_branches): Removed. 13567 (s390_add_execute): Removed. 13568 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages. 13569 (s390_mainpool_start): Likewise. 13570 (s390_mainpool_finish): Likewise. 13571 (s390_mainpool_cancel): Removed. 13572 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages. 13573 (s390_chunkify_cancel): Likewise. 13574 (s390_return_addr_rtx): Likewise. 13575 (s390_register_info): Remove split_branches_pending_p uages. 13576 (s390_optimize_register_info): Likewise. 13577 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and 13578 split_branches_pending_p usages. 13579 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages. 13580 (s390_load_got): Likewise. 13581 (s390_expand_split_stack_prologue): Likewise. 13582 (output_asm_nops): Likewise. 13583 (s390_function_profiler): Likewise. 13584 (s390_emit_call): Likewise. 13585 (s390_conditional_register_usage): Likewise. 13586 (s390_optimize_prologue): Likewise. 13587 (s390_reorg): Remove TARGET_CPU_ZARCH and 13588 split_branches_pending_p usages. 13589 (s390_option_override_internal): Remove TARGET_CPU_ZARCH 13590 usages. 13591 (s390_output_indirect_thunk_function): Likewise. 13592 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed. 13593 (TARGET_CPU_ZARCH_P): Removed. 13594 (struct machine_function): Remove split_branches_pending_p. 13595 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages. 13596 135972018-08-08 Ilya Leoshkevich <iii@linux.ibm.com> 13598 13599 * common/config/s390/s390-common.c (processor_flags_table): 13600 Remove flags. 13601 * config.gcc: Remove with_arch/with_tune support. 13602 * config/s390/2064.md: Remove cpu attribute comparisons. 13603 * config/s390/driver-native.c (s390_host_detect_local_cpu): 13604 Remove MTN. 13605 * config/s390/linux.h (ASM_SPEC): 13606 Remove -march support. 13607 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): 13608 Use a table to get an arch level. 13609 * config/s390/s390-opts.h (enum processor_type): 13610 Remove enum values. 13611 * config/s390/s390.c 13612 (processor_table): Remove entries, add arch_level values. 13613 (s390_issue_rate): Remove cases. 13614 (s390_option_override): Adjust 13615 s390_option_override_internal() call. 13616 (s390_option_override_internal): Remove deprecation warning. 13617 (s390_valid_target_attribute_tree): Adjust 13618 s390_option_override_internal() call. 13619 * config/s390/s390.h (struct s390_processor): 13620 Share with s390-c.c, add arch_level field. 13621 * config/s390/s390.md: 13622 Remove occurrences in cpu attribute. 13623 * config/s390/s390.opt: Remove -march/-mtune support. 13624 * config/s390/tpf.h (ASM_SPEC): Remove -march support. 13625 * doc/invoke.texi: Remove deprecation warning. 13626 136272018-08-08 Luis Machado <luis.machado@linaro.org> 13628 13629 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static 13630 global. 13631 (qdf24xx_tunings): Set vector cost structure to 13632 qdf24xx_vector_cost. 13633 13634 * config/aarch64/aarch64.c (qdf24xx_addrcost_table) 13635 <register_sextend>: Set to 3. 13636 136372018-08-07 Richard Sandiford <richard.sandiford@arm.com> 13638 13639 PR target/86838 13640 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete. 13641 * config/aarch64/aarch64-simd.md 13642 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into... 13643 (@aarch64_frecpe<mode>): ...here and the move FRECPX to... 13644 (aarch64_frecpx<mode>): ...this new pattern. 13645 * config/aarch64/aarch64-simd-builtins.def: Remove comment 13646 about aarch64_frecp<FRECP:frecp_suffix><mode>. 13647 136482018-08-07 Martin Liska <mliska@suse.cz> 13649 13650 PR middle-end/83023 13651 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC, 13652 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW. 13653 * predict.def (PRED_MALLOC_NONNULL): New predictor. 13654 * doc/extend.texi: Document that malloc attribute adds 13655 hit to compiler. 13656 136572018-08-06 John David Anglin <danglin@gcc.gnu.org> 13658 13659 PR target/86785 13660 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): 13661 Define to speculation_safe_value_not_needed. 13662 136632018-08-06 Jeff Law <law@redhat.com> 13664 13665 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down 13666 the vr_values instance to cprop_into_stmt. 13667 (cprop_into_stmt): Pass vr_values instance down to cprop_operand. 13668 (cprop_operand): Also query EVRP to determine if OP is a constant. 13669 136702018-08-06 Nathan Sidwell <nathan@acm.org> 13671 13672 * diagnostic.c (diagnostic_report_current_module): Reroll 13673 included-at loop. Translate text. 13674 136752018-08-06 David Malcolm <dmalcolm@redhat.com> 13676 13677 * function-tests.c (selftest::test_expansion_to_rtl): Call 13678 free_after_compilation. 13679 136802018-08-06 Alan Hayward <alan.hayward@arm.com> 13681 13682 * config/aarch64/aarch64.md: Add clobber highs to tls_desc. 13683 136842018-08-06 Andreas Krebbel <krebbel@linux.ibm.com> 13685 13686 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small 13687 loops with memory block operations from getting unrolled. 13688 136892018-08-06 Ulrich Weigand <uweigand@de.ibm.com> 13690 13691 PR target/86807 13692 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): 13693 Define to speculation_safe_value_not_needed. 13694 136952018-08-06 Jeff Law <law@redhat.com> 13696 13697 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH 13698 assert. 13699 137002018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com> 13701 13702 PR target/86662 13703 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array 13704 with all enabled __intN types. 13705 13706 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test. 13707 137082018-08-06 Alan Hayward <alan.hayward@arm.com> 13709 13710 * alias.c (record_set): Check for clobber high. 13711 * cfgexpand.c (expand_gimple_stmt): Likewise. 13712 * combine-stack-adj.c (single_set_for_csa): Likewise. 13713 * combine.c (find_single_use_1): Likewise. 13714 (set_nonzero_bits_and_sign_copies): Likewise. 13715 (get_combine_src_dest): Likewise. 13716 (is_parallel_of_n_reg_sets): Likewise. 13717 (try_combine): Likewise. 13718 (record_dead_and_set_regs_1): Likewise. 13719 (reg_dead_at_p_1): Likewise. 13720 (reg_dead_at_p): Likewise. 13721 * dce.c (deletable_insn_p): Likewise. 13722 (mark_nonreg_stores_1): Likewise. 13723 (mark_nonreg_stores_2): Likewise. 13724 * df-scan.c (df_find_hard_reg_defs): Likewise. 13725 (df_uses_record): Likewise. 13726 (df_get_call_refs): Likewise. 13727 * dwarf2out.c (mem_loc_descriptor): Likewise. 13728 * haifa-sched.c (haifa_classify_rtx): Likewise. 13729 * ira-build.c (create_insn_allocnos): Likewise. 13730 * ira-costs.c (scan_one_insn): Likewise. 13731 * ira.c (equiv_init_movable_p): Likewise. 13732 (rtx_moveable_p): Likewise. 13733 (interesting_dest_for_shprep): Likewise. 13734 * jump.c (mark_jump_label_1): Likewise. 13735 * postreload-gcse.c (record_opr_changes): Likewise. 13736 * postreload.c (reload_cse_simplify): Likewise. 13737 (struct reg_use): Add source expr. 13738 (reload_combine): Check for clobber high. 13739 (reload_combine_note_use): Likewise. 13740 (reload_cse_move2add): Likewise. 13741 (move2add_note_store): Likewise. 13742 * print-rtl.c (print_pattern): Likewise. 13743 * recog.c (decode_asm_operands): Likewise. 13744 (store_data_bypass_p): Likewise. 13745 (if_test_bypass_p): Likewise. 13746 * regcprop.c (kill_clobbered_value): Likewise. 13747 (kill_set_value): Likewise. 13748 * reginfo.c (reg_scan_mark_refs): Likewise. 13749 * reload1.c (maybe_fix_stack_asms): Likewise. 13750 (eliminate_regs_1): Likewise. 13751 (elimination_effects): Likewise. 13752 (mark_not_eliminable): Likewise. 13753 (scan_paradoxical_subregs): Likewise. 13754 (forget_old_reloads_1): Likewise. 13755 * reorg.c (find_end_label): Likewise. 13756 (try_merge_delay_insns): Likewise. 13757 (redundant_insn): Likewise. 13758 (own_thread_p): Likewise. 13759 (fill_simple_delay_slots): Likewise. 13760 (fill_slots_from_thread): Likewise. 13761 (dbr_schedule): Likewise. 13762 * resource.c (update_live_status): Likewise. 13763 (mark_referenced_resources): Likewise. 13764 (mark_set_resources): Likewise. 13765 * rtl.c (copy_rtx): Likewise. 13766 * rtlanal.c (reg_referenced_p): Likewise. 13767 (single_set_2): Likewise. 13768 (noop_move_p): Likewise. 13769 (note_stores): Likewise. 13770 * sched-deps.c (sched_analyze_reg): Likewise. 13771 (sched_analyze_insn): Likewise. 13772 137732018-08-06 Alan Hayward <alan.hayward@arm.com> 13774 13775 * cse.c (invalidate_reg): New function extracted from... 13776 (invalidate): ...here. 13777 (canonicalize_insn): Check for clobber high. 13778 (invalidate_from_clobbers): invalidate clobber highs. 13779 (invalidate_from_sets_and_clobbers): Likewise. 13780 (count_reg_usage): Check for clobber high. 13781 (insn_live_p): Likewise. 13782 * cselib.c (cselib_expand_value_rtx_1):Likewise. 13783 (cselib_invalidate_regno): Check for clobber in setter. 13784 (cselib_invalidate_rtx): Pass through setter. 13785 (cselib_invalidate_rtx_note_stores): 13786 (cselib_process_insn): Check for clobber high. 13787 * cselib.h (cselib_invalidate_rtx): Add operand. 13788 137892018-08-06 Alan Hayward <alan.hayward@arm.com> 13790 13791 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high. 13792 (mark_not_eliminable): Likewise. 13793 * lra-int.h (struct lra_insn_reg): Add clobber high marker. 13794 * lra-lives.c (process_bb_lives): Check for clobber high. 13795 * lra.c (new_insn_reg): Remember clobber highs. 13796 (collect_non_operand_hard_regs): Check for clobber high. 13797 (lra_set_insn_recog_data): Likewise. 13798 (add_regs_to_insn_regno_info): Likewise. 13799 (lra_update_insn_regno_info): Likewise. 13800 138012018-08-06 Alan Hayward <alan.hayward@arm.com> 13802 13803 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations. 13804 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function. 13805 138062018-08-06 Alan Hayward <alan.hayward@arm.com> 13807 13808 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH. 13809 (copy_insn_1): Likewise. 13810 (gen_hard_reg_clobber_high): New gen function. 13811 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH. 13812 * genemit.c (gen_exp): Likewise. 13813 (gen_emit_seq): Pass through info. 13814 (gen_insn): Check for CLOBBER_HIGH. 13815 (gen_expand): Pass through info. 13816 (gen_split): Likewise. 13817 (output_add_clobbers): Likewise. 13818 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH. 13819 (remove_clobbers): Likewise. 13820 * rtl.h (gen_hard_reg_clobber_high): New declaration. 13821 138222018-08-06 Alan Hayward <alan.hayward@arm.com> 13823 13824 * doc/rtl.texi (clobber_high): Add. 13825 (parallel): Add in clobber high 13826 * rtl.c (rtl_check_failed_code3): Add function. 13827 * rtl.def (CLOBBER_HIGH): Add expression. 13828 * rtl.h (RTL_CHECKC3): Add macro. 13829 (rtl_check_failed_code3): Add declaration. 13830 (XC3EXP): Add macro. 13831 138322018-08-05 H.J. Lu <hongjiu.lu@intel.com> 13833 13834 PR target/86386 13835 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set 13836 cfun->machine->max_used_stack_alignment if needed. 13837 138382018-08-04 Martin Sebor <msebor@redhat.com> 13839 13840 PR tree-optimization/86571 13841 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of 13842 NaN output to 4. 13843 138442018-08-03 Sandra Loosemore <sandra@codesourcery.com> 13845 13846 PR target/86799 13847 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): 13848 Define. 13849 138502018-08-03 Jeff Law <law@redhat.com> 13851 13852 PR target/86795 13853 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): 13854 Define to speculation_safe_value_not_needed. 13855 138562018-08-03 David Malcolm <dmalcolm@redhat.com> 13857 13858 * doc/gcov.texi (-x): Remove duplicate "to". 13859 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls". 13860 (-Wif-not-aligned): Remove duplicate "is". 13861 (-flto): Remove duplicate "the". 13862 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove 13863 duplicate "v5.00.b". 13864 (MSP430 Options): Remove duplicate "and" from the description 13865 of "-mgprel-sec=regexp". 13866 (x86 Options): Remove duplicate copies of "vmldLog102" and 13867 vmlsLog104 from description of "-mveclibabi=type". 13868 138692018-08-03 Richard Sandiford <richard.sandiford@arm.com> 13870 13871 * internal-fn.h (first_commutative_argument): Declare. 13872 * internal-fn.c (first_commutative_argument): New function. 13873 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra 13874 restrictions for pattern statements. Use first_commutative_argument 13875 to look for commutative operands in calls to internal functions. 13876 138772018-08-03 Aldy Hernandez <aldyh@redhat.com> 13878 13879 * Makefile.in (wide-int-range.o): New. 13880 * tree-vrp.c: Move all the wide_int_* functions to... 13881 * wide-int-range.cc: ...here. 13882 * tree-vrp.h: Move all the wide_int_* prototypes to... 13883 * wide-int-range.h: ...here. 13884 138852018-08-03 Tom de Vries <tdevries@suse.de> 13886 13887 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return 13888 UI_NONE. 13889 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define. 13890 * except.c (output_function_exception_table): Do early exit if 13891 targetm_common.except_unwind_info (&global_options) == UI_NONE. 13892 138932018-08-03 Martin Liska <mliska@suse.cz> 13894 13895 * predict.c (dump_prediction): Change to 2 digits 13896 in fraction part. 13897 138982018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org> 13899 13900 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move 13901 neon_dup_q to... 13902 (falkor_am_1_gtov_gtov): ... a new insn reservation. 13903 139042018-07-19 Ilya Leoshkevich <iii@linux.ibm.com> 13905 13906 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=. 13907 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=. 13908 * dwarf2out.c (mem_loc_descriptor): Replace > with >=. 13909 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <. 13910 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <. 13911 139122018-08-02 David Malcolm <dmalcolm@redhat.com> 13913 13914 * diagnostic-show-locus.c (diagnostic_show_locus): Use 13915 pp_take_prefix when saving the existing prefix. 13916 * diagnostic.c (diagnostic_append_note): Likewise. 13917 * langhooks.c (lhd_print_error_function): Likewise. 13918 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix" 13919 param's type. Free the existing prefix. 13920 (pp_take_prefix): New function. 13921 (pretty_printer::pretty_printer): Drop the prefix parameter. 13922 Rename the length parameter to match the comment. 13923 (pretty_printer::~pretty_printer): Free the prefix. 13924 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix 13925 parameter. 13926 (struct pretty_printer): Drop the "const" from "prefix" field's 13927 type and clarify memory management. 13928 (pp_set_prefix): Drop the "const" from the 2nd param. 13929 (pp_take_prefix): New decl. 13930 139312018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com> 13932 13933 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test 13934 for word_mode_ok here instead of passing as argument. 13935 (expand_block_compare): Change select_block_compare_mode() call. 13936 (expand_strncmp_gpr_sequence): New function. 13937 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence. 13938 139392018-08-02 Jeff Law <law@redhat.com> 13940 13941 PR target/86790 13942 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): 13943 Define to speculation_safe_value_not_needed. 13944 13945 PR target/86784 13946 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): 13947 Define to speculation_safe_value_not_needed. 13948 139492018-08-02 Tom de Vries <tdevries@suse.de> 13950 13951 PR target/86660 13952 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New 13953 function. Return UI_TARGET unconditionally. 13954 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info. 13955 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'. 13956 139572018-08-02 Richard Sandiford <richard.sandiford@arm.com> 13958 13959 * genemit.c (print_overload_test): Fix typo. 13960 139612018-08-02 Richard Biener <rguenther@suse.de> 13962 13963 PR tree-optimization/86816 13964 * tree-ssa-tail-merge.c (tail_merge_valueize): New function 13965 which checks for value availability before querying it. 13966 (gvn_uses_equal): Use it. 13967 (same_succ_hash): Likewise. 13968 (gimple_equal_p): Likewise. 13969 139702018-08-02 Nick Clifton <nickc@redhat.com> 13971 13972 PR target/86813 13973 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): 13974 Define to speculation_safe_value_not_needed. 13975 13976 PR target/86810 13977 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): 13978 Define to speculation_safe_value_not_needed. 13979 13980 PR target/86810 13981 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): 13982 Define to speculation_safe_value_not_needed. 13983 13984 PR target/86803 13985 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): 13986 Define to speculation_safe_value_not_needed. 13987 13988 PR target/86797 13989 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): 13990 Define to speculation_safe_value_not_needed. 13991 13992 PR target/86791 13993 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): 13994 Define to speculation_safe_value_not_needed. 13995 13996 PR target/86789 13997 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): 13998 Define to speculation_safe_value_not_needed. 13999 14000 PR target/86787 14001 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): 14002 Define to speculation_safe_value_not_needed. 14003 14004 PR target/86782 14005 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to 14006 speculation_safe_value_not_needed. 14007 14008 PR target/86781 14009 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define 14010 to speculation_safe_value_not_needed. 14011 140122018-08-02 Richard Sandiford <richard.sandiford@arm.com> 14013 14014 * doc/md.texi: Expand the documentation of instruction names 14015 to mention port-local uses. Document '@' in pattern names. 14016 * read-md.h (overloaded_instance, overloaded_name): New structs. 14017 (mapping): Declare. 14018 (md_reader::handle_overloaded_name): New member function. 14019 (md_reader::get_overloads): Likewise. 14020 (md_reader::m_first_overload): New member variable. 14021 (md_reader::m_next_overload_ptr): Likewise. 14022 (md_reader::m_overloads_htab): Likewise. 14023 * read-md.c (md_reader::md_reader): Initialize m_first_overload, 14024 m_next_overload_ptr and m_overloads_htab. 14025 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields. 14026 (get_mode_token, get_code_token, get_int_token): New functions. 14027 (map_attr_string): Add an optional argument that passes back 14028 the associated iterator. 14029 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p): 14030 (md_reader::handle_overloaded_name, add_overload_instance): New 14031 functions. 14032 (apply_iterators): Handle '@' names. Report an error if '@' 14033 is used without iterators. 14034 (initialize_iterators): Initialize the new iterator_group fields. 14035 * genopinit.c (handle_overloaded_code_for) 14036 (handle_overloaded_gen): New functions. 14037 (main): Use them to print declarations of maybe_code_for_* and 14038 maybe_gen_* functions, and inline definitions of code_for_* and gen_*. 14039 * genemit.c (print_overload_arguments, print_overload_test) 14040 (handle_overloaded_code_for, handle_overloaded_gen): New functions. 14041 (main): Use it to print definitions of maybe_code_for_* and 14042 maybe_gen_* functions. 14043 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use 14044 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high} 14045 instead of explicit mode checks. 14046 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine. 14047 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov. 14048 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive. 14049 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive. 14050 (aarch64_expand_compare_and_swap): Likewise 14051 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse 14052 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas. 14053 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp. 14054 (aarch64_constant_pool_reload_icode): Delete. 14055 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp 14056 instead of aarch64_constant_pool_reload_icode. Use 14057 code_for_aarch64_reload_mov instead of explicit mode checks. 14058 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete. 14059 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of 14060 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type. 14061 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete. 14062 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of 14063 get_recpe_type and gen_aarch64_frecps instead of get_recps_type. 14064 (aarch64_atomic_load_op_code): Delete. 14065 (aarch64_emit_atomic_load_op): Likewise. 14066 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of 14067 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load 14068 instead of aarch64_emit_atomic_load_op. 14069 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>) 14070 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>) 14071 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high) 14072 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@' 14073 character before the pattern name. 14074 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>) 14075 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>) 14076 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>) 14077 (aarch64_frecps<mode>): Likewise. 14078 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>) 14079 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse) 14080 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>) 14081 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>) 14082 (aarch64_atomic_load<atomic_ldop><mode>): Likewise. 14083 140842018-08-02 Richard Sandiford <richard.sandiford@arm.com> 14085 14086 * config/aarch64/aarch64.c (aarch64_float_const_representable_p): 14087 Allow HFmode constants if TARGET_FP_F16INST. 14088 140892018-08-02 Jackson Woodruff <jackson.woodruff@arm.com> 14090 14091 PR target/86014 14092 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): 14093 No longer check last store for clobber of address register. 14094 140952018-08-02 Martin Liska <mliska@suse.cz> 14096 14097 PR gcov-profile/86817 14098 * gcov.c (process_all_functions): New function. 14099 (main): Call it. 14100 (process_file): Move functions processing to 14101 process_all_functions. 14102 141032018-08-02 David Malcolm <dmalcolm@redhat.com> 14104 14105 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add 14106 "const" to the "gimple *" and "rtx_insn *" parameters. 14107 * dumpfile.h (dump_user_location_t::dump_user_location_t): 14108 Likewise. 14109 (dump_location_t::dump_location_t): Likewise. 14110 141112018-08-01 Martin Sebor <msebor@redhat.com> 14112 14113 PR tree-optimization/86650 14114 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *" 14115 rather than a "gcall *". Directly pass the data of interest 14116 to percent_K_format, rather than building a temporary CALL_EXPR 14117 to hold it. 14118 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust. 14119 (gimple_fold_builtin_strncat): Adjust. 14120 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace 14121 gcall* argument with gimple*. 14122 * gimple-ssa-warn-restrict.c (check_call): Same. 14123 (wrestrict_dom_walker::before_dom_children): Same. 14124 (builtin_access::builtin_access): Same. 14125 (check_bounds_or_overlap): Same 14126 (maybe_diag_overlap): Same. 14127 (maybe_diag_offset_bounds): Same. 14128 * tree-diagnostic.c (default_tree_printer): Move usage of 14129 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format 14130 to this callsite. 14131 * tree-pretty-print.c (percent_K_format): Add argument. 14132 * tree-pretty-print.h: Add argument. 14133 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust. 14134 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust. 14135 (maybe_diag_stxncpy_trunc): Same. 14136 (handle_builtin_stxncpy): Same. 14137 (handle_builtin_strcat): Same. 14138 141392018-08-01 Richard Sandiford <richard.sandiford@arm.com> 14140 14141 * match.pd: Optimise pointer range checks. 14142 141432018-08-01 Richard Sandiford <richard.sandiford@arm.com> 14144 14145 PR tree-optimization/86758 14146 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try 14147 to remove pattern statements. 14148 141492018-08-01 Richard Sandiford <richard.sandiford@arm.com> 14150 14151 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the 14152 result of dfs_enumerate_from when constructing stmt_vec_infos, 14153 instead of additionally calling get_loop_body. 14154 141552018-08-01 Richard Sandiford <richard.sandiford@arm.com> 14156 14157 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p 14158 parameter. 14159 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise. 14160 When creating an iv, assert that the step is not known to be zero. 14161 (vect_setup_realignment): Update call accordingly. 14162 * tree-vect-stmts.c (vectorizable_store): Likewise. 14163 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately. 14164 141652018-08-01 Richard Sandiford <richard.sandiford@arm.com> 14166 14167 * tree-vectorizer.h (vect_stmt_to_vectorize): New function. 14168 * tree-vect-loop.c (vect_update_vf_for_slp): Use it. 14169 (vectorizable_reduction): Likewise. 14170 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise. 14171 (vect_detect_hybrid_slp_stmts): Likewise. 14172 * tree-vect-stmts.c (vect_is_simple_use): Likewise. 14173 141742018-08-01 Aldy Hernandez <aldyh@redhat.com> 14175 14176 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to... 14177 (wide_int_set_zero_nonzero_bits): ...this. 14178 (zero_nonzero_bits_from_vr): Rename to... 14179 (vrp_set_zero_nonzero_bits): ...this. 14180 (extract_range_from_multiplicative_op_1): Abstract wide int 14181 code... 14182 (wide_int_range_multiplicative_op): ...here. 14183 (extract_range_from_binary_expr_1): Extract wide int binary 14184 operations into their own functions. 14185 (wide_int_range_lshift): New. 14186 (wide_int_range_can_optimize_bit_op): New. 14187 (wide_int_range_shift_undefined_p): New. 14188 (wide_int_range_bit_xor): New. 14189 (wide_int_range_bit_ior): New. 14190 (wide_int_range_bit_and): New. 14191 (wide_int_range_trunc_mod): New. 14192 (extract_range_into_wide_ints): New. 14193 (vrp_shift_undefined_p): New. 14194 (extract_range_from_multiplicative_op): New. 14195 (vrp_can_optimize_bit_op): New. 14196 * tree-vrp.h (value_range::dump): New. 14197 (wide_int_range_multiplicative_op): New. 14198 (wide_int_range_lshift):New. 14199 (wide_int_range_shift_undefined_p): New. 14200 (wide_int_range_bit_xor): New. 14201 (wide_int_range_bit_ior): New. 14202 (wide_int_range_bit_and): New. 14203 (wide_int_range_trunc_mod): New. 14204 (zero_nonzero_bits_from_bounds): Rename to... 14205 (wide_int_set_zero_nonzero_bits): ...this. 14206 (zero_nonzero_bits_from_vr): Rename to... 14207 (vrp_set_zero_nonzero_bits): ...this. 14208 (range_easy_mask_min_max): Rename to... 14209 (wide_int_range_can_optimize_bit_op): this. 14210 * vr-values.c (simplify_bit_ops_using_ranges): Rename 14211 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits. 14212 142132018-08-01 Richard Sandiford <richard.sandiford@arm.com> 14214 14215 * tree-vectorizer.h (vect_orig_stmt): New function. 14216 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it. 14217 * tree-vect-loop.c (vect_model_reduction_cost): Likewise. 14218 (vect_create_epilog_for_reduction): Likewise. 14219 (vectorizable_live_operation): Likewise. 14220 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise. 14221 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise. 14222 * tree-vect-stmts.c (vectorizable_call): Likewise. 14223 (vectorizable_simd_clone_call, vect_remove_stores): Likewise. 14224 142252018-08-01 Richard Sandiford <richard.sandiford@arm.com> 14226 14227 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store 14228 argument. 14229 * tree-vect-stmts.c (vect_transform_stmt): Likewise. 14230 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly. 14231 (vect_transform_loop): Likewise. 14232 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise. 14233 142342018-08-01 Richard Sandiford <richard.sandiford@arm.com> 14235 14236 * tree-vectorizer.h (vect_schedule_slp): Return void. 14237 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise. 14238 (vect_schedule_slp): Likewise. 14239 142402018-08-01 Richard Sandiford <richard.sandiford@arm.com> 14241 14242 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled 14243 argument. 14244 (vect_transform_loop): Update calls accordingly. Schedule SLP 14245 instances before the main loop, if any exist. 14246 142472018-08-01 Richard Sandiford <richard.sandiford@arm.com> 14248 14249 PR tree-optimization/86749 14250 * tree-vect-patterns.c (vect_determine_min_output_precision_1): 14251 If the lhs is used in a COND_EXPR, check that it is being used 14252 as the "then" or "else" value. 14253 142542018-08-01 Tom de Vries <tdevries@suse.de> 14255 14256 PR target/86800 14257 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to 14258 speculation_safe_value_not_needed. 14259 142602018-08-01 Richard Biener <rguenther@suse.de> 14261 14262 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses 14263 as base and offset. 14264 142652018-08-01 Martin Liska <mliska@suse.cz> 14266 14267 * value-prof.c (gimple_divmod_fixed_value_transform): Unify 14268 format how successful transformation is dumped. 14269 (gimple_mod_pow2_value_transform): Likewise. 14270 (gimple_mod_subtract_transform): Likewise. 14271 (gimple_stringops_transform): Likewise. 14272 142732018-08-01 Martin Liska <mliska@suse.cz> 14274 14275 PR value-prof/35543 14276 * value-prof.c (interesting_stringop_to_profile_p): 14277 Simplify the code and add BUILT_IN_MEMMOVE. 14278 (gimple_stringops_transform): Likewise. 14279 142802018-08-01 Sam Tebbs <sam.tebbs@arm.com> 14281 14282 * config/aarch64/aarch64-simd.md 14283 (*aarch64_get_lane_zero_extendsi<mode>): Rename to... 14284 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and 14285 use GPI iterator instead of SI mode. 14286 142872018-08-01 Richard Earnshaw <rearnsha@arm.com> 14288 14289 * config/rs6000/rs6000.md (speculation_barrier): Renamed from 14290 rs6000_speculation_barrier. 14291 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for 14292 new barrier pattern name. 14293 142942018-08-01 Richard Earnshaw <rearnsha@arm.com> 14295 14296 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER. 14297 (speculation_barrier): New insn. 14298 142992018-08-01 Richard Biener <rguenther@suse.de> 14300 14301 PR bootstrap/86724 14302 * graphite.h: Include isl/id.h and isl/space.h to allow build 14303 with ISL 0.20. 14304 143052018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com> 14306 14307 PR target/86651 14308 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO 14309 mode for COFF targets. 14310 * defaults.h (TARGET_COFF): Define. 14311 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END, 14312 TARGET_COFF): Define. 14313 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare. 14314 * config/i386/djgpp.c (saved_debug_info_level): New static variable. 14315 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions. 14316 143172018-07-31 Alexandre Oliva <oliva@adacore.com> 14318 14319 * gimple-streamer-in.c (input_bb): Restore BB discriminator. 14320 * gimple-streamer-out.c (output_bb): Save it. 14321 * lto-streamer-in.c (input_struct_function_base): Restore 14322 instance discriminator if available. Create map on demand. 14323 * lto-streamer-out.c (output_struct_function_base): Save it if 14324 available. 14325 * final.c (decl_to_instance_map): Document LTO strategy. 14326 143272018-07-31 Alexandre Oliva <oliva@adacore.com> 14328 Olivier Hainque <hainque@adacore.com> 14329 14330 * debug.h (decl_to_instance_map_t): New type. 14331 (decl_to_instance_map): Declare. 14332 (maybe_create_decl_to_instance_map): New inline function. 14333 * final.c (bb_discriminator, last_bb_discriminator): New statics, 14334 to track basic block discriminators. 14335 (final_start_function_1): Initialize them. 14336 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track 14337 bb_discriminator. 14338 (decl_to_instance_map): New variable. 14339 (map_decl_to_instance, maybe_set_discriminator): New functions. 14340 (notice_source_line): Set discriminator. 14341 143422018-07-31 Ian Lance Taylor <iant@golang.org> 14343 14344 * targhooks.c (default_have_speculation_safe_value): Add 14345 ATTRIBUTE_UNUSED. 14346 143472018-07-31 David Malcolm <dmalcolm@redhat.com> 14348 14349 * dump-context.h: Include "pretty-print.h". 14350 (dump_context::refresh_dumps_are_enabled): New decl. 14351 (dump_context::emit_item): New decl. 14352 (class dump_context): Add fields "m_test_pp" and 14353 "m_test_pp_flags". 14354 (temp_dump_context::temp_dump_context): Add param "test_pp_flags". 14355 (temp_dump_context::get_dumped_text): New decl. 14356 (class temp_dump_context): Add field "m_pp". 14357 * dumpfile.c (refresh_dumps_are_enabled): Convert to... 14358 (dump_context::refresh_dumps_are_enabled): ...and add a test for 14359 m_test_pp. 14360 (set_dump_file): Update for above change. 14361 (set_alt_dump_file): Likewise. 14362 (dump_loc): New overload, taking a pretty_printer *. 14363 (dump_context::dump_loc): Call end_any_optinfo. Dump the location 14364 to any test pretty-printer. 14365 (make_item_for_dump_gimple_stmt): New function, adapted from 14366 optinfo::add_gimple_stmt. 14367 (dump_context::dump_gimple_stmt): Call it, and use the result, 14368 eliminating the direct usage of dump_file and alt_dump_file in 14369 favor of indirectly using them via emit_item. 14370 (make_item_for_dump_gimple_expr): New function, adapted from 14371 optinfo::add_gimple_expr. 14372 (dump_context::dump_gimple_expr): Call it, and use the result, 14373 eliminating the direct usage of dump_file and alt_dump_file in 14374 favor of indirectly using them via emit_item. 14375 (make_item_for_dump_generic_expr): New function, adapted from 14376 optinfo::add_tree. 14377 (dump_context::dump_generic_expr): Call it, and use the result, 14378 eliminating the direct usage of dump_file and alt_dump_file in 14379 favor of indirectly using them via emit_item. 14380 (make_item_for_dump_printf_va): New function, adapted from 14381 optinfo::add_printf_va. 14382 (make_item_for_dump_printf): New function. 14383 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va, 14384 and use the result, eliminating the direct usage of dump_file and 14385 alt_dump_file in favor of indirectly using them via emit_item. 14386 (make_item_for_dump_dec): New function. 14387 (dump_context::dump_dec): Call it, and use the result, 14388 eliminating the direct usage of dump_file and alt_dump_file in 14389 favor of indirectly using them via emit_item. 14390 (make_item_for_dump_symtab_node): New function, adapted from 14391 optinfo::add_symtab_node. 14392 (dump_context::dump_symtab_node): Call it, and use the result, 14393 eliminating the direct usage of dump_file and alt_dump_file in 14394 favor of indirectly using them via emit_item. 14395 (dump_context::begin_scope): Reimplement, avoiding direct usage 14396 of dump_file and alt_dump_file in favor of indirectly using them 14397 via emit_item. 14398 (dump_context::emit_item): New member function. 14399 (temp_dump_context::temp_dump_context): Add param "test_pp_flags". 14400 Set up test pretty-printer on the underlying context. Call 14401 refresh_dumps_are_enabled. 14402 (temp_dump_context::~temp_dump_context): Call 14403 refresh_dumps_are_enabled. 14404 (temp_dump_context::get_dumped_text): New member function. 14405 (selftest::verify_dumped_text): New function. 14406 (ASSERT_DUMPED_TEXT_EQ): New macro. 14407 (selftest::test_capture_of_dump_calls): Run all tests twice, with 14408 and then without optinfo enabled. Add uses of 14409 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes. 14410 * dumpfile.h: Update comment for the dump_* API. 14411 * optinfo-emit-json.cc 14412 (selftest::test_building_json_from_dump_calls): Update for new 14413 param for temp_dump_context ctor. 14414 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param 14415 and "m_owned" field. 14416 (optinfo_item::~optinfo_item): Likewise. 14417 (optinfo::add_item): New member function. 14418 (optinfo::emit): Update comment. 14419 (optinfo::add_string): Delete. 14420 (optinfo::add_printf): Delete. 14421 (optinfo::add_printf_va): Delete. 14422 (optinfo::add_gimple_stmt): Delete. 14423 (optinfo::add_gimple_expr): Delete. 14424 (optinfo::add_tree): Delete. 14425 (optinfo::add_symtab_node): Delete. 14426 (optinfo::add_dec): Delete. 14427 * optinfo.h (class dump_context): New forward decl. 14428 (optinfo::add_item): New decl. 14429 (optinfo::add_string): Delete. 14430 (optinfo::add_printf): Delete. 14431 (optinfo::add_printf_va): Delete. 14432 (optinfo::add_gimple_stmt): Delete. 14433 (optinfo::add_gimple_expr): Delete. 14434 (optinfo::add_tree): Delete. 14435 (optinfo::add_symtab_node): Delete. 14436 (optinfo::add_dec): Delete. 14437 (optinfo::add_poly_int): Delete. 14438 (optinfo_item::optinfo_item): Remove "owned" param. 14439 (class optinfo_item): Remove field "m_owned". 14440 144412018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com> 14442 14443 PR middle-end/86705 14444 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when 14445 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the 14446 requested variable alignment. 14447 (expand_one_ssa_partition): Likewise. 14448 (expand_one_var): Likewise. 14449 144502018-07-31 Richard Earnshaw <rearnsha@arm.com> 14451 14452 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine 14453 to speculation_safe_value_not_needed. 14454 144552018-07-31 Richard Earnshaw <rearnsha@arm.com> 14456 14457 * targhooks.h (speculation_safe_value_not_needed): New prototype. 14458 * targhooks.c (speculation_safe_value_not_needed): New function. 14459 * target.def (have_speculation_safe_value): Update documentation. 14460 * doc/tm.texi: Regenerated. 14461 144622018-07-31 Richard Earnshaw <rearnsha@arm.com> 14463 14464 * config/aarch64/iterators.md (ALLI_TI): New iterator. 14465 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New 14466 expand. 14467 (despeculate_copy<ALLI:mode>_insn): New insn. 14468 (despeculate_copyti_insn): New insn. 14469 (despeculate_simple<ALLI:mode>): New insn 14470 (despeculate_simpleti): New insn. 14471 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New 14472 function. 14473 (TARGET_SPECULATION_SAFE_VALUE): Redefine to 14474 aarch64_speculation_safe_value. 14475 (aarch64_print_operand): Handle const0_rtx in modifier 'H'. 14476 144772018-07-31 Richard Earnshaw <rearnsha@arm.com> 14478 14479 * config/aarch64/aarch64-speculation.cc: New file. 14480 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add 14481 before pass_reorder_blocks. 14482 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add 14483 prototype. 14484 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix 14485 X14 and X15 when tracking speculation. 14486 * config/aarch64/aarch64.md (register name constants): Add 14487 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM. 14488 (unspec): Add UNSPEC_SPECULATION_TRACKER. 14489 (speculation_barrier): New insn attribute. 14490 (cmp<mode>): Allow SP in comparisons. 14491 (speculation_tracker): New insn. 14492 (speculation_barrier): Add speculation_barrier attribute. 14493 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o. 14494 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs. 14495 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation. 14496 144972018-07-31 Richard Earnshaw <rearnsha@arm.com> 14498 14499 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when 14500 aarch64_track_speculation is true. 14501 (tb<optab><mode>1): Likewise. 14502 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not 14503 generate CB[N]Z when tracking speculation. 14504 (aarch64_split_compare_and_swap): Likewise. 14505 (aarch64_split_atomic_op): Likewise. 14506 145072018-07-31 Richard Earnshaw <rearnsha@arm.com> 14508 14509 * config/aarch64/aarch64.opt (mtrack-speculation): New target option. 14510 145112018-07-31 Richard Earnshaw <rearnsha@arm.com> 14512 14513 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER. 14514 (speculation_barrier): New insn. 14515 145162018-07-31 Richard Earnshaw <rearnsha@arm.com> 14517 14518 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER. 14519 * config/arm/arm.md (speculation_barrier): New expand. 14520 (speculation_barrier_insn): New pattern. 14521 145222018-07-31 Richard Earnshaw <rearnsha@arm.com> 14523 14524 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type. 14525 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise. 14526 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise. 14527 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute 14528 list. 14529 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin. 14530 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin. 14531 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise. 14532 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise. 14533 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise. 14534 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise. 14535 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise. 14536 * builtins.c (expand_speculation_safe_value): New function. 14537 (expand_builtin): Call it. 14538 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE. 14539 * doc/extend.texi: Document __builtin_speculation_safe_value. 14540 * doc/md.texi: Document "speculation_barrier" pattern. 14541 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and 14542 TARGET_HAVE_SPECULATION_SAFE_VALUE. 14543 * doc/tm.texi: Regenerated. 14544 * target.def (have_speculation_safe_value, speculation_safe_value): New 14545 hooks. 14546 * targhooks.c (default_have_speculation_safe_value): New function. 14547 (default_speculation_safe_value): New function. 14548 * targhooks.h (default_have_speculation_safe_value): Add prototype. 14549 (default_speculation_safe_value): Add prototype. 14550 145512018-07-31 David Malcolm <dmalcolm@redhat.com> 14552 14553 * dump-context.h (dump_context::dump_loc): New decl. 14554 * dumpfile.c (dump_context::dump_loc): New member function. 14555 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc 14556 and dump_gimple_stmt. 14557 (dump_context::dump_gimple_expr_loc): Likewise, using 14558 dump_gimple_expr. 14559 (dump_context::dump_generic_expr_loc): Likewise, using 14560 dump_generic_expr. 14561 (dump_context::dump_printf_loc_va): Likewise, using 14562 dump_printf_va. 14563 (dump_context::begin_scope): Explicitly using the global function 14564 "dump_loc", rather than the member function. 14565 145662018-07-31 Martin Sebor <msebor@redhat.com> 14567 14568 PR tree-optimization/86741 14569 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types. 14570 145712018-07-31 Andreas Krebbel <krebbel@linux.ibm.com> 14572 14573 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to 14574 depend on whether prefetch instructions will be emitted or not. 14575 Use TARGET_SETMEM_PFD for checking whether prefetch instructions 14576 will be emitted or not. 14577 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE) 14578 (TARGET_SETMEM_PFD): New macros. 14579 145802018-07-31 Richard Sandiford <richard.sandiford@arm.com> 14581 14582 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef. 14583 (NULL_STMT_VEC_INFO): Delete. 14584 (stmt_vec_info::operator*): Likewise. 14585 (stmt_vec_info::operator gimple *): Likewise. 14586 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead 14587 of NULL_STMT_VEC_INFO. 14588 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise. 14589 (vect_reassociating_reduction_p): Likewise. 14590 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise. 14591 (vectorizable_store): Likewise. 14592 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise. 14593 (vec_info::free_stmt_vec_infos): Likewise. 14594 145952018-07-31 Richard Sandiford <richard.sandiford@arm.com> 14596 14597 * tree-vectorizer.h (vect_stmt_in_region_p): Delete. 14598 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise. 14599 146002018-07-31 Richard Sandiford <richard.sandiford@arm.com> 14601 14602 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt) 14603 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos) 14604 (vec_info::free_stmt_vec_info): New private member functions. 14605 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt) 14606 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete. 14607 * tree-parloops.c (gather_scalar_reductions): Remove calls to 14608 set_stmt_vec_info_vec and free_stmt_vec_infos. 14609 * tree-vect-loop.c (_loop_vec_info): Remove call to 14610 set_stmt_vec_info_vec. 14611 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec) 14612 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of... 14613 * tree-vectorizer.c (vec_info::new_stmt_vec_info) 14614 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos) 14615 (vec_info::free_stmt_vec_info): ...these new functions. Remove 14616 assignments in {vec_info::,}new_stmt_vec_info that are redundant 14617 with the clearing in the xcalloc. 14618 (stmt_vec_info_vec): Delete. 14619 (vec_info::vec_info): Don't call set_stmt_vec_info_vec. 14620 (vectorize_loops): Likewise. 14621 (vec_info::~vec_info): Remove argument from call to 14622 free_stmt_vec_infos. 14623 (vec_info::add_stmt): Remove vinfo argument from call to 14624 new_stmt_vec_info. 14625 146262018-07-31 Richard Sandiford <richard.sandiford@arm.com> 14627 14628 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info 14629 rather than a gimple stmt. 14630 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free 14631 information for pattern statements when passed the original 14632 statement; instead wait to be passed the pattern statement itself. 14633 Don't call set_vinfo_for_stmt here. 14634 (free_stmt_vec_infos): Update call to free_stmt_vec_info. 14635 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free 14636 stmt_vec_infos here. 14637 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise. 14638 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's 14639 stmt_vec_infos entry. 14640 146412018-07-31 Richard Sandiford <richard.sandiford@arm.com> 14642 14643 * tree-vectorizer.h (vec_info::replace_stmt): Declare. 14644 * tree-vectorizer.c (vec_info::replace_stmt): New function. 14645 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it. 14646 * tree-vect-stmts.c (vectorizable_call): Likewise. 14647 (vectorizable_simd_clone_call): Likewise. 14648 146492018-07-31 Richard Sandiford <richard.sandiford@arm.com> 14650 14651 * tree-vectorizer.h (vec_info::remove_stmt): Declare. 14652 * tree-vectorizer.c (vec_info::remove_stmt): New function. 14653 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it. 14654 * tree-vect-loop.c (vect_transform_loop): Likewise. 14655 * tree-vect-slp.c (vect_schedule_slp): Likewise. 14656 * tree-vect-stmts.c (vect_remove_stores): Likewise. 14657 146582018-07-31 Richard Sandiford <richard.sandiford@arm.com> 14659 14660 * tree-vectorizer.h (vec_info::lookup_dr): New member function. 14661 (vect_dr_stmt): Delete. 14662 * tree-vectorizer.c (vec_info::lookup_dr): New function. 14663 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead 14664 of DR_VECT_AUX. 14665 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr) 14666 (vect_analyze_data_ref_dependence, vect_record_base_alignments) 14667 (vect_verify_datarefs_alignment, vect_peeling_supportable) 14668 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list) 14669 (vect_analyze_data_refs): Likewise. 14670 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info 14671 argument. 14672 (vect_find_same_alignment_drs): Likewise. 14673 (vect_slp_analyze_node_dependences): Update calls accordingly. 14674 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr 14675 instead of DR_VECT_AUX. 14676 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead 14677 of a vector data references. Use vec_info::lookup_dr instead of 14678 DR_VECT_AUX. 14679 (vect_peeling_hash_get_lowest_cost): Update calls accordingly. 14680 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr 14681 instead of DR_VECT_AUX. 14682 146832018-07-31 Richard Sandiford <richard.sandiford@arm.com> 14684 14685 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to 14686 dr_vec_info. 14687 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update 14688 accordingly. 14689 * tree-vect-loop.c (vect_analyze_loop_2): Likewise. 14690 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise. 14691 (vect_gen_prolog_loop_niters): Likewise. 14692 146932018-07-31 Richard Sandiford <richard.sandiford@arm.com> 14694 14695 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment) 14696 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p) 14697 (vect_known_alignment_in_bytes, vect_dr_behavior) 14698 (vect_get_scalar_dr_size): Take references as dr_vec_infos 14699 instead of data_references. Update calls to other routines for 14700 which the same change has been made. 14701 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take 14702 dr_vec_infos instead of stmt_vec_infos. 14703 (vect_analyze_data_ref_dependence): Update call accordingly. 14704 (vect_slp_analyze_data_ref_dependence) 14705 (vect_record_base_alignments): Use DR_VECT_AUX. 14706 (vect_calculate_target_alignment, vect_compute_data_ref_alignment) 14707 (vect_update_misalignment_for_peel, verify_data_ref_alignment) 14708 (vector_alignment_reachable_p, vect_get_data_access_cost) 14709 (vect_peeling_supportable, vect_analyze_group_access_1) 14710 (vect_analyze_group_access, vect_analyze_data_ref_access) 14711 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align) 14712 (vect_compile_time_alias, vect_small_gap_p) 14713 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info): 14714 (vect_supportable_dr_alignment): Take references as dr_vec_infos 14715 instead of data_references. Update calls to other routines for 14716 which the same change has been made. 14717 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs) 14718 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment) 14719 (vect_slp_analyze_and_verify_node_alignment) 14720 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list) 14721 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr) 14722 (vect_setup_realignment): Use dr_vec_infos. Update calls after 14723 above changes. 14724 (_vect_peel_info::dr): Replace with... 14725 (_vect_peel_info::dr_info): ...this new field. 14726 (vect_peeling_hash_get_most_frequent) 14727 (vect_peeling_hash_choose_best_peeling): Update accordingly. 14728 (vect_peeling_hash_get_lowest_cost): 14729 (vect_enhance_data_refs_alignment): Likewise. Update calls to other 14730 routines for which the same change has been made. 14731 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a 14732 data_reference. 14733 * tree-vect-loop-manip.c (get_misalign_in_elems) 14734 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after 14735 above changes. 14736 * tree-vect-loop.c (vect_analyze_loop_2): Likewise. 14737 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost) 14738 (vect_truncate_gather_scatter_offset, compare_step_with_zero) 14739 (get_group_load_store_type, get_negative_load_store_type) 14740 (vect_get_data_ptr_increment, vectorizable_store) 14741 (vectorizable_load): Likewise. 14742 (ensure_base_align): Take a dr_vec_info instead of a data_reference. 14743 Update calls to other routines for which the same change has been made. 14744 147452018-07-31 Richard Sandiford <richard.sandiford@arm.com> 14746 14747 * tree-vectorizer.h (vec_info::move_dr): New member function. 14748 (dataref_aux): Rename to... 14749 (dr_vec_info): ...this and add "dr" and "stmt" fields. 14750 (_stmt_vec_info::dr_aux): Update accordingly. 14751 (_stmt_vec_info::data_ref_info): Delete. 14752 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT) 14753 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT) 14754 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT): 14755 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead 14756 of data_ref. 14757 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue. 14758 (STMT_VINFO_DR_INFO): New macro. 14759 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt. 14760 (set_dr_misalignment): Update after rename of dataref_aux. 14761 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt. 14762 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant 14763 initialization of STMT_VINFO_DATA_REF. 14764 * tree-vectorizer.c (vec_info::move_dr): New function. 14765 * tree-vect-patterns.c (vect_recog_bool_pattern) 14766 (vect_recog_mask_conversion_pattern) 14767 (vect_recog_gather_scatter_pattern): Use it. 14768 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize 14769 the "dr" and "stmt" fields of dr_vec_info instead of 14770 STMT_VINFO_DATA_REF. 14771 147722018-07-31 Richard Sandiford <richard.sandiford@arm.com> 14773 14774 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field. 14775 (is_pattern_stmt_p): Use it. 14776 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p 14777 on pattern statements. 14778 147792018-07-31 Richard Sandiford <richard.sandiford@arm.com> 14780 14781 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the 14782 original stmt as a stmt_vec_info rather than a gimple stmt. 14783 (vect_pattern_recog_1): Take the statement directly as a 14784 stmt_vec_info, rather than via a gimple_stmt_iterator. 14785 Update call to vect_mark_pattern_stmts. 14786 (vect_pattern_recog): Update calls accordingly. 14787 147882018-07-31 Richard Sandiford <richard.sandiford@arm.com> 14789 14790 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy) 14791 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than 14792 a vect_def_type for the first argument. 14793 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise. 14794 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original 14795 operand if it isn't defined by a vectorized statement. 14796 (vect_build_gather_load_calls): Remove the mask_dt argument and 14797 update calls to vect_get_vec_def_for_stmt_copy. 14798 (vectorizable_bswap): Likewise the dt argument. 14799 (vectorizable_call): Update calls to vectorizable_bswap and 14800 vect_get_vec_def_for_stmt_copy. 14801 (vectorizable_simd_clone_call, vectorizable_assignment) 14802 (vectorizable_shift, vectorizable_operation, vectorizable_condition) 14803 (vectorizable_comparison): Update calls to 14804 vect_get_vec_def_for_stmt_copy. 14805 (vectorizable_store): Likewise. Remove now-unnecessary calls to 14806 vect_is_simple_use. 14807 (vect_get_loop_based_defs): Remove dt argument and update call 14808 to vect_get_vec_def_for_stmt_copy. 14809 (vectorizable_conversion): Update calls to vect_get_loop_based_defs 14810 and vect_get_vec_def_for_stmt_copy. 14811 (vectorizable_load): Update calls to vect_build_gather_load_calls 14812 and vect_get_vec_def_for_stmt_copy. 14813 * tree-vect-loop.c (vect_create_epilog_for_reduction) 14814 (vectorizable_reduction, vectorizable_live_operation): Update calls 14815 to vect_get_vec_def_for_stmt_copy. 14816 148172018-07-31 Richard Sandiford <richard.sandiford@arm.com> 14818 14819 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info 14820 and gimple stmt arguments with a stmt_vec_info. 14821 (vect_record_base_alignments): Update calls accordingly. 14822 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info 14823 and gimple stmt arguments with a stmt_vec_info. 14824 (vect_build_slp_tree_1): Remove vinfo argument and update call 14825 to vect_record_max_nunits. 14826 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1 14827 and vect_record_max_nunits. 14828 148292018-07-31 Richard Sandiford <richard.sandiford@arm.com> 14830 14831 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down 14832 file and take a stmt_vec_info instead of a gimple stmt. 14833 (supportable_widening_operation, vect_finish_replace_stmt) 14834 (vect_finish_stmt_generation, vect_get_store_rhs) 14835 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand) 14836 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt) 14837 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition) 14838 (vect_get_smallest_scalar_type, vect_check_gather_scatter) 14839 (vect_create_data_ref_ptr, bump_vector_ptr) 14840 (vect_permute_store_chain, vect_setup_realignment) 14841 (vect_transform_grouped_load, vect_record_grouped_load_vectors) 14842 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation) 14843 (vectorizable_reduction, vectorizable_induction) 14844 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant) 14845 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather 14846 than gimple stmts as arguments. 14847 * tree-vect-data-refs.c (vect_get_smallest_scalar_type) 14848 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences) 14849 (can_group_stmts_p, vect_check_gather_scatter) 14850 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr) 14851 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment) 14852 (vect_permute_load_chain, vect_shift_permute_load_chain) 14853 (vect_transform_grouped_load) 14854 (vect_record_grouped_load_vectors): Likewise. 14855 * tree-vect-loop.c (vect_fixup_reduc_chain) 14856 (get_initial_def_for_reduction, vect_create_epilog_for_reduction) 14857 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction) 14858 (vectorizable_reduction, vectorizable_induction) 14859 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise. 14860 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts) 14861 (vect_get_load_store_mask): Likewise. 14862 * tree-vect-slp.c (vect_get_place_in_interleaving_chain) 14863 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise. 14864 * tree-vect-stmts.c (vect_mark_relevant) 14865 (is_simple_and_all_uses_invariant) 14866 (exist_non_indexing_operands_for_use_p, process_use) 14867 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1) 14868 (vect_get_vec_def_for_operand, vect_get_vec_defs) 14869 (vect_finish_stmt_generation_1, vect_finish_replace_stmt) 14870 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset) 14871 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type) 14872 (get_negative_load_store_type, get_load_store_type) 14873 (vect_check_load_store_mask, vect_check_store_rhs) 14874 (vect_build_gather_load_calls, vect_get_strided_load_store_ops) 14875 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call) 14876 (vect_create_vectorized_demotion_stmts, vectorizable_conversion) 14877 (vectorizable_assignment, vectorizable_shift, vectorizable_operation) 14878 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses) 14879 (vectorizable_load, vectorizable_condition, vectorizable_comparison) 14880 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores) 14881 (supportable_widening_operation): Likewise. 14882 148832018-07-31 Richard Sandiford <richard.sandiford@arm.com> 14884 14885 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take 14886 a stmt_vec_info instead of a gcall. 14887 (vect_check_gather_scatter): Update call accordingly. 14888 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead 14889 of a gphi. 14890 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer) 14891 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly. 14892 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info 14893 instead of a gimple stmt. 14894 (vect_transform_loop): Update calls accordingly. 14895 * tree-vect-slp.c (vect_split_slp_store_group): Take and return 14896 stmt_vec_infos instead of gimple stmts. 14897 (vect_analyze_slp_instance): Update use accordingly. 14898 * tree-vect-stmts.c (read_vector_array, write_vector_array) 14899 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements) 14900 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask) 14901 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops) 14902 (vect_gen_widened_results_half, vect_get_loop_based_defs) 14903 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts): 14904 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos 14905 down to subroutines. 14906 149072018-07-31 Richard Sandiford <richard.sandiford@arm.com> 14908 14909 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type 14910 of the worklist from a vector of gimple stmts to a vector of 14911 stmt_vec_infos. 14912 * tree-vect-stmts.c (vect_mark_relevant, process_use) 14913 (vect_mark_stmts_to_be_vectorized): Likewise 14914 149152018-07-31 Richard Sandiford <richard.sandiford@arm.com> 14916 14917 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the 14918 statement before passing it to vect_analyze_stmt. 14919 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk 14920 the chain of phi vector definitions. Track the exit phi via its 14921 stmt_vec_info. 14922 (vectorizable_reduction): Set cond_stmt_vinfo directly from the 14923 STMT_VINFO_REDUC_DEF. 14924 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use 14925 stmt_vec_infos to handle the statement chains. 14926 (vect_get_slp_defs): Record the first statement in the node 14927 using a stmt_vec_info. 14928 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up 14929 statements here and pass their stmt_vec_info down to subroutines. 14930 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it 14931 down to vect_finish_stmt_generation. 14932 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt) 14933 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass 14934 stmt_vec_infos to subroutines. 14935 (vect_remove_stores): Use stmt_vec_infos to handle the statement 14936 chains. 14937 149382018-07-31 Richard Sandiford <richard.sandiford@arm.com> 14939 14940 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): 14941 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr) 14942 (vect_permute_store_chain, vect_setup_realignment) 14943 (vect_permute_load_chain, vect_shift_permute_load_chain) 14944 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple 14945 stmts internally, and when passing values to other vectorizer routines. 14946 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise. 14947 * tree-vect-loop.c (vect_analyze_scalar_cycles_1) 14948 (vect_analyze_loop_operations, get_initial_def_for_reduction) 14949 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction) 14950 (vectorizable_reduction, vectorizable_induction) 14951 (vectorizable_live_operation, vect_transform_loop_stmt) 14952 (vect_transform_loop): Likewise. 14953 * tree-vect-patterns.c (vect_reassociating_reduction_p) 14954 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern) 14955 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise. 14956 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise. 14957 (vect_slp_analyze_node_operations_1): Likewise. 14958 * tree-vect-stmts.c (vect_mark_relevant, process_use) 14959 (exist_non_indexing_operands_for_use_p, vect_init_vector_1) 14960 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand) 14961 (vect_finish_stmt_generation_1, get_group_load_store_type) 14962 (get_load_store_type, vect_build_gather_load_calls) 14963 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call) 14964 (vect_create_vectorized_demotion_stmts, vectorizable_conversion) 14965 (vectorizable_assignment, vectorizable_shift, vectorizable_operation) 14966 (vectorizable_store, vectorizable_load, vectorizable_condition) 14967 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt) 14968 (supportable_widening_operation): Likewise. 14969 (vect_get_vector_types_for_stmt): Likewise. 14970 * tree-vectorizer.h (vect_dr_behavior): Likewise. 14971 149722018-07-31 Richard Sandiford <richard.sandiford@arm.com> 14973 14974 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence) 14975 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses) 14976 (vect_permute_store_chain, vect_permute_load_chain) 14977 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid 14978 repeated stmt_vec_info lookups. 14979 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise. 14980 (vect_update_ivs_after_vectorizer): Likewise. 14981 * tree-vect-loop.c (vect_is_simple_reduction): Likewise. 14982 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise. 14983 * tree-vect-patterns.c (adjust_bool_stmts): Likewise. 14984 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise. 14985 (vect_bb_slp_scalar_cost): Likewise. 14986 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise. 14987 149882018-07-31 Richard Sandiford <richard.sandiford@arm.com> 14989 14990 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the 14991 gcall rather than the generic gimple stmt to gimple_call_internal_fn. 14992 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast 14993 to get gassigns and gcalls, rather than operating on generc gimple 14994 stmts. 14995 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p) 14996 (vect_mark_stmts_to_be_vectorized, vectorizable_store) 14997 (vectorizable_load, vect_analyze_stmt): Likewise. 14998 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi. 14999 150002018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15001 15002 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and 15003 return stmt_vec_infos rather than gimple stmts. Do not accept 15004 null arguments. 15005 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead 15006 of a gimple stmt. 15007 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise. 15008 Update use of get_later_stmt. 15009 (vect_get_constant_vectors): Update call accordingly. 15010 (vect_schedule_slp_instance): Likewise 15011 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise. 15012 (vect_slp_analyze_instance_dependence): Likewise. 15013 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt. 15014 150152018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15016 15017 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with... 15018 (stmt_info_for_cost::stmt_info): ...this new field. 15019 (add_stmt_costs): Update accordingly. 15020 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost) 15021 (vect_get_known_peeling_cost): Likewise. 15022 (vect_estimate_min_profitable_iters): Likewise. 15023 * tree-vect-stmts.c (record_stmt_cost): Likewise. 15024 150252018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15026 15027 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change 15028 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>. 15029 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update 15030 accordingly. 15031 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise. 15032 150332018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15034 15035 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from 15036 a gimple stmt to a stmt_vec_info. 15037 * tree-vect-stmts.c (vectorizable_load): Update accordingly. 15038 150392018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15040 15041 * tree-vectorizer.h (vec_info::grouped_stores): Change from 15042 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>. 15043 (_loop_vec_info::reduction_chains): Likewise. 15044 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update 15045 accordingly. 15046 * tree-vect-slp.c (vect_analyze_slp): Likewise. 15047 150482018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15049 15050 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from 15051 a gimple stmt to a stmt_vec_info. 15052 (_stmt_vec_info::next_element): Likewise. 15053 * tree-vect-data-refs.c (vect_update_misalignment_for_peel) 15054 (vect_slp_analyze_and_verify_node_alignment) 15055 (vect_analyze_group_access_1, vect_analyze_group_access) 15056 (vect_small_gap_p, vect_prune_runtime_alias_test_list) 15057 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors) 15058 (vect_supportable_dr_alignment): Update accordingly. 15059 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise. 15060 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction) 15061 (vect_is_simple_reduction, vectorizable_reduction): Likewise. 15062 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise. 15063 * tree-vect-slp.c (vect_build_slp_tree_1) 15064 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p) 15065 (vect_split_slp_store_group, vect_analyze_slp_instance) 15066 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise. 15067 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost) 15068 (get_group_load_store_type, get_load_store_type) 15069 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load) 15070 (vect_transform_stmt, vect_remove_stores): Likewise. 15071 150722018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15073 15074 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather 15075 than a gimple stmt. 15076 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence) 15077 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments) 15078 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment) 15079 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment) 15080 (vector_alignment_reachable_p, vect_get_data_access_cost) 15081 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost) 15082 (vect_peeling_supportable, vect_enhance_data_refs_alignment) 15083 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment) 15084 (vect_analyze_group_access_1, vect_analyze_group_access) 15085 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses) 15086 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs) 15087 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the 15088 result of vect_dr_stmt and use the stmt_vec_info instead of 15089 the associated gimple stmt. 15090 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise. 15091 (vect_gen_prolog_loop_niters): Likewise. 15092 * tree-vect-loop.c (vect_analyze_loop_2): Likewise. 15093 150942018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15095 15096 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *> 15097 to a vec<stmt_vec_info>. 15098 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly. 15099 (vect_create_new_slp_node): Take a vec<gimple *> instead of a 15100 vec<stmt_vec_info>. 15101 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *> 15102 to a vec<stmt_vec_info>. 15103 (bst_traits::value_type, bst_traits::value_type): Likewise. 15104 (bst_traits::hash): Update accordingly. 15105 (vect_get_and_check_slp_defs): Change the stmts parameter from 15106 a vec<gimple *> to a vec<stmt_vec_info>. 15107 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise. 15108 (vect_build_slp_tree): Likewise. 15109 (vect_build_slp_tree_2): Likewise. Update uses of 15110 SLP_TREE_SCALAR_STMTS. 15111 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS. 15112 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant) 15113 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts) 15114 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp) 15115 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1) 15116 (vect_slp_analyze_node_operations, vect_slp_analyze_operations) 15117 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1) 15118 (vect_get_constant_vectors, vect_get_slp_defs) 15119 (vect_transform_slp_perm_load, vect_schedule_slp_instance) 15120 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise. 15121 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos 15122 instead of gimple stmts. 15123 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change 15124 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>. 15125 (vect_slp_analyze_instance_dependence): Update uses of 15126 SLP_TREE_SCALAR_STMTS. 15127 (vect_slp_analyze_and_verify_node_alignment): Likewise. 15128 (vect_slp_analyze_and_verify_instance_alignment): Likewise. 15129 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise. 15130 (get_initial_defs_for_reduction): Likewise. 15131 (vect_create_epilog_for_reduction): Likewise. 15132 (vectorize_fold_left_reduction): Likewise. 15133 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise. 15134 (vect_model_simple_cost, vectorizable_shift, vectorizable_load) 15135 (can_vectorize_live_stmts): Likewise. 15136 151372018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15138 15139 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an 15140 auto_vec<gimple *> to an auto_vec<stmt_vec_info>. 15141 (vect_force_simple_reduction): Take and return stmt_vec_infos rather 15142 than gimple stmts. 15143 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead 15144 of a gimple stmt. 15145 (gather_scalar_reductions): Update after above interface changes. 15146 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise. 15147 (vect_is_simple_reduction): Take and return stmt_vec_infos rather 15148 than gimple stmts. 15149 (vect_force_simple_reduction): Likewise. 15150 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of 15151 LOOP_VINFO_REDUCTIONS. 15152 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise. 15153 151542018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15155 15156 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from 15157 a gimple stmt to a stmt_vec_info. 15158 * tree-vect-loop.c (vect_active_double_reduction_p) 15159 (vect_force_simple_reduction, vectorizable_reduction): Update 15160 accordingly. 15161 151622018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15163 15164 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a 15165 vec<gimple *> to a vec<stmt_vec_info>. 15166 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change 15167 the reduction_phis argument from a vec<gimple *> to a 15168 vec<stmt_vec_info>. 15169 (vectorizable_reduction): Likewise the phis local variable that 15170 is passed to vect_create_epilog_for_reduction. Update for new type 15171 of SLP_TREE_VEC_STMTS. 15172 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS. 15173 (vectorizable_live_operation): Likewise. 15174 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise. 15175 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise. 15176 151772018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15178 15179 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from 15180 a gimple stmt to a stmt_vec_info. 15181 (vectorizable_condition, vectorizable_live_operation) 15182 (vectorizable_reduction, vectorizable_induction): Pass back the 15183 vectorized statement as a stmt_vec_info. 15184 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update 15185 use of STMT_VINFO_VEC_STMT. 15186 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise, 15187 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT 15188 as stmt_vec_infos rather than gimple stmts. 15189 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt 15190 to a stmt_vec_info. 15191 (vectorizable_live_operation): Likewise. 15192 (vectorizable_reduction, vectorizable_induction): Likewise, 15193 updating use of STMT_VINFO_VEC_STMT. 15194 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use 15195 of STMT_VINFO_VEC_STMT. 15196 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call) 15197 (vectorizable_simd_clone_call, vectorizable_conversion) 15198 (vectorizable_assignment, vectorizable_shift, vectorizable_operation) 15199 (vectorizable_store, vectorizable_load, vectorizable_condition) 15200 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt 15201 from a gimple stmt to a stmt_vec_info. 15202 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a 15203 pointer to a stmt_vec_info to the vectorizable_* routines. 15204 152052018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15206 15207 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from 15208 a gimple stmt to a stmt_vec_info. 15209 (is_pattern_stmt_p): Update accordingly. 15210 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise. 15211 (vect_record_grouped_load_vectors): Likewise. 15212 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise. 15213 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise. 15214 (vect_model_reduction_cost): Likewise. 15215 (vect_create_epilog_for_reduction): Likewise. 15216 (vectorizable_reduction, vectorizable_induction): Likewise. 15217 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise. 15218 Return the stmt_vec_info for the pattern statement. 15219 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT. 15220 (vect_split_statement, vect_mark_pattern_stmts): Likewise. 15221 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise. 15222 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise. 15223 * tree-vect-stmts.c (vect_mark_relevant): Likewise. 15224 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise. 15225 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info) 15226 (free_stmt_vec_info, vect_is_simple_use): Likewise. 15227 152282018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15229 15230 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info 15231 (vect_finish_stmt_generation): Likewise. 15232 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise. 15233 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise. 15234 (vect_build_gather_load_calls): Use the return value of the above 15235 functions instead of a separate call to vinfo_for_stmt. Use narrow 15236 scopes for the input gimple stmt and wider scopes for the associated 15237 stmt_vec_info. Use vec_info::lookup_def when setting these 15238 stmt_vec_infos from an SSA_NAME definition. 15239 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call) 15240 (vect_create_vectorized_demotion_stmts, vectorizable_conversion) 15241 (vectorizable_assignment, vectorizable_shift, vectorizable_operation) 15242 (vectorizable_store, vectorizable_load, vectorizable_condition) 15243 (vectorizable_comparison): Likewise. 15244 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise. 15245 (vectorizable_reduction): Likewise. 15246 152472018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15248 15249 * tree-vectorizer.h (vect_is_simple_use): Add an optional 15250 stmt_vec_info * parameter before the optional gimple **. 15251 * tree-vect-stmts.c (vect_is_simple_use): Likewise. 15252 (process_use, vect_get_vec_def_for_operand_1): Update callers. 15253 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise. 15254 * tree-vect-loop.c (vectorizable_reduction): Likewise. 15255 (vectorizable_live_operation): Likewise. 15256 * tree-vect-patterns.c (type_conversion_p): Likewise. 15257 (vect_look_through_possible_promotion): Likewise. 15258 (vect_recog_rotate_pattern): Likewise. 15259 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise. 15260 152612018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15262 15263 * tree-vectorizer.h (stmt_vec_info): Temporarily change from 15264 a typedef to a wrapper class. 15265 (NULL_STMT_VEC_INFO): New macro. 15266 (vec_info::stmt_infos): Change to vec<stmt_vec_info>. 15267 (stmt_vec_info::operator*): New function. 15268 (stmt_vec_info::operator gimple *): Likewise. 15269 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO. 15270 (add_stmt_costs): Likewise. 15271 * tree-vect-loop-manip.c (iv_phi_p): Likewise. 15272 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost) 15273 (vect_get_known_peeling_cost): Likewise. 15274 (vect_estimate_min_profitable_iters): Likewise. 15275 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise. 15276 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise. 15277 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise. 15278 (vectorizable_store, free_stmt_vec_infos): Likewise. 15279 (new_stmt_vec_info): Change return type of xcalloc to 15280 _stmt_vec_info *. 15281 152822018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15283 15284 * tree-vectorizer.h (vec_info::lookup_single_use): Declare. 15285 * tree-vectorizer.c (vec_info::lookup_single_use): New function. 15286 * tree-vect-loop.c (vectorizable_reduction): Use it instead of 15287 a single_imm_use-based sequence. 15288 * tree-vect-stmts.c (supportable_widening_operation): Likewise. 15289 152902018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15291 15292 * tree-vectorizer.h (vec_info::lookup_def): Declare. 15293 * tree-vectorizer.c (vec_info::lookup_def): New function. 15294 * tree-vect-patterns.c (vect_get_internal_def): Use it. 15295 (vect_widened_op_tree): Likewise. 15296 * tree-vect-stmts.c (vect_is_simple_use): Likewise. 15297 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise. 15298 (vectorizable_reduction): Likewise. 15299 (vect_valid_reduction_input_p): Take a stmt_vec_info instead 15300 of a gimple *. 15301 (vect_is_slp_reduction): Update calls accordingly. Use 15302 vec_info::lookup_def. 15303 (vect_is_simple_reduction): Likewise 15304 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def. 15305 153062018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15307 15308 * tree-vectorizer.h (vec_info::lookup_stmt): Declare. 15309 * tree-vectorizer.c (vec_info::lookup_stmt): New function. 15310 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead 15311 of vinfo_for_stmt. 15312 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1) 15313 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form) 15314 (vect_update_vf_for_slp, vect_analyze_loop_operations) 15315 (vect_is_slp_reduction, vectorizable_induction) 15316 (vect_transform_loop_stmt, vect_transform_loop): Likewise. 15317 * tree-vect-patterns.c (vect_init_pattern_stmt): 15318 (vect_determine_min_output_precision_1, vect_determine_precisions) 15319 (vect_pattern_recog): Likewise. 15320 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise. 15321 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise. 15322 * config/rs6000/rs6000.c (rs6000_density_test): Likewise. 15323 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise. 15324 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2) 15325 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info 15326 info field from a loop to a loop_vec_info. 15327 153282018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15329 15330 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file. 15331 (vec_info::add_stmt): Declare. 15332 * tree-vectorizer.c (vec_info::add_stmt): New function. 15333 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it. 15334 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise. 15335 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise. 15336 (vectorizable_induction): Likewise. 15337 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise. 15338 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise. 15339 (vectorizable_simd_clone_call, vectorizable_store): Likewise. 15340 (vectorizable_load): Likewise. 15341 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise. 15342 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern) 15343 (vect_recog_gather_scatter_pattern): Likewise. 15344 (append_pattern_def_seq): Likewise. Remove a check that is 15345 performed by add_stmt itself. 15346 153472018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15348 15349 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in 15350 which make_ssa_name was called with new_stmt before new_stmt 15351 had been created. 15352 153532018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15354 15355 * tree-vect-loop.c (vect_valid_reduction_input_p): New function, 15356 split out from... 15357 (vect_is_slp_reduction): ...here... 15358 (vect_is_simple_reduction): ...and here. Remove repetition of tests 15359 that are already known to be false. 15360 153612018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15362 15363 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter. 15364 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update 15365 STMT_VINFO_NUM_SLP_USES when it's true. 15366 (vect_free_slp_instance): Add a final_p parameter and pass it to 15367 vect_free_slp_tree. 15368 (vect_build_slp_tree_2): Update call to vect_free_slp_instance. 15369 (vect_analyze_slp_instance): Likewise. 15370 (vect_slp_analyze_operations): Likewise. 15371 (vect_slp_analyze_bb_1): Likewise. 15372 * tree-vectorizer.c (vec_info): Likewise. 15373 * tree-vect-loop.c (vect_transform_loop): Likewise. 15374 153752018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15376 15377 * tree-vect-loop.c (vectorizable_reduction): Assert that the 15378 function is not called for second and subsequent members of 15379 a reduction group. 15380 153812018-07-31 Richard Sandiford <richard.sandiford@arm.com> 15382 15383 * tree-vect-loop.c (get_initial_def_for_reduction): Move special 15384 cases for nested loops from here to ... 15385 (vect_create_epilog_for_reduction): ...here. Only call 15386 vect_is_simple_use for inner-loop reductions. 15387 153882018-07-31 Martin Liska <mliska@suse.cz> 15389 15390 PR gcov-profile/85338 15391 PR gcov-profile/85350 15392 PR gcov-profile/85372 15393 * profile.c (struct location_triplet): New. 15394 (struct location_triplet_hash): Likewise. 15395 (output_location): Do not output a BB that 15396 is already recorded for a line. 15397 (branch_prob): Use streamed_locations. 15398 153992018-07-31 Martin Liska <mliska@suse.cz> 15400 15401 PR gcov-profile/85370 15402 * coverage.c (coverage_begin_function): Do not mark target 15403 clones as artificial functions. 15404 154052018-07-31 Martin Liska <mliska@suse.cz> 15406 15407 PR gcov-profile/83813 15408 PR gcov-profile/84758 15409 PR gcov-profile/85217 15410 PR gcov-profile/85332 15411 * profile.c (branch_prob): Do not record GOTO expressions 15412 for GIMPLE statements which locations are already streamed. 15413 154142018-07-31 Olivier Hainque <hainque@adacore.com> 15415 15416 * gcc.c (handle_spec_function): Accept a soft_matched_part 15417 argument, as do_spec_1. Pass it down to ... 15418 (eval_spec_function): Accept a soft_matched_part argument, 15419 and pass it down to ... 15420 (do_spec_2): Accept a soft_matched_part argument, and pass 15421 it down to do_spec_1. 15422 (do_spec_1): Pass soft_matched_part to handle_spec_function. 15423 (handle_braces): Update call to handle_spec_function. 15424 (driver::set_up_specs): Update calls to do_spec_2. 15425 (compare_debug_dump_opt_spec_function): Likewise. 15426 (compare_debug_self_opt_spec_function): Likewise. 15427 154282018-07-31 Olivier Hainque <hainque@adacore.com> 15429 15430 * common.opt (nolibc): New option. 15431 * doc/invoke.texi (Link Options): Document it. 15432 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc. 15433 * config/alpha/linux.h: Likewise. 15434 * config/arc/elf.h: Likewise. 15435 * config/arm/uclinux-elf.h: Likewise. 15436 * config/arm/unknown-elf.h: Likewise. 15437 * config/avr/avrlibc.h: Likewise. 15438 * config/bfin/bfin.h: Likewise. 15439 * config/bfin/linux.h: Likewise. 15440 * config/bfin/uclinux.h: Likewise. 15441 * config/darwin.h: Likewise. 15442 * config/darwin10.h: Likewise. 15443 * config/darwin12.h: Likewise. 15444 * config/gnu-user.h: Likewise. 15445 * config/lm32/uclinux-elf.h: Likewise. 15446 * config/pa/pa-hpux11.h: Likewise. 15447 * config/pa/pa64-hpux.h: Likewise. 15448 * config/sparc/sparc.h: Likewise. 15449 154502018-07-31 Olivier Hainque <hainque@adacore.com> 15451 15452 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed 15453 undefined variables. 15454 154552018-07-30 Segher Boessenkool <segher@kernel.crashing.org> 15456 15457 PR target/86640 15458 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode 15459 instead of GEN_INT. 15460 154612018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de> 15462 15463 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL 15464 terminated string literal. 15465 154662018-07-30 Segher Boessenkool <segher@kernel.crashing.org> 15467 15468 PR rtl-optimization/85160 15469 * combine.c (is_just_move): New function. 15470 (try_combine): Allow combining two instructions into two if neither of 15471 the original instructions was a move. 15472 154732018-07-30 Alexander Monakov <amonakov@ispras.ru> 15474 15475 PR target/86673 15476 * doc/extend.texi (Global Register Variables): Discourage use of type 15477 qualifiers. 15478 (Local Register Variables): Likewise. 15479 154802018-07-30 Richard Sandiford <richard.sandiford@arm.com> 15481 15482 PR tree-optimization/86506 15483 * hwint.h (ceil_log2): Resync with hwint.c implementation. 15484 154852018-07-30 Ilya Leoshkevich <iii@linux.ibm.com> 15486 15487 PR target/86547 15488 * lra-constraints.c (spill_hard_reg_in_range): When selecting the 15489 hard_regno, make sure no insn between `from` and `to` clobbers it. 15490 154912018-07-30 Cesar Philippidis <cesar@codesourcery.com> 15492 Tom de Vries <tdevries@suse.de> 15493 15494 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ... 15495 (PTX_DEFAULT_RUNTIME_DIM): ... this. 15496 (nvptx_goacc_validate_dims): Set default worker and gang dims to 15497 PTX_DEFAULT_RUNTIME_DIM. 15498 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER. 15499 155002018-07-29 John David Anglin <danglin@gcc.gnu.org> 15501 15502 * config/pa/pa.c (pa_output_addr_vec): Align address table. 15503 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment. 15504 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define. 15505 155062018-07-27 Michael Meissner <meissner@linux.ibm.com> 15507 15508 * config/rs6000/constraints.md (wG constraint): Delete, no longer 15509 used. 15510 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename 15511 predicate to reflect toc fusion has been deleted. 15512 (toc_fusion_mem_raw): Delete, no longer used. 15513 (toc_fusion_mem_wrapped): Likewise. 15514 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc 15515 fusion mask bit. 15516 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address): 15517 Delete, no longer used. 15518 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields 15519 meant to be used for toc fusion. 15520 (rs6000_debug_print_mode): Delete toc fusion debugging. 15521 (rs6000_debug_reg_global): Likewise. 15522 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc 15523 fusion and secondary reload support that were never used. 15524 (rs6000_option_override_internal): Delete TOC fusion, that was only 15525 partially defined, and it did not work unless you also used the 15526 -mcmodel= switch. 15527 (rs6000_legitimate_address_p): Delete TOC fusion support. 15528 (rs6000_opt_masks): Likewise. 15529 (fusion_wrap_memory_address): Delete function, no longer used. 15530 (fusion_split_address); Delete TOC fusion support. 15531 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no 15532 longer used with toc fusion being deleted. 15533 (TARGET_TOC_FUSION_FP): Likewise. 15534 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion 15535 UNSPEC. 15536 (toc fusion spliter): Delete TOC fusion support. 15537 (toc_fusionload_<mode>): Likewise. 15538 (toc_fusionload_di): Likewise. 15539 (fusion_gpr_load_<mode>): Delete generator function, this insn no 15540 longer needs to be named. Rename predicate to delete TOC fusion. 15541 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise. 15542 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise. 15543 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise. 15544 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise. 15545 (p9 fusion peephole2s): Rename predicate to delete TOC fusion. 15546 155472018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org> 15548 15549 * doc/extend.texi (Basic PowerPC Built-in Functions Available on 15550 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with 15551 __int128 in built-in function prototypes. 15552 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise. 15553 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise. 15554 155552018-07-27 Martin Sebor <msebor@redhat.com> 15556 15557 PR tree-optimization/86696 15558 * tree-ssa-strlen.c (get_min_string_length): Handle all integer 15559 types, including enums. 15560 (handle_char_store): Be prepared for the above function to fail. 15561 155622018-07-26 Qing Zhao <qing.zhao@oracle.com> 15563 15564 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining 15565 when optimization level is lower than 2 or optimize for size. 15566 155672018-07-26 Martin Sebor <msebor@redhat.com> 15568 15569 PR tree-optimization/86043 15570 PR tree-optimization/86042 15571 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps. 15572 (get_string_cst_length): Rename... 15573 (get_min_string_length): ...to this. Add argument. 15574 (handle_char_store): Extend to handle multi-character stores by 15575 MEM_REF. 15576 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF. 15577 * tree.h (initializer_zerop): Add argument. 15578 155792018-07-26 Jakub Jelinek <jakub@redhat.com> 15580 15581 PR middle-end/86660 15582 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for 15583 declare target to variables if they have always,{to,from,tofrom} map 15584 kinds. 15585 155862018-07-26 Martin Liska <mliska@suse.cz> 15587 15588 PR lto/86548 15589 * lto-wrapper.c: Add linker_output as prefix 15590 for ltrans_output_file. 15591 155922018-07-26 Segher Boessenkool <segher@kernel.crashing.org> 15593 15594 PR rtl-optimization/85805 15595 * combine.c (reg_nonzero_bits_for_combine): Only use the last set 15596 value for hard registers if that was written in the same mode. 15597 155982018-07-26 Martin Liska <mliska@suse.cz> 15599 15600 PR gcov-profile/86536 15601 * gcov.c (format_gcov): Use printf format %.*f directly 15602 and do not handle special values. 15603 156042018-07-25 Claudiu Zissulescu <claziss@synopsys.com> 15605 15606 * common/config/arc/arc-common.c (arc_option_optimization_table): 15607 Update default optimizations for size. 15608 156092018-07-25 Claudiu Zissulescu <claziss@synopsys.com> 15610 15611 * config/arc/arc.md (movsf_insn): Add short instruction selection. 15612 * config/arc/constraints.md (CfZ): New constraint. 15613 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint. 15614 (subsf3_fpu): Likewise. 15615 (cmpsf_fpu): Likewise. 15616 (cmpsf_fpu_uneq): Likewise. 15617 156182018-07-25 Claudiu Zissulescu <claziss@synopsys.com> 15619 15620 * config/arc/arc.c (compact_memory_operand_p): Check for uncached 15621 accesses as well. 15622 (arc_is_uncached_mem_p): uncached applies to both the variable and 15623 the pointer. 15624 156252018-07-25 Claudiu Zissulescu <claziss@synopsys.com> 15626 15627 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional 15628 register names. 15629 156302018-07-25 David Malcolm <dmalcolm@redhat.com> 15631 15632 * optinfo-emit-json.cc (class optrecord_json_writer): Convert 15633 field "m_scopes" from vec to auto_vec. 15634 156352018-07-25 Martin Liska <mliska@suse.cz> 15636 15637 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix 15638 return type. 15639 156402018-07-25 Richard Biener <rguenther@suse.de> 15641 15642 PR debug/86654 15643 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions 15644 special wrt context_die late. 15645 (gen_subprogram_die): Re-use DIEs in local scope. 15646 156472018-07-25 Richard Sandiford <richard.sandiford@arm.com> 15648 15649 PR tree-optimization/86644 15650 * hwint.c (ceil_log2): Fix comment. Return 0 for 0. 15651 156522018-07-25 Martin Liska <mliska@suse.cz> 15653 15654 PR middle-end/86645 15655 * dumpfile.c: And excluded values with TDF_ALL_VALUES. 15656 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES. 15657 156582018-07-25 Martin Liska <mliska@suse.cz> 15659 15660 PR sanitizer/79635 15661 * params.def: Explain ASan abbreviation and provide 15662 a documentation link. 15663 156642018-07-24 Martin Sebor <msebor@redhat.com> 15665 15666 PR tree-optimization/86622 15667 PR tree-optimization/86532 15668 * builtins.h (string_length): Declare. 15669 * builtins.c (c_strlen): Correct handling of non-constant offsets. 15670 (check_access): Be prepared for non-constant length ranges. 15671 (string_length): Make extern. 15672 * expr.c (string_constant): Only handle the minor non-constant 15673 array index. Use string_constant to compute the length of 15674 a generic string constant. 15675 156762018-07-24 Richard Sandiford <richard.sandiford@arm.com> 15677 15678 PR tree-optimization/86618 15679 * tree-vect-stmts.c (vectorizable_call): Don't take the address 15680 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null. 15681 156822018-07-24 David Malcolm <dmalcolm@redhat.com> 15683 15684 PR tree-optimization/86636 15685 * json.cc (json::object::set): Fix comment. Add assertions. 15686 (json::array::append): Move here from json.h. Add comment and an 15687 assertion. 15688 (json::string::string): Likewise. 15689 * json.h (json::array::append): Move to json.cc. 15690 (json::string::string): Likewise. 15691 * optinfo-emit-json.cc 15692 (optrecord_json_writer::impl_location_to_json): Assert that we 15693 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc 15694 wrapper around it. Expand the location once, rather than three 15695 times. 15696 (optrecord_json_writer::inlining_chain_to_json): Fix the check for 15697 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc 15698 wrappers. 15699 (optrecord_json_writer::optinfo_to_json): Likewise, in four 15700 places. Fix some overlong lines. 15701 157022018-07-24 Matthew Malcomson <matthew.malcomson@arm.com> 15703 15704 * config/aarch64/aarch64-simd.md 15705 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into... 15706 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This... 15707 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this. 15708 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into... 15709 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This... 15710 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this. 15711 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into... 15712 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This... 15713 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this. 15714 157152018-07-24 Jakub Jelinek <jakub@redhat.com> 15716 15717 PR middle-end/86627 15718 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN 15719 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT 15720 and abs_d == d, do the power of two handling if profitable. 15721 157222018-07-24 Richard Biener <rguenther@suse.de> 15723 15724 * match.pd: Add BIT_FIELD_REF canonicalizations. 15725 157262018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de> 15727 15728 PR c/86617 15729 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values. 15730 157312018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de> 15732 15733 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL 15734 terminated STRING_CST object. 15735 157362018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de> 15737 15738 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf. 15739 157402018-07-23 Segher Boessenkool <segher@kernel.crashing.org> 15741 15742 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust. 15743 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete. 15744 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force 15745 the elements into a register. 15746 (rs6000_split_v4si_init_di_reg): Delete. 15747 (rs6000_split_v4si_init): Delete. 15748 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT. 15749 (vsx_init_v4si): Rewrite as a define_expand. 15750 157512018-07-23 Segher Boessenkool <segher@kernel.crashing.org> 15752 15753 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the 15754 zero_extend argument from memory): New. 15755 157562018-07-22 Martin Sebor <msebor@redhat.com> 15757 15758 PR bootstrap/86621 15759 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid 15760 diagnosing calls with unknown arguments unless -Walloca-larger-than 15761 is restricted to less than PTRDIFF_MAX bytes. 15762 157632018-07-22 Gerald Pfeifer <gerald@pfeifer.com> 15764 15765 * doc/gcov.texi (Invoking Gcov): Editorial changes. 15766 157672018-07-20 David Malcolm <dmalcolm@redhat.com> 15768 15769 * pretty-print.c (text_info::set_location): Remove redundant 15770 "line_table" parameter from call to rich_location::set_range. 15771 157722018-07-20 Martin Sebor <msebor@redhat.com> 15773 15774 PR middle-end/82063 15775 * builtins.c (expand_builtin_alloca): Adjust. 15776 * calls.c (alloc_max_size): Simplify. 15777 * cgraphunit.c (cgraph_node::expand): Adjust. 15778 * common.opt (larger_than_size, warn_frame_larger_than): Remove 15779 variables. 15780 (frame_larger_than_size): Same. 15781 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options 15782 to take a HOST_WIDE_INT argument and accept a byte-size suffix. 15783 Initialize. 15784 * doc/invoke.texi (GCC Command Options): Document option arguments. 15785 Explain byte-size arguments and suffixes. 15786 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update. 15787 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same. 15788 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same. 15789 * doc/options.texi (UInteger): Expand. 15790 (Host_Wide_Int, ByteSize): Document new properties. 15791 * final.c (final_start_function_1): Include sizes in an error message. 15792 * function.c (frame_offset_overflow): Same. 15793 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust. 15794 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT. 15795 Diagnose unbounded alloca calls only for limits of less than 15796 PTRDIFF_MAX. 15797 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca 15798 calls and VLA size only for limits of less than PTRDIFF_MAX. Same 15799 for alloca(0). 15800 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops 15801 only for limits of less than PTRDIFF_MAX. 15802 * langhooks-def.h (lhd_handle_option): Change function argument 15803 to HOST_WIDE_INT. 15804 * langhooks.c (lhd_handle_option): Same. 15805 * langhooks.h (handle_option): Same. 15806 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and 15807 ByteSize flags. 15808 (var_type, var_type_struct): Same. 15809 (var_set): Handle ByteSize flag. 15810 * optc-gen.awk: Add comments to output to ease debugging. Make 15811 use of HOST_WIDE_INT where appropriate. 15812 * opts-gen-save.awk: Use %lx to format unsigned long. 15813 * opth-gen.awk: Change function argument to HOST_WIDE_INT. 15814 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add 15815 arguments. Parse bytes-size suffixes. 15816 (enum_arg_to_value): Change function argument to HOST_WIDE_INT. 15817 (enum_value_to_arg): Same. 15818 (decode_cmdline_option): Handle cl_host_wide_int. Adjust. 15819 (handle_option): Adjust. 15820 (generate_option): Change function argument to HOST_WIDE_INT. 15821 (cmdline_handle_error): Adjust. 15822 (read_cmdline_option): Change function argument to HOST_WIDE_INT. 15823 (set_option): Change function argument to HOST_WIDE_INT. 15824 (option_enabled): Handle cl_host_wide_int. 15825 (get_option_state): Handle CLVC_SIZE. 15826 (control_warning_option): Same. 15827 * opts.c (common_handle_option): Change function argument to 15828 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and 15829 OPT_Wvla_larger_than_. 15830 * opts.h (enum cl_var_type): Add an enumerator. 15831 * stor-layout.c (layout_decl): Print a more meaningful warning. 15832 * toplev.c (output_stack_usage): Adjust. 15833 158342018-07-20 Qing Zhao <qing.zhao@oracle.com> 15835 15836 * builtins.c (expand_builtin_memcmp): Delete the last parameter for 15837 call to inline_expand_builtin_string_cmp. 15838 (expand_builtin_strcmp): Likewise. 15839 (expand_builtin_strncmp): Likewise. 15840 (inline_string_cmp): Delete the last parameter, change char_type_node 15841 to unsigned_char_type_node for strcmp/strncmp, add conversions to the 15842 two operands. 15843 (inline_expand_builtin_string_cmp): Delete the last parameter, give up 15844 the inlining expansion on target where the type of the call has same or 15845 narrower precision than unsigned char. 15846 158472018-07-20 David Malcolm <dmalcolm@redhat.com> 15848 15849 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o. 15850 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME. 15851 * common.opt (fsave-optimization-record): New option. 15852 * coretypes.h (struct kv_pair): Move here from dumpfile.c. 15853 * doc/invoke.texi (-fsave-optimization-record): New option. 15854 * dumpfile.c: Include "optinfo-emit-json.h". 15855 (struct kv_pair): Move to coretypes.h. 15856 (optgroup_options): Make non-static. 15857 (dump_context::end_scope): Call 15858 optimization_records_maybe_pop_dump_scope. 15859 * dumpfile.h (optgroup_options): New decl. 15860 * json.cc: New file. 15861 * json.h: New file. 15862 * optinfo-emit-json.cc: New file. 15863 * optinfo-emit-json.h: New file. 15864 * optinfo.cc: Include "optinfo-emit-json.h". 15865 (optinfo::emit): Call optimization_records_maybe_record_optinfo. 15866 (optinfo_enabled_p): Check optimization_records_enabled_p. 15867 (optinfo_wants_inlining_info_p): Likewise. 15868 * optinfo.h: Update comment. 15869 * profile-count.c (profile_quality_as_string): New function. 15870 * profile-count.h (profile_quality_as_string): New decl. 15871 (profile_count::quality): New accessor. 15872 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests 15873 and optinfo_emit_json_cc_tests. 15874 * selftest.h (selftest::json_cc_tests): New decl. 15875 (selftest::optinfo_emit_json_cc_tests): New decl. 15876 * toplev.c: Include "optinfo-emit-json.h". 15877 (compile_file): Call optimization_records_finish. 15878 (do_compile): Call optimization_records_start. 15879 * tree-ssa-live.c: Include optinfo.h. 15880 (remove_unused_scope_block_p): Retain inlining information if 15881 optinfo_wants_inlining_info_p returns true. 15882 158832018-07-20 Richard Biener <rguenther@suse.de> 15884 15885 PR debug/86585 15886 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p 15887 to cover -flto-partition=none. 15888 158892018-07-20 Martin Liska <mliska@suse.cz> 15890 15891 * tree.h (DECL_LOCATION_RANGE): Remove unused macro. 15892 (get_decl_source_range): Remove unused function. 15893 158942018-07-20 Richard Biener <rguenther@suse.de> 15895 15896 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member. 15897 (struct vn_phi_s): Likewise. 15898 (struct vn_reference_s): Likewise. 15899 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut 15900 for searching the slot of an entry known to be in the hash itself. 15901 (vn_phi_hasher::equal): Likewise. 15902 (vn_reference_hasher::equal): Likewise. 15903 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New 15904 globals. 15905 (optimistic_info, current_info): Remove, keeping only valid_info. 15906 (vn_reference_lookup_1): Remove fallback lookup. 15907 (vn_reference_lookup_2): Likewise. 15908 (vn_nary_op_lookup_1): Likewise. 15909 (vn_phi_lookup): Likewise. 15910 (vn_nary_build_or_lookup_1): Make sure to not chain the built 15911 hash element. 15912 (vn_reference_insert): Adjust, chain the inserted hash element 15913 at last_inserted_ref. 15914 (vn_reference_insert_pieces): Likewise. 15915 (visit_reference_op_call): Likewise. 15916 (vn_nary_op_insert_into): Chain the inserted hash element at 15917 last_inserted_nary. 15918 (vn_nary_op_insert_pieces): Adjust. 15919 (vn_nary_op_insert): Likewise. 15920 (vn_nary_op_insert_stmt): Likewise. 15921 (vn_phi_insert): Adjust, chain the inserted hash element at 15922 last_inserted_phi. 15923 (process_scc): Remove clearing and copying the optimistic 15924 table. Instead remove elements inserted during an optimistic 15925 iteration from the single table we maintain. 15926 (init_scc_vn): Adjust. 15927 (free_scc_vn): Likewise. 15928 (sccvn_dom_walker::record_cond): Likewise. 15929 (sccvn_dom_walker::after_dom_children): Likewise. 15930 159312018-07-19 Martin Sebor <msebor@redhat.com> 15932 15933 PR tree-optimization/84047 15934 PR tree-optimization/83776 15935 * tree-vrp.c (vrp_prop::check_mem_ref): New function. 15936 (check_array_bounds): Call it. 15937 159382018-07-19 Martin Sebor <msebor@redhat.com> 15939 15940 * align.h (align_flags): Use member initialization. 15941 159422018-07-19 David Malcolm <dmalcolm@redhat.com> 15943 15944 * Makefile.in (OBJS): Add optinfo.o. 15945 * coretypes.h (class symtab_node): New forward decl. 15946 (struct cgraph_node): New forward decl. 15947 (class varpool_node): New forward decl. 15948 * dump-context.h: New file. 15949 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h", 15950 "tree-pass.h". 15951 (refresh_dumps_are_enabled): Use optinfo_enabled_p. 15952 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed. 15953 (set_alt_dump_file): Likewise. 15954 (dump_context::~dump_context): New dtor. 15955 (dump_gimple_stmt): Move implementation to... 15956 (dump_context::dump_gimple_stmt): ...this new member function. 15957 Add the stmt to any pending optinfo, creating one if need be. 15958 (dump_gimple_stmt_loc): Move implementation to... 15959 (dump_context::dump_gimple_stmt_loc): ...this new member function. 15960 Start a new optinfo and add the stmt to it. 15961 (dump_gimple_expr): Move implementation to... 15962 (dump_context::dump_gimple_expr): ...this new member function. 15963 Add the stmt to any pending optinfo, creating one if need be. 15964 (dump_gimple_expr_loc): Move implementation to... 15965 (dump_context::dump_gimple_expr_loc): ...this new member function. 15966 Start a new optinfo and add the stmt to it. 15967 (dump_generic_expr): Move implementation to... 15968 (dump_context::dump_generic_expr): ...this new member function. 15969 Add the tree to any pending optinfo, creating one if need be. 15970 (dump_generic_expr_loc): Move implementation to... 15971 (dump_context::dump_generic_expr_loc): ...this new member 15972 function. Add the tree to any pending optinfo, creating one if 15973 need be. 15974 (dump_printf): Move implementation to... 15975 (dump_context::dump_printf_va): ...this new member function. Add 15976 the text to any pending optinfo, creating one if need be. 15977 (dump_printf_loc): Move implementation to... 15978 (dump_context::dump_printf_loc_va): ...this new member function. 15979 Start a new optinfo and add the stmt to it. 15980 (dump_dec): Move implementation to... 15981 (dump_context::dump_dec): ...this new member function. Add the 15982 value to any pending optinfo, creating one if need be. 15983 (dump_context::dump_symtab_node): New member function. 15984 (dump_context::get_scope_depth): New member function. 15985 (dump_context::begin_scope): New member function. 15986 (dump_context::end_scope): New member function. 15987 (dump_context::ensure_pending_optinfo): New member function. 15988 (dump_context::begin_next_optinfo): New member function. 15989 (dump_context::end_any_optinfo): New member function. 15990 (dump_context::s_current): New global. 15991 (dump_context::s_default): New global. 15992 (dump_scope_depth): Delete global. 15993 (dumpfile_ensure_any_optinfo_are_flushed): New function. 15994 (dump_symtab_node): New function. 15995 (get_dump_scope_depth): Reimplement in terms of dump_context. 15996 (dump_begin_scope): Likewise. 15997 (dump_end_scope): Likewise. 15998 (selftest::temp_dump_context::temp_dump_context): New ctor. 15999 (selftest::temp_dump_context::~temp_dump_context): New dtor. 16000 (selftest::verify_item): New function. 16001 (ASSERT_IS_TEXT): New macro. 16002 (ASSERT_IS_TREE): New macro. 16003 (ASSERT_IS_GIMPLE): New macro. 16004 (selftest::test_capture_of_dump_calls): New test. 16005 (selftest::dumpfile_c_tests): Call it. 16006 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block) 16007 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc) 16008 (dump_gimple_stmt, dump_dec): Gather these related decls and add a 16009 descriptive comment. 16010 (dump_function, print_combine_total_stats, enable_rtl_dump_file) 16011 (dump_node, dump_bb): Move these unrelated decls. 16012 (class dump_manager): Add leading comment. 16013 * optinfo.cc: New file. 16014 * optinfo.h: New file. 16015 160162018-07-19 Michael Collison <michael.collison@arm.com> 16017 Richard Henderson <rth@redhat.com> 16018 16019 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns. 16020 (subti): Handle op1 zero. 16021 (subvti4, usub4ti4): New. 16022 (*sub<GPI>3_compare1_imm): New. 16023 (sub<GPI>3_carryinCV): New. 16024 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New. 16025 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New. 16026 160272018-07-19 Michael Collison <michael.collison@arm.com> 16028 Richard Henderson <rth@redhat.com> 16029 16030 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New. 16031 (addti3): Create simpler code if low part is already known to be 0. 16032 (addvti4, uaddvti4): New. 16033 (*add<GPI>3_compareC_cconly_imm): New. 16034 (*add<GPI>3_compareC_cconly): New. 16035 (*add<GPI>3_compareC_imm): New. 16036 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not 16037 handle constants within this pattern.. 16038 (*add<GPI>3_compareV_cconly_imm): New. 16039 (*add<GPI>3_compareV_cconly): New. 16040 (*add<GPI>3_compareV_imm): New. 16041 (add<GPI>3_compareV): New. 16042 (add<GPI>3_carryinC, add<GPI>3_carryinV): New. 16043 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New. 16044 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New. 16045 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator 16046 with 'comparison' operator. 16047 (*add<GPI>3_compareV_cconly_imm): Ditto. 16048 (*add<GPI>3_compareV_cconly): Ditto. 16049 (*add<GPI>3_compareV_imm): Ditto. 16050 (add<GPI>3_compareV): Ditto. 16051 (add<mode>3_carryinC): Ditto. 16052 (*add<mode>3_carryinC_zero): Ditto. 16053 (*add<mode>3_carryinC): Ditto. 16054 (add<mode>3_carryinV): Ditto. 16055 (*add<mode>3_carryinV_zero): Ditto. 16056 (*add<mode>3_carryinV): Ditto. 16057 160582018-07-19 Michael Collison <michael.collison@arm.com> 16059 Richard Henderson <rth@redhat.com> 16060 16061 * config/aarch64/aarch64-modes.def (CC_V): New. 16062 * config/aarch64/aarch64-protos.h 16063 (aarch64_addti_scratch_regs): Declare 16064 (aarch64_subvti_scratch_regs): Declare. 16065 (aarch64_expand_subvti): Declare. 16066 (aarch64_gen_unlikely_cbranch): Declare 16067 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test 16068 for signed overflow using CC_Vmode. 16069 (aarch64_get_condition_code_1): Handle CC_Vmode. 16070 (aarch64_gen_unlikely_cbranch): New function. 16071 (aarch64_addti_scratch_regs): New function. 16072 (aarch64_subvti_scratch_regs): New function. 16073 (aarch64_expand_subvti): New function. 16074 160752018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com> 16076 16077 * config/aarch64/aarch64-option-extensions.def: New entry for profile 16078 extension. 16079 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New. 16080 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile 16081 extension. 16082 160832018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com> 16084 16085 PR target/83009 16086 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make 16087 address check not strict. 16088 160892018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com> 16090 16091 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace 16092 Umq with Umn. 16093 (store_pair_lanes<mode>): Likewise. 16094 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new 16095 enum value 'ADDR_QUERY_LDP_STP_N'. 16096 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise. 16097 (aarch64_print_address_internal): Add declaration. 16098 (aarch64_print_ldpstp_address): Remove. 16099 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'. 16100 (aarch64_print_operand): Change printing of 'y'. 16101 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use 16102 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use 16103 'true' rather than '1'. 16104 * config/aarch64/constraints.md (Uml): Likewise. 16105 (Uml): Rename to Umn. 16106 (Umq): Remove. 16107 161082018-07-19 Richard Biener <rguenther@suse.de> 16109 16110 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member 16111 a trailing array. 16112 * tree-ssa-sccvn.c: Remove alloc-pool.h use. 16113 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method. 16114 (vn_reference_hasher): Likewise. 16115 (struct vn_tables_s): Remove obstack and alloc-pool members. 16116 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks. 16117 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack. 16118 (vn_reference_insert): Allocate from obstack instead of from alloc-pool. 16119 (vn_reference_insert_pieces): Likewise. 16120 (alloc_vn_nary_op_noinit): Adjust. 16121 (vn_nary_op_insert_stmt): Allocate phiargs in-place. 16122 (vn_phi_eq): Adjust. 16123 (shared_lookup_phiargs): Remove. 16124 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack. 16125 (vn_phi_insert): Allocate from obstack instead of from alloc-pool. 16126 (visit_reference_op_call): Likewise. 16127 (copy_nary, copy_phi, copy_reference): Remove. 16128 (process_scc): Rewind the obstack when iterating. Do not 16129 copy the elements to valid_info but just move them from one 16130 hashtable to the other. 16131 (allocate_vn_table): Adjust. 16132 (free_vn_table): Likewise. 16133 (init_scc_vn): Likewise. 16134 (free_scc_vn): Likewise. 16135 161362018-07-19 H.J. Lu <hongjiu.lu@intel.com> 16137 16138 PR target/86560 16139 * config/i386/i386.c (rest_of_insert_endbranch): Lookup 16140 indirect_return as function type attribute. 16141 (ix86_attribute_table): Change indirect_return to function 16142 type attribute. 16143 * doc/extend.texi: Update indirect_return attribute. 16144 161452018-07-19 Aldy Hernandez <aldyh@redhat.com> 16146 16147 * wide-int.h (widest2_int): New. 16148 * gimple-fold.c (arith_overflowed_p): Use it. 16149 * tree.h (widest2_int_cst): New. 16150 * tree-vrp.c (wide_int_binop_overflow): Rename from 16151 vrp_int_const_binop. 16152 Rewrite to work on trees. 16153 (extract_range_from_multiplicative_op_1): Abstract code to... 16154 (wide_int_range_min_max): ...here. 16155 (wide_int_range_cross_product): ...and here. 16156 (extract_range_from_binary_expr_1): Abstract overflow code to... 16157 (wide_int_range_mult_wrapping): ...here. 16158 * tree-vrp.h (wide_int_range_cross_product): New. 16159 (wide_int_range_mult_wrapping): New. 16160 161612018-07-19 Andrew Senkevich <andrew.senkevich@intel.com> 16162 Julia Koval <julia.koval@intel.com> 16163 16164 * config/i386/x86-tune-costs.h (skylake_memcpy, 16165 skylake_memset): Replace rep_prefix with unrolling for size 512. 16166 161672018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org> 16168 16169 PR middle-end/86544 16170 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle 16171 comparision with EQ_EXPR in last stmt. 16172 161732018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org> 16174 16175 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename 16176 this subsection to "PowerPC AltiVec/VSX Built-in Functions". 16177 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection 16178 previously known as "PowerPC AltiVec Built-in Functions". Move 16179 some material to new subsubsections "PowerPC AltiVec Built-in 16180 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on 16181 ISA 2.07". 16182 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection. 16183 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise. 16184 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise. 16185 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise. 16186 161872018-07-18 Richard Biener <rguenther@suse.de> 16188 16189 PR tree-optimization/86557 16190 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle 16191 EXACT_DIV_EXPR. 16192 161932018-07-18 Ilya Leoshkevich <iii@linux.ibm.com> 16194 16195 * config/s390/s390.c (s390_function_profiler): Generate CFI. 16196 161972018-07-17 Jeff Law <law@redhat.com> 16198 16199 * config/arm/arm.c (get_label_padding): Update for recent 16200 changes to label_to_alignment. 16201 16202 PR tree-optimization/86010 16203 * tree-ssa-dse.c (compute_trims): Fix typo/thinko. 16204 16205 * config/mips/mips.c (vr4130_align_insns): Update for recent 16206 changes to label_to_alignment. 16207 16208 * config/frv/frv.c (frv_label_align): Update for recent changes 16209 to label_to_alignment. 16210 16211 * config/nios2/nios2.c (nios2_label_align): Update for recent 16212 changes which dropped ALIGN_LABELS_LOG. 16213 162142018-07-17 Andreas Schwab <schwab@linux-m68k.org> 16215 16216 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart) 16217 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT. 16218 162192018-07-17 Claudiu Zissulescu <claziss@synopsys.com> 16220 16221 * config/arc/arc.c (arc_label_align): Use align_labels instead of 16222 deprecated align_labels_log. 16223 162242018-07-17 Richard Biener <rguenther@suse.de> 16225 16226 PR lto/86456 16227 * dwarf2out.c (init_sections_and_labels): Always generate 16228 a debug_line_str_section for early LTO debug. 16229 (dwarf2out_finish): Reset debug_line_str_hash output early. 16230 Bump counter for extra dwarf5 .debug_loc labels to not conflict 16231 with fat LTO part. 16232 (dwarf2out_early_finish): Output debug_line_str. 16233 162342018-07-17 Robin Dapp <rdapp@linux.ibm.com> 16235 16236 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with 16237 index register on z196 or later. 16238 162392018-07-17 Robin Dapp <rdapp@linux.ibm.com> 16240 16241 * config/s390/s390.c (s390_default_align): Set default function 16242 alignment to 16. 16243 (s390_override_options_after_change): Call s390_default align. 16244 (s390_option_override_internal): Call s390_default_align. 16245 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define. 16246 162472018-07-17 Jakub Jelinek <jakub@redhat.com> 16248 16249 PR middle-end/86542 16250 * omp-low.c (create_task_copyfn): Copy over also fields corresponding 16251 to _looptemp_ clauses, other than the first two. 16252 162532018-07-17 Martin Liska <mliska@suse.cz> 16254 16255 * opts.c: Do not enable OPT_falign_* for -Os. 16256 162572018-07-17 Martin Liska <mliska@suse.cz> 16258 16259 * align.h (MAX_CODE_ALIGN): New. 16260 (MAX_CODE_ALIGN_VALUE): New. 16261 * common/config/i386/i386-common.c (ix86_handle_option): 16262 (MAX_CODE_ALIGN): Moved to align.h. 16263 * final.c (MAX_CODE_ALIGN): Likewise. 16264 * opts.c (parse_and_check_align_values): 16265 (MAX_CODE_ALIGN): Likewise. 16266 (MAX_CODE_ALIGN_VALUE): Likewise. 16267 162682018-07-17 Martin Liska <mliska@suse.cz> 16269 16270 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing 16271 in order to fulfil coding style. 16272 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise. 16273 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise. 16274 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise. 16275 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise. 16276 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise. 16277 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise. 16278 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise. 16279 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise. 16280 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise. 16281 162822018-07-17 Martin Liska <mliska@suse.cz> 16283 16284 * align.h: New file. 16285 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions 16286 directly. 16287 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type 16288 align_flags of label_to_alignment. 16289 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into 16290 align_flags class. 16291 * config/m68k/m68k.c: Do not use removed align_labels_value and 16292 align_loops_value. 16293 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class. 16294 (LOOP_ALIGN): Likewise. 16295 (LABEL_ALIGN): Likewise. 16296 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP): 16297 Remove not used macro. 16298 (rs6000_loop_align): Change return type to align_flags. 16299 (rs6000_loop_align_max_skip): Remove. 16300 * config/rs6000/rs6000-protos.h (rs6000_loop_align): 16301 Change return type to align_flags. 16302 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP): 16303 Remove not used macro. 16304 (rs6000_loop_align): Change return type to align_flags. 16305 (rs6000_loop_align_max_skip): Remove. 16306 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values 16307 * config/rx/rx-protos.h (rx_align_for_label): Make it 16308 static function. 16309 * config/rx/rx.c (rx_align_for_label): Change return type 16310 to align_flags. 16311 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP 16312 macro definitions. 16313 into align_flags class. 16314 (LABEL_ALIGN): Likewise. 16315 (LOOP_ALIGN): Likewise. 16316 * config/s390/s390.c (s390_label_align): Use align_flags 16317 class member. 16318 (s390_asm_output_function_label): Likewise. 16319 * config/sh/sh.c (sh_override_options_after_change): 16320 Use align_flags class directly without macros. 16321 (find_barrier): Likewise. 16322 (barrier_align): Likewise. 16323 (sh_loop_align): Likewise. 16324 * config/spu/spu.c (spu_option_override): 16325 Use align_flags_tuple::get_value instead of removed macros. 16326 (spu_sched_init): Likewise. 16327 * config/spu/spu.h (GTY): Likewise. 16328 * config/visium/visium.c (visium_option_override): 16329 Set "8" as default secondary alignment. 16330 * config/visium/visium.h (SUBALIGN_LOG): Define to 3 16331 in order to guarantee secondary alignment of 8. 16332 * coretypes.h: Include align.h header file. 16333 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP, 16334 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP 16335 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros. 16336 * doc/tm.texi.in: Likewise. 16337 * final.c (struct label_alignment): Remove not used structure. 16338 (LABEL_ALIGN): Change type to align_flags. 16339 (LOOP_ALIGN): Likewise. 16340 (JUMP_ALIGN): Likewise. 16341 (default_loop_align_max_skip): Remove. 16342 (default_label_align_max_skip): Likewise. 16343 (default_jump_align_max_skip): Likewise. 16344 (default_label_align_after_barrier_max_skip): 16345 (LABEL_TO_ALIGNMENT): Change to access label_align vector. 16346 (LABEL_TO_MAX_SKIP): Remove. 16347 (label_to_alignment): Return align_flags type instead of integer. 16348 (label_to_max_skip): Remove. 16349 (align_fuzz): Use align_flags type. 16350 (compute_alignments): Use align_flags type and use align_flags::max 16351 to combine multiple alignments. 16352 (grow_label_align): Grow vec instead of C array. 16353 (update_alignments): Assign just LABEL_TO_ALIGNMENT. 16354 (shorten_branches): Use align_flags type and use align_flags::max 16355 to combine multiple alignments. 16356 (final_scan_insn_1): Remove usage of secondary alignment that comes 16357 from label alignment, but instead use proper secondary alignment 16358 which is computed in grow_label_align. 16359 * flags.h (struct align_flags_tuple): Move to align.h. 16360 (struct align_flags): Likewise. 16361 (state_align_loops): Rename to align_loops. 16362 (state_align_jumps): Rename to align_jumps. 16363 (state_align_labels): Rename to align_labels. 16364 (state_align_functions): Rename to align_functions. 16365 (align_loops_log): Remove. 16366 (align_jumps_log): Remove. 16367 (align_labels_log): Remove. 16368 (align_functions_log): Remove. 16369 (align_loops_max_skip): Remove. 16370 (align_jumps_max_skip): Remove. 16371 (align_labels_max_skip): Remove. 16372 (align_functions_max_skip): Remove. 16373 (align_loops_value): Remove. 16374 (align_jumps_value): Remove. 16375 (align_labels_value): Remove. 16376 (align_functions_value): Remove. 16377 * output.h (label_to_alignment): Change return type to align_flags. 16378 (label_to_max_skip): Remove. 16379 * target.def: Remove loop_align_max_skip, label_align_max_skip, 16380 jump_align_max_skip macros. 16381 * targhooks.h (default_loop_align_max_skip): Remove. 16382 (default_label_align_max_skip): Likewise. 16383 (default_jump_align_max_skip): Likewise. 16384 (default_label_align_after_barrier_max_skip): Remove. 16385 * toplev.c (read_log_maxskip): Use ::normalize function. 16386 (parse_N_M): Remove not used argument and also call ::normalize. 16387 (parse_alignment_opts): Do not pass unused arguments. 16388 * varasm.c (assemble_start_function): Use directly align_functions 16389 instead of removed macros. 16390 * system.h: Do not poison removed macros. 16391 163922018-07-17 Jakub Jelinek <jakub@redhat.com> 16393 16394 PR middle-end/86539 16395 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init 16396 and cond temporaries don't have reference type if iterator has 16397 pointer type. For init use &for_pre_body instead of pre_p if 16398 for_pre_body is non-empty. 16399 164002018-07-16 Segher Boessenkool <segher@kernel.crashing.org> 16401 16402 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of 16403 double-double modes to SFmode directly directly. 16404 (trunc<mode>sf2_fprs): Delete. 16405 164062018-07-16 Segher Boessenkool <segher@kernel.crashing.org> 16407 16408 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names 16409 for conversions between IFmode and the decimal floating point modes. 16410 (init_float128_ieee): Use the correct names for conversions between 16411 KFmode and the decimal floating point modes. 16412 164132018-07-16 Segher Boessenkool <segher@kernel.crashing.org> 16414 16415 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names 16416 for the conversions between TDmode and IFmode. 16417 (init_float128_ieee): Use more correct names for the conversions 16418 between TDmode and KFmode. 16419 164202018-07-16 Jakub Jelinek <jakub@redhat.com> 16421 16422 PR tree-optimization/86526 16423 * builtins.c (expand_builtin_memcmp): Formatting fixes. 16424 (inline_expand_builtin_string_cmp): Likewise. 16425 (inline_string_cmp): Likewise. Use c_readstr instead of 16426 builtin_memcpy_read_str. Add unit_mode temporary. 16427 164282018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de> 16429 16430 PR middle-end/86528 16431 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST. 16432 * expr.c (string_constant): Fix the element size of ARRAY_TYPE. 16433 164342018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org> 16435 16436 * doc/extend.texi (PowerPC AltiVec Built-in Functions): 16437 Alphabetize prototypes of built-in functions, separating out 16438 built-in functions that are listed in this section but should be 16439 described elsewhere. 16440 164412018-07-16 Uros Bizjak <ubizjak@gmail.com> 16442 16443 PR target/86511 16444 * expmed.c (emit_store_flag): Do not emit setcc followed by a 16445 conditional move when trapping comparison was split to a 16446 non-trapping one (and vice versa). 16447 164482018-07-16 Ilya Leoshkevich <iii@linux.ibm.com> 16449 16450 * config/s390/s390.c (s390_function_profiler): Generate nops 16451 instead of profiler call sequences. 16452 * config/s390/s390.opt: Add the new option. 16453 164542018-07-16 Ilya Leoshkevich <iii@linux.ibm.com> 16455 16456 * config/s390/s390.c (s390_function_profiler): Generate 16457 __mcount_loc section. 16458 * config/s390/s390.opt: Add the new option. 16459 164602018-07-16 Ilya Leoshkevich <iii@linux.ibm.com> 16461 16462 * common.opt: Add the new warning. 16463 * config/s390/s390.c (s390_function_profiler): Emit "brasl 16464 %r0,__fentry__" when -mfentry is specified. 16465 (s390_option_override_internal): Disallow -mfentry for 31-bit 16466 CPUs. 16467 * config/s390/s390.opt: Add the new option. 16468 164692018-07-16 Richard Biener <rguenther@suse.de> 16470 16471 PR lto/86523 16472 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents 16473 for function-local FUNCTION_DECL and RESULT_DECL immediately. 16474 164752018-07-16 Martin Liska <mliska@suse.cz> 16476 16477 PR ipa/86529 16478 * ipa-pure-const.c (malloc_candidate_p): Revert ::get 16479 to ::get_create. 16480 164812017-07-16 Claudiu Zissulescu <claziss@synopsys.com> 16482 16483 * config/arc/arcHS.md: Update ARCHS scheduling rules. 16484 164852017-07-16 Claudiu Zissulescu <claziss@synopsys.com> 16486 16487 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters 16488 for ARCHS4x. 16489 * config/arc/arc-cpus.def (hs4x): New cpu. 16490 (hs4xd): Likewise. 16491 * config/arc/arc-tables.opt: Regenerate. 16492 * config/arc/arc.c (arc_sched_issue_rate): New function. 16493 (TARGET_SCHED_ISSUE_RATE): Define. 16494 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise. 16495 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv, 16496 fpu_cvt. 16497 (attr tune): Add ARCHS4x tune values. 16498 (attr tune_dspmpy): Define. 16499 (*tst): Correct instruction type. 16500 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus. 16501 * config/arc/arcHS4x.md: New file. 16502 * config/arc/fpu.md: Update instruction type attributes. 16503 * config/arc/t-multilib: Regenerate. 16504 165052018-07-16 Tom de Vries <tdevries@suse.de> 16506 16507 PR debug/86455 16508 * var-tracking.c (vt_initialize): Fix pre_dec handling. 16509 165102018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com> 16511 16512 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add 16513 early clobber. 16514 165152018-07-16 Eric Botcazou <ebotcazou@adacore.com> 16516 16517 PR tree-optimization/86514 16518 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a 16519 conversion to a boolean type from a type with greater precision. 16520 165212018-07-16 Tom de Vries <tdevries@suse.de> 16522 16523 * var-tracking.c (vt_initialize): Print adjusted insn slim if 16524 dump_flags request TDF_SLIM. 16525 165262018-07-16 Aldy Hernandez <aldyh@redhat.com> 16527 16528 * fold-const.c (int_const_binop_1): Abstract... 16529 (wide_int_binop): ...wide int code here. 16530 (poly_int_binop): ...poly int code here. 16531 Abstract the rest of int_const_binop_1 into int_const_binop. 16532 * fold-const.h (wide_int_binop): New. 16533 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop. 16534 Remove useless PLUS/MINUS_EXPR case. 16535 (zero_nonzero_bits_from_vr): Move wide int code... 16536 (zero_nonzero_bits_from_bounds): ...here. 16537 (extract_range_from_binary_expr_1): Move mask optimization code... 16538 (range_easy_mask_min_max): ...here. 16539 * tree-vrp.h (zero_nonzero_bits_from_bounds): New. 16540 (range_easy_mask_min_max): New. 16541 165422018-07-15 Jeff Law <law@redhat.com> 16543 16544 PR target/85993 16545 * config/sh/sh.c (output_mi_thunk): Remove dead conditional 16546 block. 16547 165482018-07-14 Jim Wilson <jimw@sifive.com> 16549 16550 * config/riscv/linux.h (TARGET_ASM_FILE_END): New. 16551 165522018-07-14 Paul Koning <ni1d@arrl.net> 16553 16554 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes. 16555 165562018-07-13 Jan Hubicka <hubicka@ucw.cz> 16557 16558 * lto-streamer-out.c (copy_function_or_variable): Dump info about 16559 copying section. 16560 165612018-07-13 Bill Schmidt <wschmidt@linux.ibm.com> 16562 Steve Munroe <munroesj52@gmail.com> 16563 16564 * config/rs6000/emmintrin.h (_mm_and_si128): New function. 16565 (_mm_andnot_si128): Likewise. 16566 (_mm_or_si128): Likewise. 16567 (_mm_xor_si128): Likewise. 16568 165692018-07-13 Qing Zhao <qing.zhao@oracle.com> 16570 16571 PR middle-end/78809 16572 * builtins.c (expand_builtin_memcmp): Inline the calls first 16573 when result_eq is false. 16574 (expand_builtin_strcmp): Inline the calls first. 16575 (expand_builtin_strncmp): Likewise. 16576 (inline_string_cmp): New routine. Expand a string compare 16577 call by using a sequence of char comparison. 16578 (inline_expand_builtin_string_cmp): New routine. Inline expansion 16579 a call to str(n)cmp/memcmp. 16580 * doc/invoke.texi (--param builtin-string-cmp-inline-length): 16581 New option. 16582 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New. 16583 165842018-07-13 Richard Earnshaw <rearnsha@arm.com> 16585 16586 * config/arm/driver-arm.c: Include arm-native.h. 16587 (host_detect_local_cpu): Use auto-generated data tables. 16588 (vendors, arm_cpu_table): Delete. Move part information to ... 16589 * config/arm/arm-cpus.in: ... here. 16590 * config/arm/parsecpu.awk (gen_native): New function. 16591 (vendor, part): New CPU fields. 16592 (END): Add support for building the native CPU detection tables. 16593 * config/arm/t-arm (arm-native.h): Add build rule. 16594 (driver-arm.o): Add dependency on arm-native.h. 16595 165962018-07-13 Richard Biener <rguenther@suse.de> 16597 16598 PR middle-end/85974 16599 * match.pd (addr1 - addr2): Allow either of the operand to 16600 have a conversion. 16601 166022018-07-13 Tom de Vries <tdevries@suse.de> 16603 16604 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated 16605 in remap_ssa_name. 16606 166072018-07-13 Jackson Woodruff <jackson.woodruff@arm.com> 16608 16609 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use 16610 arrays instead of numbered variables. 16611 166122018-07-13 Eric Botcazou <ebotcazou@adacore.com> 16613 16614 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete. 16615 * config/sparc/sparc.c (sparc_compute_frame_size): Make static. 16616 166172018-07-13 Richard Biener <rguenther@suse.de> 16618 16619 PR debug/86452 16620 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for 16621 instead of get_context_die. 16622 166232018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org> 16624 Richard Biener <rguenther@suse.de> 16625 16626 PR middle-end/86489 16627 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check 16628 that the loop latch destination where phi is defined. 16629 166302018-07-12 Kito Cheng <kito.cheng@gmail.com> 16631 16632 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE. 16633 (riscv_expand_epilogue): Add assertion to check interrupt mode. 16634 (riscv_set_current_function): Extract getting interrupt type to new 16635 function. 16636 (riscv_get_interrupt_type): New function. 16637 (riscv_merge_decl_attributes): New function, checking interrupt type is 16638 same. 16639 (TARGET_MERGE_DECL_ATTRIBUTES): Define. 16640 166412018-07-12 Paul Koning <ni1d@arrl.net> 16642 16643 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set 16644 directive. 16645 166462018-07-12 Paul Koning <ni1d@arrl.net> 16647 16648 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until 16649 zero reference, add doloop_end instead. 16650 * doc/md.texi (decrement_and_branch_until_zero): Remove. 16651 (Looping patterns): Remove decrement_and_branch_until_zero. Add 16652 detail for doloop_end. 16653 166542018-07-12 Martin Sebor <msebor@redhat.com> 16655 16656 PR c/86453 16657 * attribs.c (decl_attributes): Reject conflicting attributes before 16658 calling attribute handlers. 16659 166602018-07-12 Jan Hubicka <hubicka@ucw.cz> 16661 16662 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART 16663 parameter. 16664 (gcc::dump_manager::get_dump_file_name): likewise. 16665 (dump_begin): Likewise. 16666 * dumpfile.h (dump_begin): Update prototype. 16667 (gcc::dump_manager::get_dump_file_name, 16668 gcc::dump_manager::get_dump_file_name): Update prototype. 16669 166702018-07-12 Richard Sandiford <richard.sandiford@linaro.org> 16671 16672 * internal-fn.h (vectorizable_internal_fn_p): New function. 16673 * tree-vect-slp.c (compatible_calls_p): Likewise. 16674 (vect_build_slp_tree_1): Remove nops argument. Handle calls 16675 to internal functions. 16676 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1. 16677 166782018-07-12 Richard Sandiford <richard.sandiford@linaro.org> 16679 16680 * fold-const.h (inverse_conditions_p): Declare. 16681 * fold-const.c (inverse_conditions_p): New function. 16682 * match.pd: Use inverse_conditions_p. Add folds of view_converts 16683 that test the inverse condition of a conditional internal function. 16684 * internal-fn.h (vectorized_internal_fn_supported_p): Declare. 16685 * internal-fn.c (internal_fn_mask_index): Handle conditional 16686 internal functions. 16687 (vectorized_internal_fn_supported_p): New function. 16688 * tree-if-conv.c: Include internal-fn.h and fold-const.h. 16689 (any_pred_load_store): Replace with... 16690 (need_to_predicate): ...this new variable. 16691 (redundant_ssa_names): New variable. 16692 (ifcvt_can_use_mask_load_store): Move initial checks to... 16693 (ifcvt_can_predicate): ...this new function. Handle tree codes 16694 for which a conditional internal function exists. 16695 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate 16696 instead of ifcvt_can_use_mask_load_store. Update after variable 16697 name change. 16698 (predicate_load_or_store): New function, split out from 16699 predicate_mem_writes. 16700 (check_redundant_cond_expr): New function. 16701 (value_available_p): Likewise. 16702 (predicate_rhs_code): Likewise. 16703 (predicate_mem_writes): Rename to... 16704 (predicate_statements): ...this. Use predicate_load_or_store 16705 and predicate_rhs_code. 16706 (combine_blocks, tree_if_conversion): Update after above name changes. 16707 (ifcvt_local_dce): Handle redundant_ssa_names. 16708 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle 16709 general conditional functions. 16710 * tree-vect-stmts.c (vectorizable_call): Likewise. 16711 167122018-07-12 Richard Sandiford <richard.sandiford@linaro.org> 16713 Alan Hayward <alan.hayward@arm.com> 16714 David Sherwood <david.sherwood@arm.com> 16715 16716 * internal-fn.h (can_interpret_as_conditional_op_p): Declare. 16717 * internal-fn.c (can_interpret_as_conditional_op_p): New function. 16718 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional 16719 plus and minus and convert them into IFN_COND_FMA-based sequences. 16720 (convert_mult_to_fma): Handle conditional plus and minus. 16721 167222018-07-12 Richard Sandiford <richard.sandiford@linaro.org> 16723 16724 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document. 16725 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab) 16726 (cond_fnms_optab): New optabs. 16727 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New 16728 internal functions. 16729 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN. 16730 * internal-fn.h (get_conditional_internal_fn): Declare. 16731 (get_unconditional_internal_fn): Likewise. 16732 * internal-fn.c (cond_ternary_direct): New macro. 16733 (expand_cond_ternary_optab_fn): Likewise. 16734 (direct_cond_ternary_optab_supported_p): Likewise. 16735 (FOR_EACH_COND_FN_PAIR): Likewise. 16736 (get_conditional_internal_fn): New function. 16737 (get_unconditional_internal_fn): Likewise. 16738 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5. 16739 (gimple_match_op::gimple_match_op): Add a new overload for 5 16740 operands. 16741 (gimple_match_op::set_op): Likewise. 16742 (gimple_resimplify5): Declare. 16743 * genmatch.c (decision_tree::gen): Generate simplifications for 16744 5 operands. 16745 * gimple-match-head.c (gimple_simplify): Define an overload for 16746 5 operands. Handle calls with 5 arguments in the top-level overload. 16747 (convert_conditional_op): Handle conversions from unconditional 16748 internal functions to conditional ones. 16749 (gimple_resimplify5): New function. 16750 (build_call_internal): Pass a fifth operand. 16751 (maybe_push_res_to_seq): Likewise. 16752 (try_conditional_simplification): Try converting conditional 16753 internal functions to unconditional internal functions. 16754 Handle 3-operand unconditional forms. 16755 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists. 16756 Define ternary equivalents of the current rules for binary conditional 16757 internal functions. 16758 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle 16759 ternary operations. 16760 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS) 16761 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs. 16762 (optab): Handle them. 16763 (SVE_COND_FP_TERNARY): New int iterator. 16764 (sve_fmla_op, sve_fmad_op): New int attributes. 16765 * config/aarch64/aarch64-sve.md (cond_<optab><mode>) 16766 (*cond_<optab><mode>_2, *cond_<optab><mode_4) 16767 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns. 16768 167692018-07-12 Richard Sandiford <richard.sandiford@linaro.org> 16770 16771 * target.def (preferred_else_value): New target hook. 16772 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook. 16773 * doc/tm.texi: Regenerate. 16774 * targhooks.h (default_preferred_else_value): Declare. 16775 * targhooks.c (default_preferred_else_value): New function. 16776 * internal-fn.h (conditional_internal_fn_code): Declare. 16777 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro. 16778 (get_conditional_internal_fn): Use it. 16779 (conditional_internal_fn_code): New function. 16780 * gimple-match.h (gimple_match_cond): New struct. 16781 (gimple_match_op): Add a cond member function. 16782 (gimple_match_op::gimple_match_op): Update all forms to take a 16783 gimple_match_cond. 16784 * genmatch.c (expr::gen_transform): Use the same condition as res_op 16785 for the suboperation, but don't specify a particular else_value. 16786 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3) 16787 (visit_nary_op, visit_reference_op_load): Pass 16788 gimple_match_cond::UNCOND to the gimple_match_op constructor. 16789 * gimple-match-head.c: Include tree-eh.h 16790 (convert_conditional_op): New function. 16791 (maybe_resimplify_conditional_op): Likewise. 16792 (gimple_resimplify1): Call maybe_resimplify_conditional_op. 16793 (gimple_resimplify2): Likewise. 16794 (gimple_resimplify3): Likewise. 16795 (gimple_resimplify4): Likewise. 16796 (maybe_push_res_to_seq): Return null for conditional operations. 16797 (try_conditional_simplification): New function. 16798 (gimple_simplify): Call it. Pass conditions to the gimple_match_op 16799 constructor. 16800 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new 16801 IFN_COND_* call. 16802 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New 16803 function. 16804 (TARGET_PREFERRED_ELSE_VALUE): Redefine. 16805 168062018-07-12 Jan Hubicka <hubicka@ucw.cz> 16807 16808 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream 16809 DECL_FCONTEXT 16810 (hash_tree): Do not hash DECL_FCONTEXT 16811 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers): 16812 Do not stream DECL_FCONTEXT. 16813 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise. 16814 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT. 16815 168162018-07-12 Richard Biener <rguenther@suse.de> 16817 16818 PR debug/86462 16819 * dwarf2out.c (gen_block_die): Only output blocks when they have 16820 at least one !DECL_IGNORED_P variable. 16821 168222018-07-12 Richard Biener <rguenther@suse.de> 16823 16824 PR target/84829 16825 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC): 16826 Remove -mieee-fp handling. 16827 168282018-07-12 Richard Biener <rguenther@suse.de> 16829 16830 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus 16831 left-over from last patch. 16832 168332018-07-12 Jakub Jelinek <jakub@redhat.com> 16834 16835 PR tree-optimization/86492 16836 * gimple-ssa-store-merging.c 16837 (imm_store_chain_info::coalesce_immediate_stores): Call 16838 check_no_overlap even for the merge_overlapping case. Formatting fix. 16839 168402018-07-12 Richard Biener <rguenther@suse.de> 16841 16842 PR middle-end/86479 16843 * fold-const.c (fold_binary_op_with_conditional_arg): Do not 16844 move possibly trapping operations into the conditional. 16845 168462018-07-12 Richard Biener <rguenther@suse.de> 16847 16848 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove. 16849 (vn_lookup_simplify_result): Remove recursion limit applied 16850 here. 16851 (vn_nary_build_or_lookup_1): Adjust. 16852 (try_to_simplify): Likewise. 16853 * gimple-match-head.c (gimple_resimplify1): Instead apply one 16854 here. 16855 (gimple_resimplify2): Likewise. 16856 (gimple_resimplify3): Likewise. 16857 (gimple_resimplify4): Likewise. 16858 168592018-07-11 Jakub Jelinek <jakub@redhat.com> 16860 16861 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask): 16862 Use __mmask64 type instead of __mmask8 for __M argument. 16863 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64, 16864 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for 16865 __U argument. 16866 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of 16867 __mmask16 for __M argument. 16868 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4, 16869 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument 16870 to __mmask16 instead of __mmask8. 16871 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps, 16872 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps, 16873 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps, 16874 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type 16875 instead of __mmask16 for __U argument. 16876 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use 16877 __mmask16 instead of __mmask8 for __U argument. 16878 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for 16879 __U argument. 16880 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of 16881 __mmask16. 16882 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U 16883 argument. 16884 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for 16885 __U argument. 16886 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of 16887 __mmask16. 16888 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead 16889 of __mmask16. 16890 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for 16891 __U argument. 16892 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for 16893 __U argument. 16894 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for 16895 __U argument. 16896 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for 16897 __U argument. 16898 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask, 16899 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change 16900 return type as well as __M argument type and all casts from __mmask8 16901 to __mmask32. 16902 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask, 16903 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change 16904 return type as well as __M argument type and all casts from __mmask8 16905 to __mmask16. 16906 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask, 16907 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change 16908 return type as well as __M argument type and all casts from __mmask8 16909 to __mmask32. 16910 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask, 16911 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change 16912 return type as well as __M argument type and all casts from __mmask8 16913 to __mmask16. 16914 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32, 16915 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of 16916 __mmask16. 16917 169182018-07-11 Grazvydas Ignotas <notasas@gmail.com> 16919 16920 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask, 16921 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32 16922 for __U argument. 16923 169242018-07-11 Paul Koning <ni1d@arrl.net> 16925 16926 * doc/md.texi (define_subst): Document how multiple occurrences of 16927 the same argument in the replacement pattern are handled. 16928 169292018-07-11 Paul Koning <ni1d@arrl.net> 16930 16931 * doc/extend.texi (Common Variable Attributes): Move "mode" into 16932 alphabetical order. 16933 (Common Type Attributes): Add "mode" attribute. 16934 169352018-07-11 Jan Hubicka <hubicka@ucw.cz> 16936 16937 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not 16938 stream DECL_ORIGINAL_TYPE. 16939 (DFS::DFS_write_tree_body): Drop hack handling local external decls. 16940 (hash_tree): Do not walk DECL_ORIGINAL_TYPE. 16941 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers): 16942 Do not walk original type. 16943 * tree-streamer-out.c (streamer_write_chain): Drop hack handling 16944 external decls. 16945 (write_ts_decl_non_common_tree_pointers): Do not stream 16946 DECL_ORIGINAL_TYPE 16947 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE. 16948 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE. 16949 169502018-07-11 Aldy Hernandez <aldyh@redhat.com> 16951 16952 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump 16953 thread twice from the same starting edge. 16954 169552018-07-11 Aldy Hernandez <aldyh@redhat.com> 16956 16957 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to... 16958 * gimple.c (gimple_call_nonnull_result_p): ...here... 16959 (gimple_call_nonnull_arg): ...and here. 16960 * gimple.h (gimple_call_nonnull_result_p): New. 16961 (gimple_call_nonnull_arg): New. 16962 169632018-07-11 Richard Earnshaw <rearnsha@arm.com> 16964 16965 * config/arm/arm-cpus.in: Move information from fpu field of each 16966 cpu definition to the isa field. 16967 * config/arm/parsecpu.awk (fpu): Delete match rule. 16968 (gen_comm_data): Don't add bits from the CPU's FPU entry. 16969 169702018-07-11 Richard Biener <rguenther@suse.de> 16971 16972 PR debug/86457 16973 * dwarf2out.c (init_sections_and_labels): Use 16974 output_asm_line_debug_info consistently. 16975 (dwarf2out_early_finish): Likewise. 16976 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated 16977 type units. 16978 169792018-07-11 Richard Biener <rguenther@suse.de> 16980 16981 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): 16982 Rework father_bb setting in a way to avoid propagating constants 16983 multiple times on a loop body. 16984 169852018-07-10 Mark Wielaard <mark@klomp.org> 16986 16987 PR debug/86459 16988 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert. 16989 169902018-07-10 Richard Biener <rguenther@suse.de> 16991 16992 * hash-map.h (hash_map::iterator::operator*): Return 16993 references to key and value. 16994 169952018-07-10 Jakub Jelinek <jakub@redhat.com> 16996 16997 PR c++/86443 16998 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition 16999 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers, 17000 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between. 17001 (gimplify_omp_for): For composite loops, move outer 17002 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost 17003 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class 17004 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain 17005 TREE_LIST for both the original class iterator and the "last" helper 17006 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite 17007 loop, remember has_decl_expr from outer composite loops for the 17008 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT. 17009 170102018-07-09 Martin Sebor <msebor@redhat.com> 17011 17012 PR middle-end/77357 17013 PR middle-end/86428 17014 * builtins.c (c_strlen): Avoid out-of-bounds warnings when 17015 accessing implicitly initialized array elements. 17016 * expr.c (string_constant): Handle string initializers of 17017 character arrays within aggregates. 17018 * gimple-fold.c (fold_array_ctor_reference): Add argument. 17019 Store element offset. As a special case, handle zero size. 17020 (fold_nonarray_ctor_reference): Same. 17021 (fold_ctor_reference): Add argument. Store subobject offset. 17022 * gimple-fold.h (fold_ctor_reference): Add argument. 17023 170242018-07-09 Paul Koning <ni1d@arrl.net> 17025 17026 * config/pdp11/pdp11.c (pdp11_addr_cost): New function. 17027 (pdp11_insn_cost): New function. 17028 (pdp11_md_asm_adjust): New function. 17029 (TARGET_INVALID_WITHIN_DOLOOP): Define. 17030 (pdp11_rtx_costs): Update to match machine better. 17031 (output_addr_const_pdp11): Correct format mismatch warnings. 17032 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition. 17033 * config/pdp11/pdp11.md: General change to add base_cost and/or 17034 length attributes for use by new pdp11_insn_cost function. 17035 (MIN_BRANCH): Correct definition. 17036 (MIN_SOB): Ditto. 17037 (doloop_end): Use standard pattern name for looping pattern. 17038 (doloop_end_nocc): New. 17039 (movsf): Add another constraint alternative. 17040 (zero_extendqihi2): Add constraint alternatives for not in place 17041 extend. 17042 (zero_extendhisi2): Remove. 17043 (shift patterns): Add CC handling variants. 17044 (bswaphi2): New. 17045 (bswapsi2): New. 17046 (rothi3): New. 17047 (define_peephole2): New peephole to recognize mov that sets CC for 17048 subsequent test. 17049 170502018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE> 17051 17052 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST, 17053 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg, 17054 wi::add. 17055 170562018-07-09 Jakub Jelinek <jakub@redhat.com> 17057 17058 PR c/86420 17059 * real.c (real_nextafter): Return true if result is denormal. 17060 170612018-07-09 Martin Liska <mliska@suse.cz> 17062 17063 * common.opt: Add back wrongly removed attribute. 17064 170652018-07-09 Richard Biener <rguenther@suse.de> 17066 17067 PR debug/86413 17068 * dwarf2out.c (gen_block_die): For an early generated DIE 17069 always output high/low PC attributes. 17070 170712018-07-09 Tom de Vries <tdevries@suse.de> 17072 17073 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL. 17074 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping 17075 onto VAR_DECL with abstract origin. 17076 170772018-07-07 Jim Wilson <jimw@sifive.com> 17078 17079 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New. 17080 170812018-07-07 Tom de Vries <tdevries@suse.de> 17082 17083 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL. 17084 170852018-07-07 Aldy Hernandez <aldyh@redhat.com> 17086 17087 * tree-vrp.c (vrp_int_const_binop): Change overflow type to 17088 overflow_type. 17089 (combine_bound): Use wide-int overflow calculation instead of 17090 rolling our own. 17091 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to 17092 overflow_type. 17093 * fold-const.c (int_const_binop_2): Same. 17094 (extract_muldiv_1): Same. 17095 (fold_div_compare): Same. 17096 (fold_abs_const): Same. 17097 * match.pd: Same. 17098 * poly-int.h (add): Same. 17099 (sub): Same. 17100 (neg): Same. 17101 (mul): Same. 17102 * predict.c (predict_iv_comparison): Same. 17103 * profile-count.c (slow_safe_scale_64bit): Same. 17104 * simplify-rtx.c (simplify_const_binary_operation): Same. 17105 * tree-chrec.c (tree_fold_binomial): Same. 17106 * tree-data-ref.c (split_constant_offset_1): Same. 17107 * tree-if-conv.c (idx_within_array_bound): Same. 17108 * tree-scalar-evolution.c (iv_can_overflow_p): Same. 17109 * tree-ssa-phiopt.c (minmax_replacement): Same. 17110 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same. 17111 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same. 17112 * vr-values.c (vr_values::adjust_range_with_scev): Same. 17113 * wide-int.cc (wi::add_large): Same. 17114 (wi::mul_internal): Same. 17115 (wi::sub_large): Same. 17116 (wi::divmod_internal): Same. 17117 * wide-int.h: Change overflow type to overflow_type for neg, add, 17118 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round, 17119 mod_trunc, mod_ceil, mod_round, add_large, sub_large, 17120 mul_internal, divmod_internal. 17121 (overflow_type): New enum. 17122 (accumulate_overflow): New. 17123 171242018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org> 17125 17126 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New. 17127 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern. 17128 171292018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org> 17130 17131 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount 17132 argument is checked for zero before entering loop, avoid checking again. 17133 171342018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org> 17135 17136 * gimplify.h (generic_expr_could_trap_p): Set as global function. 17137 * gimplify.h (generic_expr_could_trap_p): Likwise. 17138 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR. 17139 171402018-07-06 Jakub Jelinek <jakub@redhat.com> 17141 17142 PR tree-optimization/86401 17143 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the 17144 ((A & N) + B) & M -> (A + B) & M etc. optimization into ... 17145 (fold_bit_and_mask): ... here. New helper function for match.pd. 17146 * fold-const.h (fold_bit_and_mask): Declare. 17147 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization. 17148 171492018-07-06 Peter Bergner <bergner@linux.ibm.com> 17150 17151 PR target/86324 17152 * target.def (translate_mode_attribute): New hook. 17153 * targhooks.h (default_translate_mode_attribute): Declare. 17154 * targhooks.c (default_translate_mode_attribute): New function. 17155 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook. 17156 * doc/tm.texi: Regenerate. 17157 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define. 17158 (rs6000_translate_mode_attribute): New function. 17159 171602018-07-06 Paul Koning <ni1d@arrl.net> 17161 17162 * doc/md.texi (define_split): Document DONE and FAIL. 17163 (define_peephole2): Ditto. 17164 171652018-07-05 Jeff Law <law@redhat.com> 17166 17167 PR tree-optimization/86010 17168 * tree-ssa-dse.c (compute_trims): More aggressively trim at 17169 both the head and tail of mem* and str* calls. 17170 171712018-07-05 Jim Wilson <jimw@sifive.com> 17172 17173 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*. 17174 171752018-07-05 Indu Bhagat <indu.bhagat@oracle.com> 17176 17177 * config/aarch64/aarch64-simd.md: correct flags text for 17178 MIN_EXPR replacement. 17179 171802018-07-05 James Clarke <jrtc27@jrtc27.com> 17181 17182 * configure: Regenerated. 17183 171842018-07-05 Carl Love <cel@us.ibm.com> 17185 17186 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for 17187 float argument to VSX_BUILTIN_DOUBLEH_V4SF. 17188 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to 17189 VSX_BUILTIN_DOUBLEL_V4SF. 17190 171912018-07-05 Martin Sebor <msebor@redhat.com> 17192 17193 PR c++/86400 17194 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather 17195 than its domain to compute its the upper bound of a char array. 17196 171972018-07-05 Nathan Sidwell <nathan@acm.org> 17198 17199 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C. 17200 * doc/cpp.texi: Update comment. 17201 * doc/tm.texi: Rebuilt. 17202 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ... 17203 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense. 17204 * doc/extend.texi (Backwards Compatibility): Clarify it is system 17205 headers affected by extern "C". 17206 * system.h: Poison NO_IMPLICIT_EXTERN_C. 17207 * config/alpha/alpha.h, config/arm/uclinux-elf.h, 17208 config/bfin/elf.h, config/cris/cris.h, config/darwin.h, 17209 config/dragonfly.h, config/freebsd.h, config/gnu-user.h, 17210 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h, 17211 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h, 17212 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h, 17213 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h, 17214 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h, 17215 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h, 17216 config/sparc/openbsd64.h, config/sparc/sp-elf.h, 17217 config/sparc/sp64-elf.h, config/spu/spu.h, 17218 config/stormy16/stormy16.h, config/v850/v850.h, 17219 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't 17220 define NO_IMPLICIT_EXTERN_C. 17221 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C. 17222 172232018-07-05 Tamar Christina <tamar.christina@arm.com> 17224 17225 PR target/84711 17226 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE 17227 instead of GET_MODE_SIZE when comparing Units. 17228 172292018-07-05 Tamar Christina <tamar.christina@arm.com> 17230 17231 PR target/84711 17232 * rtlanal.c (set_noop_p): Constrain on mode change, 17233 include hard-reg-set.h 17234 172352018-07-05 Tamar Christina <tamar.christina@arm.com> 17236 17237 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size. 17238 172392018-07-05 Jakub Jelinek <jakub@redhat.com> 17240 17241 Revert 17242 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com> 17243 17244 PR sanitizer/84250 17245 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static 17246 libasan. 17247 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan. 17248 172492018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com> 17250 17251 PR sanitizer/84250 17252 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static 17253 libasan. 17254 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan. 17255 172562018-07-04 Eric Botcazou <ebotcazou@adacore.com> 17257 17258 PR middle-end/86380 17259 * expmed.c (choose_multiplier): Fix incorrect comparison with mask. 17260 172612018-07-04 Aldy Hernandez <aldyh@redhat.com> 17262 17263 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize 17264 neg_*_op* variables. 17265 172662018-07-04 Martin Liska <mliska@suse.cz> 17267 17268 * tree-switch-conversion.c: Define 17269 max_ratio_for_speed and max_ratio_for_size constants. 17270 172712018-07-04 Denys Vlasenko <dvlasenk@redhat.com> 17272 Martin Liska <mliska@suse.cz> 17273 17274 PR middle-end/66240 17275 PR target/45996 17276 PR c/84100 17277 * common.opt: Rename align options with 'str_' prefix. 17278 * common/config/i386/i386-common.c (set_malign_value): New 17279 function. 17280 (ix86_handle_option): Use it to set -falign-* options/ 17281 * config/aarch64/aarch64-protos.h (struct tune_params): Change 17282 type from int to string. 17283 * config/aarch64/aarch64.c: Update default values from int 17284 to string. 17285 * config/alpha/alpha.c (alpha_override_options_after_change): 17286 Likewise. 17287 * config/arm/arm.c (arm_override_options_after_change_1): Likewise. 17288 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print 17289 max skip conditionally. 17290 * config/i386/freebsd.h (SUBALIGN_LOG): New. 17291 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print 17292 max skip conditionally. 17293 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print 17294 max skip conditionally. 17295 * config/i386/gnu-user.h (SUBALIGN_LOG): New. 17296 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print 17297 max skip conditionally. 17298 * config/i386/i386.c (struct ptt): Change type from int to 17299 string. 17300 (ix86_default_align): Set default values. 17301 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print 17302 max skip conditionally. 17303 * config/i386/iamcu.h (SUBALIGN_LOG): New. 17304 (ASM_OUTPUT_MAX_SKIP_ALIGN): 17305 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): 17306 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print 17307 max skip conditionally. 17308 * config/i386/openbsdelf.h (SUBALIGN_LOG): New. 17309 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.: 17310 * config/i386/x86-64.h (SUBALIGN_LOG): New. 17311 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print 17312 max skip conditionally. 17313 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise. 17314 * config/ia64/ia64.c (ia64_option_override): Set default values 17315 for alignment options. 17316 * config/m68k/m68k.c: Handle new str_align_* options. 17317 * config/mips/mips.c (mips_set_compression_mode): Change 17318 type of constants. 17319 (mips_option_override): Set default values for options. 17320 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal): 17321 Likewise. 17322 * config/rs6000/rs6000.c (rs6000_option_override_internal): 17323 Likewise. 17324 * config/rx/rx.c (rx_option_override): Likewise. 17325 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log. 17326 (LABEL_ALIGN): Use align_labels_log. 17327 (LOOP_ALIGN): Use align_loops_align. 17328 * config/s390/s390.c (s390_asm_output_function_label): Use new 17329 macros. 17330 * config/sh/sh.c (sh_override_options_after_change): 17331 Change type of constants. 17332 * config/spu/spu.c (spu_sched_init): Likewise. 17333 * config/sparc/sparc.c (sparc_option_override): Set default 17334 values for options. 17335 * config/visium/visium.c (visium_option_override): Likewise. 17336 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not 17337 emit p2align format with last argument if it's not needed. 17338 * doc/invoke.texi: Document extended format of -falign-*. 17339 * final.c: Use align_labels alignment. 17340 * flags.h (struct target_flag_state): Change type to use 17341 align_flags. 17342 (struct align_flags_tuple): New. 17343 (struct align_flags): Likewise. 17344 (align_loops_log): Redefine macro to use new types. 17345 (align_loops_max_skip): Redefine macro to use new types. 17346 (align_jumps_log): Redefine macro to use new types. 17347 (align_jumps_max_skip): Redefine macro to use new types. 17348 (align_labels_log): Redefine macro to use new types. 17349 (align_labels_max_skip): Redefine macro to use new types. 17350 (align_functions_log): Redefine macro to use new types. 17351 (align_loops): Redefine macro to use new types. 17352 (align_jumps): Redefine macro to use new types. 17353 (align_labels): Redefine macro to use new types. 17354 (align_functions): Redefine macro to use new types. 17355 (align_functions_max_skip): Redefine macro to use new types. 17356 (align_loops_value): New macro. 17357 (align_jumps_value): New macro. 17358 (align_labels_value): New macro. 17359 (align_functions_value): New macro. 17360 * function.c (invoke_set_current_function_hook): Propagate 17361 alignment values from flags to global variables default in 17362 topleev.h. 17363 * ipa-icf.c (sem_function::equals_wpa): Use 17364 cl_optimization_option_eq instead of memcmp. 17365 * lto-streamer.h (cl_optimization_stream_out): Support streaming 17366 of string types. 17367 (cl_optimization_stream_in): Likewise. 17368 * optc-save-gen.awk: Support strings in cl_optimization. 17369 * opth-gen.awk: Likewise. 17370 * opts.c (finish_options): Remove error checking of invalid 17371 value ranges. 17372 (MAX_CODE_ALIGN): Remove. 17373 (MAX_CODE_ALIGN_VALUE): Likewise. 17374 (parse_and_check_align_values): New function. 17375 (check_alignment_argument): Likewise. 17376 (common_handle_option): Use check_alignment_argument. 17377 * opts.h (parse_and_check_align_values): Declare. 17378 * toplev.c (init_alignments): Remove. 17379 (read_log_maxskip): New. 17380 (parse_N_M): Likewise. 17381 (parse_alignment_opts): Likewise. 17382 (backend_init_target): Remove usage of init_alignments. 17383 * toplev.h (parse_alignment_opts): Declare. 17384 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new 17385 argument. 17386 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise. 17387 * tree.c (cl_option_hasher::equal): New. 17388 * varasm.c: Use new global macros. 17389 173902018-07-04 Denys Vlasenko <dvlasenk@redhat.com> 17391 17392 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN): 17393 Use a simpler align directive also if MAXSKIP = ALIGN-1. 17394 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise. 17395 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise. 17396 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise. 17397 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise. 17398 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N 17399 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG 17400 define. Use a simpler align directive also if MAXSKIP = ALIGN-1. 17401 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise. 17402 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise. 17403 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise. 17404 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise. 17405 174062018-07-04 Martin Liska <mliska@suse.cz> 17407 Jonathan Wakely <jwakely@redhat.com> 17408 17409 * coverage.c: Use correct type. 17410 * doc/invoke.texi: Language correction. 17411 174122018-07-03 H.J. Lu <hongjiu.lu@intel.com> 17413 17414 PR target/85620 17415 * config/i386/i386.c (rest_of_insert_endbranch): Also generate 17416 ENDBRANCH for non-tail call which may return via indirect branch. 17417 * doc/extend.texi: Document indirect_return attribute. 17418 174192018-07-03 Martin Sebor <msebor@redhat.com> 17420 17421 PR tree-optimization/86274 17422 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify 17423 precondition. 17424 (format_floating): Correct handling of infinities and NaNs. 17425 174262018-07-03 Martin Sebor <msebor@redhat.com> 17427 17428 * print-tree.c (print_real_cst): New function. 17429 (print_node_brief): Call it. 17430 (print_node): Ditto. 17431 174322018-07-03 Jeff Law <law@redhat.com> 17433 17434 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge 17435 into a single pattern. 17436 17437 * config/h8300/h8300.md (ors code_iterator): New. 17438 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into 17439 a single pattern and single splitter. 17440 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern. 17441 (iorqi3_1, xorqi3_1): Likewise. 17442 (iorqi3, xorqi3 expanders): Similarly. 17443 17444 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with 17445 (movmd_internal) into a single pattern using the P mode iterator. 17446 (movmd splitters): Similarly. 17447 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns. 17448 (movsd splitters): Similarly. 17449 17450 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate 17451 ADDB, ADDW and ADDL into a single ADD attribute which selects the 17452 right table based on the size of the operand. 17453 * config/h8300/h8300.md (length_table): Corresponding changes. All 17454 references to "addb", "addw" and "addl" changed to "add". 17455 (btst patterns): Merge two variants into a single pattern. 17456 (tstqi, tsthi): Likewise. 17457 (addhi3_incdec, addsi3_incdec): Likewise. 17458 (subhi3_h8300hs, subsi3_h8300hs): Likewise. 17459 (mulhi3, mulsi3): Likewise. 17460 (udivhi3, udivsi3): Likewise. 17461 (divhi3, divsi3): Likewise. 17462 (andorqi3, andorhi3, andorsi3): Likewise. 17463 174642018-07-03 Uros Bizjak <ubizjak@gmail.com> 17465 17466 PR target/85694 17467 * config/i386/sse.md (uavg<mode>3_ceil): New expander. 17468 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander. 17469 174702018-07-03 Richard Sandiford <richard.sandiford@arm.com> 17471 17472 PR tree-optimization/85694 17473 * config/aarch64/iterators.md (HADD, RHADD): New int iterators. 17474 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and 17475 UNSPEC_URHADD. 17476 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor) 17477 (<u>avg<mode>3_ceil): New patterns. 17478 174792018-07-03 David Malcolm <dmalcolm@redhat.com> 17480 17481 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from 17482 scan-tree-dump directive. 17483 * gcc.dg/vect/slp-perm-2.c: Likewise. 17484 * gcc.dg/vect/slp-perm-3.c: Likewise. 17485 * gcc.dg/vect/slp-perm-5.c: Likewise. 17486 * gcc.dg/vect/slp-perm-6.c: Likewise. 17487 * gcc.dg/vect/slp-perm-7.c: Likewise. 17488 * gcc.dg/vect/slp-perm-8.c: Likewise. 17489 174902018-07-03 Marek Polacek <polacek@redhat.com> 17491 17492 PR middle-end/86202 17493 * gimple-fold.c (size_must_be_zero_p): Check the type of the size. 17494 174952018-07-03 Richard Biener <rguenther@suse.de> 17496 17497 PR ipa/86389 17498 * tree-ssa-structalias.c (find_func_clobbers): Properly 17499 handle indirect calls. 17500 175012018-07-03 Jeff Law <law@redhat.com> 17502 17503 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators. 17504 (shifts): New code iterator. 17505 (movqi, movhi, movsi, movsf expanders): Consolidate into a single 17506 expander. Fix HImode handling on H8/SX. 17507 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander. 17508 (subqi3, subhi3, subsi3 expanders): Likewise. 17509 (andqi3, andhi3, andsi3 expanders): Likewise. 17510 (iorqi3, iorhi3, iorsi3 expanders): Likewise. 17511 (xorqi3, xorhi3, xorsi3 expanders): Likewise. 17512 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise. 17513 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise. 17514 (zero_extendqihi2, zero_extendqisi2): Likewise. 17515 (extendqihi2, extendqisi2): Likewise. 17516 (rotlqi3, rotlhi3, rotlsi3): Likewise. 17517 (neghi2_h8300, negsi2_h8300): Likewise for these patterns. 17518 (rotlqi3_1, rotlhi3_1): Likewise. 17519 (logicalhi3_sn, logicalsi3_sn): Likewise. 17520 (logicalhi3, logicalsi3): Likewise. 17521 175222018-07-03 Richard Sandiford <richard.sandiford@arm.com> 17523 17524 * tree-vect-patterns.c (vect_recog_rotate_pattern) 17525 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern) 17526 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast) 17527 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector 17528 type to append_pattern_def_seq instead of creating a stmt_vec_info 17529 directly. 17530 (build_mask_conversion): Likewise. Remove vinfo argument. 17531 (vect_add_conversion_to_patterm): Likewise, renaming to... 17532 (vect_add_conversion_to_pattern): ...this. 17533 (vect_recog_mask_conversion_pattern): Update call to 17534 build_mask_conversion. Pass the vector type to 17535 append_pattern_def_seq here too. 17536 (vect_recog_gather_scatter_pattern): Update call to 17537 vect_add_conversion_to_pattern. 17538 175392018-07-03 Richard Sandiford <richard.sandiford@arm.com> 17540 17541 * tree-vect-patterns.c (new_pattern_def_seq): Delete. 17542 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern) 17543 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern) 17544 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set 17545 STMT_VINFO_PATTERN_DEF_SEQ to null here. 17546 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern) 17547 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use 17548 append_pattern_def_seq instead of new_pattern_def_seq. 17549 (vect_recog_divmod_pattern): Do both of the above. 17550 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ 17551 is null. 17552 175532018-07-03 Richard Sandiford <richard.sandiford@arm.com> 17554 17555 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): 17556 (vect_recog_sad_pattern, vect_recog_widen_op_pattern) 17557 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern): 17558 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern) 17559 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern) 17560 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern) 17561 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant) 17562 (vect_recog_mult_pattern, vect_recog_divmod_pattern) 17563 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern) 17564 (vect_recog_mask_conversion_pattern): Replace vec<gimple *> 17565 parameter with a single stmt_vec_info. 17566 (vect_recog_func_ptr): Likewise. 17567 (vect_recog_gather_scatter_pattern): Likewise, folding in... 17568 (vect_try_gather_scatter_pattern): ...this. 17569 (vect_pattern_recog_1): Remove stmts_to_replace and just pass 17570 the stmt_vec_info of the statement to be matched. Don't clear 17571 STMT_VINFO_RELATED_STMT. 17572 (vect_pattern_recog): Update call accordingly. 17573 175742018-07-03 Richard Sandiford <richard.sandiford@arm.com> 17575 17576 PR tree-optimization/85694 17577 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil) 17578 (uavgM3_ceil): Document new optabs. 17579 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector. 17580 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal 17581 functions. 17582 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab) 17583 (savg_ceil_optab): New optabs. 17584 * tree-vect-patterns.c (vect_recog_average_pattern): New function. 17585 (vect_vect_recog_func_ptrs): Add it. 17586 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero 17587 constant directly from the associated lhs. 17588 175892018-07-03 Richard Sandiford <richard.sandiford@arm.com> 17590 17591 * tree-vect-patterns.c (vect_split_statement): New function. 17592 (vect_convert_input): Use it to try to split an existing cast. 17593 175942018-07-03 Richard Sandiford <richard.sandiford@arm.com> 17595 17596 * poly-int.h (print_hex): New function. 17597 * dumpfile.h (dump_dec, dump_hex): Declare. 17598 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions. 17599 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision, 17600 min_input_precision, operation_precision and operation_sign. 17601 * tree-vect-patterns.c (vect_get_range_info): New function. 17602 (vect_same_loop_or_bb_p, vect_single_imm_use) 17603 (vect_operation_fits_smaller_type): Delete. 17604 (vect_look_through_possible_promotion): Add an optional 17605 single_use_p parameter. 17606 (vect_recog_over_widening_pattern): Rewrite to use new 17607 stmt_vec_info infomration. Handle one operation at a time. 17608 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p) 17609 (vect_truncatable_operation_p, vect_set_operation_type) 17610 (vect_set_min_input_precision): New functions. 17611 (vect_determine_min_output_precision_1): Likewise. 17612 (vect_determine_min_output_precision): Likewise. 17613 (vect_determine_precisions_from_range): Likewise. 17614 (vect_determine_precisions_from_users): Likewise. 17615 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise. 17616 (vect_vect_recog_func_ptrs): Put over_widening first. 17617 Add cast_forwprop. 17618 (vect_pattern_recog): Call vect_determine_precisions. 17619 176202018-07-03 Richard Sandiford <richard.sandiford@arm.com> 17621 17622 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern 17623 statements that have been replaced by further pattern statements. 17624 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure. 17625 176262018-07-03 Richard Biener <rguenther@suse.de> 17627 17628 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping, 17629 always set *dt. Dump vectype in vectype overload. 17630 * dumpfile.h (dump_gimple_expr): New function. 17631 (dump_gimple_expr_loc): Likewise. 17632 * dumpfile.c (dump_gimple_expr): New function. 17633 (dump_gimple_expr_loc): Likewise. 17634 176352018-07-02 Jeff Law <law@redhat.com> 17636 17637 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate 17638 the H8/300, H8/300H and H8/S variants into a single pattern. 17639 (movhi_h8300, movqi_h8300hs): Similarly. 17640 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern. 17641 (QHI mode iterator): New. 17642 17643 * config/h8300/h8300.md: Remove trailing whitespace. 17644 176452018-07-02 Jim Wilson <jimw@sifive.com> 17646 17647 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn 17648 instead of emit_insn for interrupt returns. 17649 * config/riscv/riscv.md (riscv_met): Add (return) to rtl. 17650 (riscv_sret, riscv_uret): Likewise. 17651 176522018-07-02 David Malcolm <dmalcolm@redhat.com> 17653 17654 * pretty-print.c (selftest::test_pp_format): Move save and restore 17655 of quotes to class auto_fix_quotes, and add an instance. 17656 * selftest.c: Include "intl.h". 17657 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor. 17658 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor. 17659 * selftest.h (selftest::auto_fix_quotes): New class. 17660 176612018-07-02 Richard Henderson <richard.henderson@linaro.org> 17662 17663 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c 17664 (aarch64_sve_prepare_conditional_op): Remove. 17665 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>): 17666 Allow aarch64_simd_reg_or_zero as select operand; remove 17667 the aarch64_sve_prepare_conditional_op call. 17668 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise. 17669 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise. 17670 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern. 17671 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern. 17672 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern. 17673 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern. 17674 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern. 17675 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern 17676 and a splitters to match all of the *_any patterns. 17677 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New. 17678 17679 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove. 17680 (SVE_COND_FP_BINARY_REV): Remove. 17681 (sve_int_op_rev, sve_fp_op_rev): New. 17682 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New. 17683 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New. 17684 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New. 17685 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative. 17686 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly. 17687 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly. 17688 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev. 17689 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly. 17690 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev. 17691 17692 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>): 17693 Remove match_dup 1 from the inner unspec. 17694 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise. 17695 17696 * config/aarch64/aarch64.md (movprfx): New attr. 17697 (length): Default movprfx to 8. 17698 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt. 17699 (*madd<SVE_I>, *msub<SVE_I): Likewise. 17700 (*<su>mul<SVE_I>3_highpart): Likewise. 17701 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise. 17702 (*v<ASHIFT><SVE_I>3): Likewise. 17703 (*<su><MAXMIN><SVE_I>3): Likewise. 17704 (*<su><MAXMIN><SVE_F>3): Likewise. 17705 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise. 17706 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise. 17707 (*div<SVE_F>4): Likewise. 17708 177092018-07-02 Richard Sandiford <richard.sandiford@arm.com> 17710 17711 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo 17712 in dump string. 17713 177142018-07-02 Richard Biener <rguenther@suse.de> 17715 17716 PR tree-optimization/86363 17717 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the 17718 memset argument refers to a non-variable address. 17719 177202018-07-02 Aldy Hernandez <aldyh@redhat.com> 17721 17722 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the 17723 {PLUS,MINUS}_EXPR code to... 17724 (adjust_symbolic_bound): ...here, 17725 (combine_bound): ...here, 17726 (set_value_range_with_overflow): ...and here. 17727 177282018-07-02 Aldy Hernandez <aldyh@redhat.com> 17729 17730 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR 17731 code... 17732 (extract_range_from_abs_expr): ...here. 17733 177342018-07-02 Eric Botcazou <ebotcazou@adacore.com> 17735 17736 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule 17737 -fno-omit-frame-pointer when not optimizing. 17738 177392018-07-02 Martin Liska <mliska@suse.cz> 17740 17741 PR ipa/86279 17742 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get. 17743 (propagate_nothrow): Likewise. 17744 177452018-07-02 Martin Liska <mliska@suse.cz> 17746 17747 PR ipa/86323 17748 * ipa-inline.c (early_inliner): Revert wrongly added ::get call. 17749 177502018-07-02 David Malcolm <dmalcolm@redhat.com> 17751 17752 * dumpfile.c (dump_generic_expr_loc): Undo removal of this 17753 function in r262149, changing "loc" param from source_location to 17754 const dump_location_t &. 17755 * dumpfile.h (dump_generic_expr_loc): Undo removal of this 17756 declaration, as above. 17757 177582018-07-01 Paul Koning <ni1d@arrl.net> 17759 17760 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle 17761 -munit-asm, -mgnu-asm, -mdec-asm. 17762 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New. 17763 (pdp11_output_labelref): New. 17764 (pdp11_output_def): New. 17765 (pdp11_output_addr_vec_elt): New. 17766 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use 17767 %# and %@ format codes. 17768 (pdp11_option_override): New. 17769 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define. 17770 (pdp11_output_ident): New. 17771 (pdp11_asm_named_section): New. 17772 (pdp11_asm_init_sections): New. 17773 (pdp11_file_start): New. 17774 (pdp11_file_end): New. 17775 (output_ascii): Use .ascii/.asciz for -mdec-asm. 17776 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add 17777 %o, like %c but octal. 17778 (pdp11_option_override): New. 17779 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for 17780 -mdec-asm. 17781 (DATA_SECTION_ASM_OP): Ditto. 17782 (READONLY_DATA_SECTION_ASM_OP): New. 17783 (IS_ASM_LOGICAL_LINE_SEPARATOR): New. 17784 (ASM_GENERATE_INTERNAL_LABEL): Use new function. 17785 (ASM_OUTPUT_LABELREF): Ditto. 17786 (ASM_OUTPUT_DEF): Ditto. 17787 (ASM_OUTPUT_EXTERNAL): New. 17788 (ASM_OUTPUT_SOURCE_FILENAME): New. 17789 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function. 17790 (ASM_OUTPUT_SKIP): Update for -mdec-asm. 17791 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use 17792 %# and %@ format codes. 17793 * config/pdp11/pdp11.opt (mgnu-asm): New. 17794 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm. 17795 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm. 17796 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm. 17797 177982018-07-01 Aldy Hernandez <aldyh@redhat.com> 17799 17800 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid 17801 dereferencing path[] beyond its length. 17802 (debug_path): New. 17803 (debug_all_paths): New. 17804 (rewire_first_differing_edge): New. 17805 (adjust_paths_after_duplication): New. 17806 (duplicate_thread_path): Call adjust_paths_after_duplication. 17807 Add new argument. 17808 (thread_through_all_blocks): Add new argument to 17809 duplicate_thread_path. 17810 178112018-06-30 Jim Wilson <jimw@sifive.com> 17812 17813 * config/riscv/predicates.md (p2m1_shift_operand): New. 17814 (high_mask_shift_operand): New. 17815 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner 17816 pattern using p2m1_shift_operand. 17817 (lshsi3_zero_extend_3+2): New combiner pattern using 17818 high_mask_shift_operand. 17819 178202018-06-30 Richard Sandiford <richard.sandiford@arm.com> 17821 17822 * tree-vect-patterns.c (vect_get_external_def_edge): New function, 17823 split out from... 17824 (vect_recog_rotate_pattern): ...here. 17825 (vect_convert_input): Try to insert casts of invariants in the 17826 preheader. 17827 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the 17828 preheader to be empty. 17829 178302018-06-30 Richard Sandiford <richard.sandiford@arm.com> 17831 17832 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional 17833 vector type. If given, install it in the new statement's 17834 STMT_VINFO_VECTYPE. 17835 (vect_element_precision): New function. 17836 (vect_unpromoted_value): New struct. 17837 (vect_unpromoted_value::vect_unpromoted_value): New function. 17838 (vect_unpromoted_value::set_op): Likewise. 17839 (vect_look_through_possible_promotion): Likewise. 17840 (vect_joust_widened_integer, vect_joust_widened_type): Likewise. 17841 (vect_widened_op_tree, vect_convert_input): Likewise. 17842 (vect_convert_inputs, vect_convert_output): Likewise. 17843 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion 17844 to handle the optional cast of the multiplication result and 17845 vect_widened_op_tree to detect the widened multiplication itself. 17846 Do not require the input and output of promotion casts to have 17847 the same sign, but base the signedness of the operation on the 17848 input rather than the result. If the pattern includes two 17849 promotions, check that those promotions have the same sign. 17850 Do not restrict the MULT_EXPR handling to a double-width result; 17851 handle quadruple-width results and wider. Use vect_convert_inputs 17852 to convert the inputs to the common type. 17853 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion 17854 to handle the optional cast of the ABS result. Also allow a sign 17855 change or a sign extension between the ABS and MINUS. 17856 Use vect_widened_op_tree to detect the widened subtraction and use 17857 vect_convert_inputs to convert the inputs to the common type. 17858 (vect_handle_widen_op_by_const): Delete. 17859 (vect_recog_widen_op_pattern): New function. 17860 (vect_recog_widen_mult_pattern): Use it. 17861 (vect_recog_widen_shift_pattern): Likewise. 17862 (vect_recog_widen_sum_pattern): Use 17863 vect_look_through_possible_promotion to handle the promoted 17864 PLUS_EXPR operand. 17865 178662018-06-30 Richard Sandiford <richard.sandiford@arm.com> 17867 17868 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes 17869 the containing gimple_seq *. 17870 * gimple-iterator.h (gsi_for_stmt): Declare it. 17871 * tree-vect-patterns.c (vect_recog_dot_prod_pattern) 17872 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern) 17873 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern) 17874 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern) 17875 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P 17876 checks. 17877 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions, 17878 split out from... 17879 (vect_mark_pattern_stmts): ...here. Handle cases in which the 17880 statement being replaced is part of an existing pattern 17881 definition sequence, inserting the new pattern statements before 17882 the original one. 17883 (vect_pattern_recog_1): Don't return a bool. If the statement 17884 is already part of a pattern, instead apply pattern matching 17885 to the pattern definition statements. Don't clear the 17886 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p. 17887 (vect_pattern_recog): Don't break after the first match; 17888 continue processing the pattern definition statements instead. 17889 Don't bail out for STMT_VINFO_IN_PATTERN_P here. 17890 178912018-06-30 Richard Sandiford <richard.sandiford@arm.com> 17892 17893 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function. 17894 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern) 17895 (vect_recog_widen_sum_pattern): Use it. 17896 178972018-06-30 Richard Sandiford <richard.sandiford@arm.com> 17898 17899 * tree-vect-loop.c (vectorizable_reduction): Assert that the 17900 phi is not a pattern statement and has not been replaced by 17901 a pattern statement. 17902 * tree-vect-patterns.c (type_conversion_p): Don't check 17903 STMT_VINFO_IN_PATTERN_P. 17904 (vect_recog_vector_vector_shift_pattern): Likewise. 17905 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return 17906 the pattern statement rather than the original statement; check 17907 directly for a WIDEN_MULT_EXPR here. 17908 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect 17909 vect_is_simple_use to return the pattern statement rather 17910 than the original statement; use is_pattern_stmt_p to check 17911 for such a pattern statement. 17912 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use 17913 to return the pattern statement rather than the original statement; 17914 don't do the same transformation here. 17915 (vect_is_simple_use): If the defining statement has been replaced 17916 by a pattern statement, return the pattern statement instead. 17917 Remove the corresponding (local) transformation from the vectype 17918 overload. 17919 179202018-06-30 Richard Sandiford <richard.sandiford@arm.com> 17921 17922 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the 17923 end and default to null. 17924 * tree-vect-loop.c (vect_create_epilog_for_reduction) 17925 (vectorizable_reduction): Update calls accordingly, dropping the 17926 gimple ** argument if the passed-back statement isn't needed. 17927 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p) 17928 (vect_recog_rotate_pattern): Likewise. 17929 (vect_recog_mask_conversion_pattern): Likewise. 17930 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise. 17931 (vect_mask_constant_operand_p): Likewise. 17932 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use): 17933 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise. 17934 (get_group_load_store_type, get_load_store_type): Likewise. 17935 (vect_check_load_store_mask, vect_check_store_rhs): Likewise. 17936 (vectorizable_call, vectorizable_simd_clone_call): Likewise. 17937 (vectorizable_conversion, vectorizable_assignment): Likewise. 17938 (vectorizable_shift, vectorizable_operation): Likewise. 17939 (vectorizable_store, vect_is_simple_cond): Likewise. 17940 (vectorizable_condition, vectorizable_comparison): Likewise. 17941 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise. 17942 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out 17943 and move it to the end. Cope with null def_stmt_outs. 17944 179452018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de> 17946 17947 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET. 17948 179492018-06-29 Jeff Law <law@redhat.com> 17950 17951 * config/v850/v850.c (v850_legitimate_address_p): Handle large 17952 displacements for TARGET_V850E2V3 and newer. 17953 (TARGET_LRA_P): Remove. Defaults to LRA now. 17954 * config/v850/v850.md (sign23byte_load): Remove. 17955 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise. 17956 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise. 17957 179582018-06-29 Martin Liska <mliska@suse.cz> 17959 17960 PR lto/85759 17961 * coverage.c (coverage_init): Mangle full path name. 17962 * doc/invoke.texi: Document the change. 17963 * gcov-io.c (mangle_path): New. 17964 * gcov-io.h (mangle_path): Likewise. 17965 * gcov.c (mangle_name): Use mangle_path for path mangling. 17966 179672018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 17968 17969 * config/arm/arm.c (output_move_double): Don't allow STRD instructions 17970 if starting source register is not even. 17971 179722018-06-29 Martin Liska <mliska@suse.cz> 17973 17974 PR tree-optimization/86263 17975 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion): 17976 Make edge redirection. 17977 179782018-06-29 David Malcolm <dmalcolm@redhat.com> 17979 17980 * dumpfile.c (dump_loc): Add indentation based on scope depth. 17981 (dump_scope_depth): New variable. 17982 (get_dump_scope_depth): New function. 17983 (dump_begin_scope): New function. 17984 (dump_end_scope): New function. 17985 * dumpfile.h (get_dump_scope_depth): New declaration. 17986 (dump_begin_scope): New declaration. 17987 (dump_end_scope): New declaration. 17988 (class auto_dump_scope): New class. 17989 (AUTO_DUMP_SCOPE): New macro. 17990 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of 17991 AUTO_DUMP_SCOPE. 17992 179932018-06-29 Richard Biener <rguenther@suse.de> 17994 17995 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert 17996 compute_all_dependences succeeds. 17997 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we 17998 exceed --param loop-max-datarefs-for-datadeps. 17999 180002018-06-29 Jakub Jelinek <jakub@redhat.com> 18001 18002 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H. 18003 180042018-06-28 Uros Bizjak <ubizjak@gmail.com> 18005 18006 PR target/86348 18007 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use 18008 alternative 0 in preferred_for_speed attribute. 18009 180102018-06-28 Paul Koning <ni1d@arrl.net> 18011 18012 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function. 18013 * config/pdp11/pdp11.c (pdp11_shift_length): New function. 18014 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove. 18015 * config/pdp11/pdp11.md: Correct "length" attribute calculation 18016 for shift insn patterns. 18017 180182018-06-28 David Malcolm <dmalcolm@redhat.com> 18019 18020 * cgraph.c (cgraph_node::get_body): Replace assignments to 18021 "dump_file" with calls to set_dump_file. 18022 * dumpfile.c (alt_dump_file): Make static, and group with... 18023 (alt_flags): ...this definition. 18024 (dumps_are_enabled): New variable. 18025 (refresh_dumps_are_enabled): New function. 18026 (set_dump_file): New function. 18027 (set_alt_dump_file): New function. 18028 (gcc::dump_manager::dump_start): Replace assignments to 18029 "dump_file" and "alt_dump_file" with calls to set_dump_file and 18030 set_alt_dump_file. 18031 (gcc::dump_manager::dump_finish): Likewise. 18032 * dumpfile.h (alt_dump_file): Delete decl. 18033 (dumps_are_enabled): New variable decl. 18034 (set_dump_file): New function decl. 18035 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled" 18036 global. 18037 * tree-nested.c (lower_nested_functions): Replace assignments to 18038 "dump_file" with calls to set_dump_file. 18039 180402018-06-28 Eric Botcazou <ebotcazou@adacore.com> 18041 18042 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the 18043 goto_locus of each outgoing edge of each basic block. 18044 180452018-06-28 Richard Biener <rguenther@suse.de> 18046 18047 * dwarf2out.c (decl_scope_table): Remove. 18048 (push_decl_scope): Likewise. 18049 (pop_decl_scope): Likewise. 18050 (gen_type_die_for_member): Do not call push/pop_decl_scope. 18051 (gen_struct_or_union_type_die): Likewise. 18052 (gen_tagged_type_die): Likewise. 18053 (dwarf2out_init): Do not initialize decl_scope_table. 18054 (dwarf2out_c_finalize): Do not free it. 18055 180562018-06-28 Richard Biener <rguenther@suse.de> 18057 18058 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when 18059 deciding whether to not re-use a DIE. 18060 180612018-06-28 Richard Biener <rguenther@suse.de> 18062 18063 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an 18064 DW_AT_abstract_origin attribute. 18065 180662018-06-28 Martin Liska <mliska@suse.cz> 18067 18068 * tree-switch-conversion.c (jump_table_cluster::can_be_handled): 18069 Use newly introduced constants. 18070 * tree-switch-conversion.h (struct jump_table_cluster): 18071 Define max_ratio_for_size and max_ratio_for_speed. 18072 180732018-06-28 Martin Liska <mliska@suse.cz> 18074 18075 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables): 18076 Add new checking assert to catch invalid state. 18077 (jump_table_cluster::can_be_handled): Handle single case 18078 clusters. 18079 (jump_table_cluster::is_beneficial): Bail out for such case. 18080 (bit_test_cluster::find_bit_tests): 18081 Add new checking assert to catch invalid state. 18082 (bit_test_cluster::can_be_handled): Handle single case 18083 clusters. 18084 (bit_test_cluster::is_beneficial): Bail out for such case. 18085 (switch_decision_tree::analyze_switch_statement): 18086 Fix comment. 18087 180882018-06-28 Martin Liska <mliska@suse.cz> 18089 18090 * common.opt: Introduce -completion option. 18091 * gcc.c (driver_handle_option): Handle it. 18092 (driver::main): Print completions if completion 18093 is set. 18094 * opt-suggestions.c (option_proposer::get_completions): 18095 New function. 18096 (option_proposer::suggest_completion): Likewise. 18097 (option_proposer::find_param_completions): Likewise. 18098 (verify_autocompletions): Likewise. 18099 (test_completion_valid_options): Likewise. 18100 (test_completion_valid_params): Likewise. 18101 (in_completion_p): Likewise. 18102 (empty_completion_p): Likewise. 18103 (test_completion_partial_match): Likewise. 18104 (test_completion_garbage): Likewise. 18105 (opt_proposer_c_tests): Likewise. 18106 * opt-suggestions.h: Declare new functions. 18107 * opts.c (common_handle_option): Handle OPT__completion_. 18108 * selftest-run-tests.c (selftest::run_tests): Add 18109 opt_proposer_c_tests. 18110 * selftest.c (assert_str_startswith): New. 18111 * selftest.h (assert_str_startswith): Likewise. 18112 (opt_proposer_c_tests): New. 18113 (ASSERT_STR_STARTSWITH): Likewise. 18114 181152018-06-28 Martin Liska <mliska@suse.cz> 18116 18117 * Makefile.in: Add opt-suggestions.o. 18118 * gcc-main.c: Include opt-suggestions.h. 18119 * gcc.c (driver::driver): Likewise. 18120 (driver::~driver): Remove m_option_suggestions. 18121 (driver::build_option_suggestions): Moved to option_proposer. 18122 (driver::suggest_option): Likewise. 18123 (driver::handle_unrecognized_options): Use option_proposer. 18124 * gcc.h (class driver): Add new memver m_option_proposer. 18125 * opt-suggestions.c: New file. 18126 * opt-suggestions.h: New file. 18127 181282018-06-28 Martin Liska <mliska@suse.cz> 18129 18130 * vec.h (class auto_string_vec): New (moved from auto_argvec). 18131 (auto_string_vec::~auto_string_vec): Likewise. 18132 181332018-06-28 Eric Botcazou <ebotcazou@adacore.com> 18134 18135 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and 18136 prevent_decl_creation_for_types fields up and add reset_location field. 18137 * tree-inline.c (remap_gimple_stmt): Force input_location on the new 18138 statement if id->reset_location is true. 18139 (copy_edges_for_bb): Do not set goto_locus on the new edges if 18140 id->reset_location is true. 18141 (copy_phis_for_bb): Force input_location on the arguments if 18142 id->reset_location is true. 18143 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P 18144 is set on the function to be inlined. 18145 181462018-06-27 Stephan Bergmann <sbergman@redhat.com> 18147 18148 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf. 18149 181502018-06-27 Dimitar Dimitrov <dimitar@dinux.eu> 18151 18152 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard 18153 registers for Pmode. 18154 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning 18155 hard registers for the clobbered pseudo. 18156 181572018-06-27 Paul Koning <ni1d@arrl.net> 18158 18159 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle 18160 mutually exclusive options. 18161 * config/pdp11/constraints.md (h): New constraint. 18162 (O): Update definition to match shift code generation. 18163 (D): New constraint. 18164 * config/pdp11/pdp11-modes.def (CCNZ): Define mode. 18165 (CCFP): Remove. 18166 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New 18167 function. 18168 (output_jump): Change arguments. 18169 (pdp11_fixed_cc_regs): New function. 18170 (pdp11_cc_mode): Ditto. 18171 (pdp11_expand_shift): Ditto. 18172 (pdp11_assemble_shift): Ditto. 18173 (pdp11_small_shift): Ditto. 18174 (pdp11_branch_cost): Remove. 18175 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments 18176 from output. 18177 (pdp11_register_move_cost): Update for CC registers. 18178 (pdp11_rtx_costs): Add case for LSHIFTRT. 18179 (pdp11_output_jump): Add CCNZ mode conditional branches. 18180 (notice_update_cc_on_set): Remove. 18181 (pdp11_cc_mode): New function. 18182 (simple_memory_operand): Correct pre/post decrement case. 18183 (no_side_effect_operand): New function. 18184 (pdp11_regno_reg_class): Add CC_REGS class. 18185 (pdp11_fixed_cc_regs): New function. 18186 (pdp11_small_shift): New function. 18187 (pdp11_expand_shift): New function to expand shift insns. 18188 (pdp11_assemble_shift): New function to output shifts. 18189 (pdp11_branch_cost): Remove. 18190 (pdp11_modes_tieable_p): Make QI/HI modes tieable. 18191 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type. 18192 (WCHAR_TYPE): Ditto. 18193 (PTRDIFF_TYPE): Ditto. 18194 (ADJUST_INSN_LENGTH): New macro. 18195 (FIXED_REGISTERS): Add CC registers. 18196 (CALL_USED_REGISTERS): Ditto. 18197 (reg_class): Ditto. 18198 (REG_CLASS_NAMES): Ditto. 18199 (REG_CLASS_CONTENTS): Ditto. 18200 (SELECT_CC_MODE): Use new function. 18201 (TARGET_FLAGS_REGNUM): New macro. 18202 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto. 18203 (cc0_reg_rtx): Remove. 18204 (CC_STATUS_MDEP): Remove. 18205 (CC_STATUS_MDEFP_INIT): Remove. 18206 (CC_IN_FPU): Remove. 18207 (NOTICE_UPDATE_CC): Remove. 18208 (REGISTER_NAMES): Add CC registers. 18209 (BRANCH_COST): Change to constant 1. 18210 * config/pdp11/pdp11.md: Rewrite for CCmode condition code 18211 handling. 18212 * config/pdp11/pdp11.opt (mbcopy): Remove. 18213 (mbcopy-builtin): Remove. 18214 (mbranch-cheap): Remove. 18215 (mbranch-expensive): Remove. 18216 * config/pdp11/predicates.md (expand_shift_operand): Update to 18217 match shift code generation. 18218 (ccnz_operator): New predicate. 18219 * doc/invoke.texi (PDP-11 Options): Remove deleted options 18220 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive. 18221 Remove non-existent option -mabshi, -mno-abshi. Document mutually 18222 exclusive options. 18223 * doc/md.texi (PDP-11): Document new D and h constraints. Update 18224 description of O constraint. 18225 182262018-06-27 Jeff Law <law@redhat.com> 18227 Austin Law <austinklaw@gmail.com> 18228 18229 * config/v850/v850.md (addsi3_set_flags): New pattern. 18230 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise. 18231 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise. 18232 (zero_extendhisi2_v850_set_flags): Likewise. 18233 (zero_extendqisi2_v850_set_flags): Likewise. 18234 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise. 18235 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise. 18236 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise. 18237 18238 * config/v850/v850-protos.h (notice_update_cc): Remove. 18239 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove. 18240 (v850_print_operand): Handle 'D' and "d". 18241 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument. 18242 Add handling of arithmetic/logical operations compared against zero. 18243 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument. 18244 Do not look at v850_compare_op, instead get mode from last argument. 18245 (v850_gen_compare): Remove 18246 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure 18247 after reload for prologue insns. 18248 (expand_prologue): Account for CLOBBER of CC_REGNUM in various 18249 patterns. 18250 (construct_save_jarl): Likewise. 18251 (TARGET_FLAGS_REGNUM): Define. 18252 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove. 18253 (NOTICE_UPDATE_CC): Remove. 18254 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather 18255 than cc0. Conditionalize on reload_completed. 18256 (cmpsi_insn, setfcc_insn): Likewise. 18257 (tst1 splitter): Turn into define_and_split which sets the flags 18258 after reload. 18259 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise. 18260 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here. 18261 (cstoresf4, cstoredf4): Clobber the flags. 18262 (cmpsi, cmpsf, cmpdf): Remove expanders. 18263 (setf_insn): Remove pattern. 18264 (addsi3): Turn into define_and_split which clobbers the flags after 18265 reload and a suitable pattern (addsi3_clobber_flags) for use after 18266 reload. 18267 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise. 18268 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise. 18269 (ashrsi3, ashrsi3_v850e2): Likewise. 18270 (bins): Clobber the flags. 18271 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise. 18272 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise. 18273 (fix_loop_counter, call_internal_short, call_internal_long): Likewise. 18274 (call_value_internal_short, call_value_internal_long): Likewise. 18275 (callt_save_interrupt, callt_return_interrupt): Likewise. 18276 (save_interrupt, return_interrupt): Likewise. 18277 (callt_save_all_interrupt, save_all_interrupt): Likewise. 18278 (_save_all_interrupt, callt_restore_all_interrupt): Likewise. 18279 (restore_all_interrupt, _restore_all_interrupt): Likewise. 18280 (All FP comparisons): Only allow after reload has completed. 18281 (trfsr): Likewise. 18282 (divh, divhu): Tweak output template. 18283 (branch_z_normal, branch_z_invert): Remove 18284 (branch_nz_normal, branch_nz_invert): Likewise. 18285 (extendhisi_insn, extendqisi_insn): Do not clobber flags. 18286 18287 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes. 18288 * config/v850/v850.c (notice_update_cc): Remove. 18289 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove 18290 (CC_NO_CARRY): Likewise. 18291 (NOTICE_UPDATE_CC): Define to nothing. 18292 * config/v850/v850.md: Remove block comment on cc0 handling 18293 Remove "cc" attribute from all patterns. Remove cc_status handling 18294 from all patterns. Minor formatting fixes. 18295 182962018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 18297 18298 * config/aarch64/aarch64-cores.def (cortex-a76): New entry. 18299 (cortex-a76.cortex-a55): Likewise. 18300 * config/aarch64/aarch64-tune.md: Regenerate. 18301 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and 18302 cortex-a76.cortex-a55. 18303 183042018-06-27 Jeff Law <law@redhat.com> 18305 18306 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align. 18307 (MULTILIB_DIRNAMES): Similarly. 18308 183092018-06-27 Eric Botcazou <ebotcazou@adacore.com> 18310 18311 * gimple.h (gimple_return_retbnd): Delete. 18312 (gimple_return_set_retbnd): Likewise. 18313 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to 18314 gimple_return_set_retbnd. 18315 * gimple-pretty-print.c (dump_gimple_return): Remove call to 18316 gimple_return_retbnd and adjust. 18317 * tree-inline.h (struct copy_body_data): Remove retbnd field. 18318 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd. 18319 Explicitly return NULL in a couple more cases. Move assertion 18320 on debug statements and remove unreachable code. 18321 (reset_debug_binding): Do not test id->retbnd. 18322 (expand_call_inline): Do not set it. 18323 183242018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk> 18325 18326 * configure.ac: Add --disable-gcov option. 18327 * configure: Regenerate. 18328 * Makefile.in: Honour @enable_gcov@. 18329 * doc/install.texi: Document --disable-gcov. 18330 183312018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 18332 18333 * config/arm/arm-cpus.in (cortex-a76): New entry. 18334 (cortex-a76.cortex-a55): Likewise. 18335 * config/arm/arm-tables.opt: Regenerate. 18336 * config/arm/arm-tune.md: Likewise. 18337 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry. 18338 * doc/invoke.texi (ARM Options): Document cortex-a76 and 18339 cortex-a76.cortex-a55. 18340 183412018-06-27 Tamar Christina <tamar.christina@arm.com> 18342 18343 PR target/85769 18344 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern. 18345 183462018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org> 18347 18348 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete 18349 comment. 18350 (EPILOGUE_USES): Likewise. 18351 183522018-06-26 Eric Botcazou <ebotcazou@adacore.com> 18353 18354 * tree-inline.c (remap_location): New function extracted from... 18355 (copy_edges_for_bb): Add ID parameter. Remap goto_locus. 18356 (copy_phis_for_bb): ...here. Call remap_location. 18357 (copy_cfg_body): Adjust call to copy_edges_for_bb. 18358 183592018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com> 18360 18361 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use 18362 unaligned vsx for 16B memset. 18363 183642018-06-26 Segher Boessenkool <segher@kernel.crashing.org> 18365 18366 PR target/86285 18367 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set 18368 ieee128_float_type_node to long_double_type_node unless 18369 TARGET_LONG_DOUBLE_128 is set. 18370 183712018-06-26 David Malcolm <dmalcolm@redhat.com> 18372 18373 * cfgloop.c (get_loop_location): Convert return type from 18374 location_t to dump_user_location_t, replacing INSN_LOCATION lookups 18375 by implicit construction from rtx_insn *, and using 18376 dump_user_location_t::from_function_decl for the fallback case. 18377 * cfgloop.h (get_loop_location): Convert return type from 18378 location_t to dump_user_location_t. 18379 * cgraphunit.c (walk_polymorphic_call_targets): Update call to 18380 dump_printf_loc to pass in a dump_location_t rather than a 18381 location_t, via the gimple stmt. 18382 * coverage.c (get_coverage_counts): Update calls to 18383 dump_printf_loc to pass in dump_location_t rather than a 18384 location_t. 18385 * doc/optinfo.texi (Dump types): Convert example of 18386 dump_printf_loc from taking "locus" to taking "insn". Update 18387 description of the "_loc" calls to cover dump_location_t. 18388 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and 18389 "selftest.h". 18390 (dump_user_location_t::dump_user_location_t): New constructors, 18391 from gimple *stmt and rtx_insn *. 18392 (dump_user_location_t::from_function_decl): New function. 18393 (dump_loc): Make static. 18394 (dump_gimple_stmt_loc): Convert param "loc" from location_t to 18395 const dump_location_t &. 18396 (dump_generic_expr_loc): Delete. 18397 (dump_printf_loc): Convert param "loc" from location_t to 18398 const dump_location_t &. 18399 (selftest::test_impl_location): New function. 18400 (selftest::dumpfile_c_tests): New function. 18401 * dumpfile.h: Include "profile-count.h". 18402 (class dump_user_location_t): New class. 18403 (struct dump_impl_location_t): New struct. 18404 (class dump_location_t): New class. 18405 (dump_printf_loc): Convert 2nd param from source_location to 18406 const dump_location_t &. 18407 (dump_generic_expr_loc): Delete. 18408 (dump_gimple_stmt_loc): Convert 2nd param from source_location to 18409 const dump_location_t &. 18410 * gimple-fold.c (fold_gimple_assign): Update call to 18411 dump_printf_loc to pass in a dump_location_t rather than a 18412 location_t, via the gimple stmt. 18413 (gimple_fold_call): Likewise. 18414 * gimple-loop-interchange.cc 18415 (loop_cand::analyze_iloop_reduction_var): Update for change to 18416 check_reduction_path. 18417 (tree_loop_interchange::interchange): Update for change to 18418 find_loop_location. 18419 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for 18420 change in return-type of find_loop_location. 18421 (graphite_regenerate_ast_isl): Likewise. 18422 * graphite-optimize-isl.c (optimize_isl): Likewise. 18423 * graphite.c (graphite_transform_loops): Likewise. 18424 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to 18425 pass in a dump_location_t rather than a location_t, via the 18426 gimple stmt. 18427 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise. 18428 * ipa.c (walk_polymorphic_call_targets): Likewise. 18429 * loop-unroll.c (report_unroll): Convert "locus" param from 18430 location_t to dump_location_t. 18431 (decide_unrolling): Update for change to get_loop_location's 18432 return type. 18433 * omp-grid.c (struct grid_prop): Convert field "target_loc" from 18434 location_t to dump_user_location_t. 18435 (grid_find_single_omp_among_assignments_1): Updates calls to 18436 dump_printf_loc to pass in a dump_location_t rather than a 18437 location_t, via the gimple stmt. 18438 (grid_parallel_clauses_gridifiable): Convert "tloc" from 18439 location_t to dump_location_t. Updates calls to dump_printf_loc 18440 to pass in a dump_location_t rather than a location_t, via the 18441 gimple stmt. 18442 (grid_inner_loop_gridifiable_p): Likewise. 18443 (grid_dist_follows_simple_pattern): Likewise. 18444 (grid_gfor_follows_tiling_pattern): Likewise. 18445 (grid_target_follows_gridifiable_pattern): Likewise. 18446 (grid_attempt_target_gridification): Convert initialization 18447 of local "grid" from memset to zero-initialization; FIXME: does 18448 this require C++11? Update call to dump_printf_loc to pass in a 18449 optinfo_location rather than a location_t, via the gimple stmt. 18450 * profile.c (read_profile_edge_counts): Updates call to 18451 dump_printf_loc to pass in a dump_location_t rather than a 18452 location_t 18453 (compute_branch_probabilities): Likewise. 18454 * selftest-run-tests.c (selftest::run_tests): Call 18455 dumpfile_c_tests. 18456 * selftest.h (dumpfile_c_tests): New decl. 18457 * tree-loop-distribution.c (pass_loop_distribution::execute): 18458 Update for change in return type of find_loop_location. 18459 * tree-parloops.c (parallelize_loops): Likewise. 18460 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert 18461 "locus" from location_t to dump_user_location_t. 18462 (canonicalize_loop_induction_variables): Likewise. 18463 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update 18464 for change in return type of find_loop_location. 18465 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call 18466 to dump_printf_loc to pass in a dump_location_t rather than a 18467 location_t, via the stmt. 18468 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children): 18469 Likewise. 18470 * tree-vect-loop-manip.c (find_loop_location): Convert return 18471 type from source_location to dump_user_location_t. 18472 (vect_do_peeling): Update for above change. 18473 (vect_loop_versioning): Update for change in type of 18474 vect_location. 18475 * tree-vect-loop.c (check_reduction_path): Convert "loc" param 18476 from location_t to dump_user_location_t. 18477 (vect_estimate_min_profitable_iters): Update for change in type 18478 of vect_location. 18479 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from 18480 location_t to dump_location_t. 18481 (vect_slp_bb): Update for change in type of vect_location. 18482 * tree-vectorizer.c (vect_location): Convert from source_location 18483 to dump_user_location_t. 18484 (try_vectorize_loop_1): Update for change in vect_location's type. 18485 (vectorize_loops): Likewise. 18486 (increase_alignment): Likewise. 18487 * tree-vectorizer.h (vect_location): Convert from source_location 18488 to dump_user_location_t. 18489 (find_loop_location): Convert return type from source_location to 18490 dump_user_location_t. 18491 (check_reduction_path): Convert 1st param from location_t to 18492 dump_user_location_t. 18493 * value-prof.c (check_counter): Update call to dump_printf_loc to 18494 pass in a dump_user_location_t rather than a location_t; update 18495 call to error_at for change in type of "locus". 18496 (check_ic_target): Update call to dump_printf_loc to 18497 pass in a dump_user_location_t rather than a location_t, via the 18498 call_stmt. 18499 185002018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com> 18501 18502 * config/s390/s390.h (enum processor_flags): Do not use 18503 default tune parameter when -march was specified. 18504 185052018-06-26 Jakub Jelinek <jakub@redhat.com> 18506 18507 PR target/86314 18508 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s): 18509 Check reg_overlap_mentioned_p in addition to reg_set_p with the same 18510 operands. 18511 185122018-06-26 Richard Biener <rguenther@suse.de> 18513 18514 PR tree-optimization/86287 18515 PR bootstrap/86316 18516 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free. 18517 (vect_analyze_loop): Initialize n_stmts. 18518 185192018-06-26 Richard Biener <rguenther@suse.de> 18520 18521 PR middle-end/86271 18522 * fold-const.c (fold_convertible_p): Pointer extension 18523 isn't valid. 18524 185252018-06-26 Alexandre Oliva <aoliva@redhat.com> 18526 18527 PR debug/86064 18528 * dwarf2out.c (loc_list_has_views): Adjust comments. 18529 (dw_loc_list): Split single cross-partition range with 18530 nonzero locview. 18531 185322018-06-25 Jeff Law <law@redhat.com> 18533 18534 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn 18535 on -mbig-switch by default. 18536 18537 * config/v850/predicates.md (const_float_1_operand): Fix match_code 18538 test. 18539 (const_float_0_operand): Remove unused predicate. 18540 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP. 18541 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants. 18542 (recipsf2): New expander. Original pattern now called 18543 (recipsf2_insn). 18544 (recipdf2, recipdf2_insn): Similarly. 18545 (rsqrtsf2, rsqrtsf2_insn): Similarly 18546 (rsqrtdf2, rsqrtdf2_insn): Similarly 18547 185482018-06-26 Gerald Pfeifer <gerald@pfeifer.com> 18549 18550 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5. 18551 Simplify logic for FreeBSD (twice). 18552 185532018-06-25 Martin Sebor <msebor@redhat.com> 18554 18555 PR tree-optimization/86204 18556 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing 18557 a strnlen result if it's less than the length of the string. 18558 185592018-06-25 Martin Sebor <msebor@redhat.com> 18560 18561 PR tree-optimization/85700 18562 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment. 18563 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction. 18564 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy. 18565 185662018-06-25 Martin Sebor <msebor@redhat.com> 18567 18568 * doc/extend.texi (Zero-length arrays): Update and clarify. 18569 185702018-06-25 Michael Meissner <meissner@linux.ibm.com> 18571 18572 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that 18573 added IEEE/IBM long double multilib support on PowerPC little 18574 endian Linux systems. 18575 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise. 18576 (MULTILIB_DEFAULTS): Likewise. 18577 * config/rs6000/rs6000.c (rs6000_option_override_internal): 18578 Likewise. 18579 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise. 18580 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used. 18581 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used. 18582 185832018-06-25 Alexander Monakov <amonakov@ispras.ru> 18584 18585 PR middle-end/86311 18586 * sort.cc (REORDER_23): Avoid memcpy with same destination and source. 18587 (REORDER_45): Likewise. 18588 185892018-06-25 Jeff Law <law@redhat.com> 18590 18591 * config/v850/v850.md (divmodhi4): Make sure to sign extend the 18592 dividend to 32 bits. Adjust length. 18593 (udivmodhi4): Cleanup output template. Fix length. 18594 185952018-06-25 Carl Love <cel@us.ibm.com> 18596 18597 * config/rs6000/vsx.md: Change word selector to prefered location. 18598 185992018-06-25 Richard Biener <rguenther@suse.de> 18600 18601 PR tree-optimization/86304 18602 * tree-vectorizer.c (vectorize_loops): Walk over new possibly 18603 epilogue-if-converted loops as well. 18604 186052018-06-25 Jan Hubicka <hubicka@ucw.cz> 18606 18607 * lto-section-out.c (lto_begin_section): Do not print section 18608 name for noaddr and unnumbered dumps. 18609 186102018-06-25 Richard Biener <rguenther@suse.de> 18611 18612 * tree-vectorizer.h (struct vec_info_shared): New structure 18613 with parts split out from struct vec_info and loop_nest from 18614 struct _loop_vec_info. 18615 (struct vec_info): Adjust accordingly. 18616 (struct _loop_vec_info): Likewise. 18617 (LOOP_VINFO_LOOP_NEST): Adjust. 18618 (LOOP_VINFO_DATAREFS): Likewise. 18619 (LOOP_VINFO_DDRS): Likewise. 18620 (struct _bb_vec_info): Likewise. 18621 (BB_VINFO_DATAREFS): Likewise. 18622 (BB_VINFO_DDRS): Likewise. 18623 (struct _stmt_vec_info): Add dr_aux member. 18624 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo. 18625 (DR_MISALIGNMENT_UNINITIALIZED): New. 18626 (set_dr_misalignment): Adjust. 18627 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED. 18628 (vect_analyze_loop): Adjust prototype. 18629 (vect_analyze_loop_form): Likewise. 18630 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): 18631 Compute dependences lazily. 18632 (vect_record_base_alignments): Use shared datarefs/ddrs. 18633 (vect_verify_datarefs_alignment): Likewise. 18634 (vect_analyze_data_refs_alignment): Likewise. 18635 (vect_analyze_data_ref_accesses): Likewise. 18636 (vect_analyze_data_refs): Likewise. 18637 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add 18638 constructor parameter for shared part. 18639 (vect_analyze_loop_form): Pass in shared part and adjust. 18640 (vect_analyze_loop_2): Pass in storage for the number of 18641 stmts. Move loop nest finding to the caller. Compute 18642 datarefs lazily. 18643 (vect_analyze_loop): Pass in shared part. 18644 (vect_transform_loop): Verify shared datarefs are unchanged. 18645 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add 18646 constructor parameter for shared part. 18647 (vect_slp_analyze_bb_1): Pass in shared part and adjust. 18648 (vect_slp_bb): Verify shared datarefs are unchanged before 18649 transform. 18650 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX 18651 change. 18652 (new_stmt_vec_info): Initialize DR_AUX misalignment to 18653 DR_MISALIGNMENT_UNINITIALIZED. 18654 * tree-vectorizer.c (vec_info::vec_info): Add constructor 18655 parameter for shared part. 18656 (vec_info::~vec_info): Adjust. 18657 (vec_info_shared::vec_info_shared): New. 18658 (vec_info_shared::~vec_info_shared): Likewise. 18659 (vec_info_shared::save_datarefs): Likewise. 18660 (vec_info_shared::check_datarefs): Likewise. 18661 (try_vectorize_loop_1): Construct shared part live for analyses 18662 of a single loop for multiple vector sizes. 18663 * tree-parloops.c (gather_scalar_reductions): Adjust. 18664 186652018-06-25 Richard Biener <rguenther@suse.de> 18666 18667 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify 18668 DR for SIMD lane accesses here and mark DR with (void *)-1 aux. 18669 (vect_analyze_data_refs): Remove similar code from here and 18670 simplify accordingly. 18671 186722018-06-25 Richard Biener <rguenther@suse.de> 18673 18674 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail 18675 for reverse storage order accesses rather than asserting 18676 they cannot happen here. 18677 186782018-06-25 Tom de Vries <tdevries@suse.de> 18679 18680 PR debug/86257 18681 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"): 18682 Use data16 instead of .byte for insn prefix. 18683 186842018-06-25 Andreas Krebbel <krebbel@linux.ibm.com> 18685 18686 PR C++/86082 18687 * parser.c (make_char_string_pack): Pass this literal chars 18688 through cpp_interpret_string. 18689 (cp_parser_userdef_numeric_literal): Check the result of 18690 make_char_string_pack. 18691 186922018-06-24 Maya Rashish <coypu@sdf.org> 18693 18694 * ginclude/stddef.h: Simplify conditions around avoiding 18695 re-definition of __size_t. 18696 186972018-06-22 Jan Hubicka <hubicka@ucw.cz> 18698 18699 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable 18700 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL. 18701 187022018-06-22 Maya Rashish <coypu@sdf.org> 18703 18704 * doc/invoke.texi (mno-fancy-math-387): Update for changes 18705 made to OpenBSD and NetBSD through the years. 18706 187072018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org> 18708 18709 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change 18710 behavior of vec_pack (vector double, vector double) to match 18711 behavior of vec_float2 (vector double, vector double). 18712 187132018-06-22 Olivier Hainque <hainque@adacore.com> 18714 18715 * gimplify.c (gimplify_function_tree): Prevent creation 18716 of a trampoline for the address of the current function 18717 passed to entry/exit instrumentation hooks. 18718 187192018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com> 18720 18721 PR target/86222 18722 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32 18723 correctly. 18724 187252018-06-22 Martin Liska <mliska@suse.cz> 18726 18727 PR tree-optimization/86263 18728 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables): 18729 Bail out if is_enabled is false. 18730 * tree-switch-conversion.h (jump_table_cluster::is_enabled): 18731 New declaration. 18732 (jump_table_cluster::is_enabled): New function. 18733 187342018-06-22 Jan Hubicka <hubicka@ucw.cz> 18735 18736 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream 18737 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD. 18738 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise. 18739 (lto_input_ts_binfo_tree_pointers): Likewise. 18740 * tree-streamer-out.c (streamer_write_tree_bitfields, 18741 write_ts_binfo_tree_pointers): Likewise. 18742 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD. 18743 187442018-06-22 Jan Hubicka <hubicka@ucw.cz> 18745 18746 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs. 18747 187482018-06-22 Martin Liska <mliska@suse.cz> 18749 18750 * symbol-summary.h (get): Make it pure and inline move 18751 functionality from ::get function. 18752 (get): Remove and inline into ::get and ::get_create. 18753 (get_create): Move code from ::get function. 18754 187552018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE> 18756 18757 PR target/85994 18758 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for 18759 -x assembler-with-cpp. 18760 187612018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE> 18762 18763 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define 18764 _FILE_OFFSET_BITS=64 for C++. 18765 187662018-06-21 Michael Meissner <meissner@linux.ibm.com> 18767 18768 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit 18769 conversion insn that shows up when pr85657-3.c is compiled using 18770 IEEE 128-bit long double. 18771 (neg<mode>2_internal): Use the correct mode to check whether the 18772 mode is IBM extended. 18773 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex 18774 multiply and divide external functions from being created more 18775 than once. 18776 187772018-06-21 Eric Botcazou <ebotcazou@adacore.com> 18778 18779 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P 18780 functions. 18781 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of 18782 the edge can be forwarded. 18783 (cfg_layout_merge_blocks): Likewise. 18784 187852018-06-21 Eric Botcazou <ebotcazou@adacore.com> 18786 18787 * except.c (finish_eh_generation): Commit edge insertions only after 18788 the EH edges have been redirected from post-landing to landing pads. 18789 187902018-06-21 Eric Botcazou <ebotcazou@adacore.com> 18791 18792 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of 18793 create_tmp_var_for to create the FRAME decl. 18794 (finalize_nesting_tree_1): Do not unchain the FRAME decl. 18795 187962018-06-21 Eric Botcazou <ebotcazou@adacore.com> 18797 18798 * tree-inline.c (copy_edges_for_bb): Minor tweak. 18799 (maybe_move_debug_stmts_to_successors): Also reset the locus of the 18800 debug statement when resetting its value. 18801 (expand_call_inline): Copy the locus of the call onto the assignment 18802 of the return value, if any. Use local variable in more cases. 18803 188042018-06-21 Martin Liska <mliska@suse.cz> 18805 18806 * ipa-pure-const.c (propagate_nothrow): Use 18807 funct_state_summaries->get. 18808 (dump_malloc_lattice): Likewise. 18809 (propagate_malloc): Likewise. 18810 188112018-06-21 Richard Biener <rguenther@suse.de> 18812 18813 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated 18814 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally. 18815 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update 18816 comment. 18817 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream 18818 BLOCK_ABSTRACT_ORIGIN unconditionally. 18819 188202018-06-21 David Malcolm <dmalcolm@redhat.com> 18821 18822 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after 18823 deleting it. 18824 * ipa-reference.c (ipa_reference_c_finalize): Delete 18825 ipa_ref_opt_sum_summaries and set it to NULL. 18826 188272018-06-21 Tom de Vries <tdevries@suse.de> 18828 18829 PR tree-optimization/85859 18830 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call 18831 test with comment from bb_no_side_effects_p. 18832 188332018-06-21 Richard Biener <rguenther@suse.de> 18834 18835 PR tree-optimization/86232 18836 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust 18837 max for constant niter. 18838 188392018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com> 18840 18841 * config/aarch64/aarch64-simd.md 18842 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New. 18843 188442018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com> 18845 18846 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi): 18847 Make opernads of the unspec commutative. 18848 188492018-06-21 Richard Biener <rguenther@suse.de> 18850 18851 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP. 18852 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr): 18853 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters. 18854 (vect_analyze_data_ref_dependence): Re-order checks to deal with 18855 NULL DR_STEP. 18856 (vect_record_base_alignments): Do not record base alignment 18857 for gathers or scatters. 18858 (vect_compute_data_ref_alignment): Drop return value that is always 18859 true. Bail out early for gathers or scatters. 18860 (vect_enhance_data_refs_alignment): Bail out early for gathers 18861 or scatters. 18862 (vect_find_same_alignment_drs): Likewise. 18863 (vect_analyze_data_refs_alignment): Remove dead code. 18864 (vect_slp_analyze_and_verify_node_alignment): Likewise. 18865 (vect_analyze_data_refs): For possible gathers or scatters do 18866 not create an alternate DR, just check their possible validity 18867 and mark them. Adjust DECL_NONALIASED handling to not rely 18868 on DR_BASE_ADDRESS. 18869 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not 18870 update inits of gathers or scatters. 18871 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): 18872 Also copy gather/scatter flag to pattern vinfo. 18873 188742018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org> 18875 18876 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change 18877 behavior of vec_packsu (vector unsigned long long, vector unsigned 18878 long long) to match behavior of vec_packs with same signature. 18879 188802018-06-20 Chung-Lin Tang <cltang@codesourcery.com> 18881 Thomas Schwinge <thomas@codesourcery.com> 18882 Cesar Philippidis <cesar@codesourcery.com> 18883 18884 * gimplify.c (gimplify_scan_omp_clauses): Add support for 18885 OMP_CLAUSE_{IF_PRESENT,FINALIZE}. 18886 (gimplify_adjust_omp_clauses): Likewise. 18887 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove 18888 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}. 18889 (gimplify_omp_target_update): Update handling of acc update and 18890 enter/exit data. 18891 * omp-low.c (install_var_field): Remove unused parameter 18892 base_pointers_restrict. 18893 (scan_sharing_clauses): Remove base_pointers_restrict parameter. 18894 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT, 18895 FINALIZE} 18896 (omp_target_base_pointers_restrict_p): Delete. 18897 (scan_omp_target): Update call to scan_sharing_clauses. 18898 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT, 18899 FINALIZE}. 18900 * tree-nested.c (convert_nonlocal_omp_clauses): Handle 18901 OMP_CLAUSE_{IF_PRESENT,FINALIZE}. 18902 (convert_local_omp_clauses): Likewise. 18903 * tree-pretty-print.c (dump_omp_clause): Likewise. 18904 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT, 18905 FINALIZE}. 18906 (omp_clause_code_name): Likewise. 18907 189082018-06-20 Jakub Jelinek <jakub@redhat.com> 18909 18910 PR debug/86194 18911 * var-tracking.c (use_narrower_mode_test): Check if shift amount can 18912 be narrowed. 18913 18914 PR tree-optimization/86231 18915 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and 18916 anti-range don't overwrite *vr0min before using it to compute *vr0max. 18917 189182018-06-20 Tom de Vries <tdevries@suse.de> 18919 18920 PR tree-optimization/86097 18921 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to 18922 iv type if signedness of iv type is not the same as that of *nit. 18923 189242018-06-20 Jakub Jelinek <jakub@redhat.com> 18925 18926 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any 18927 EDGE_EH edges, verify they are all EDGE_EH. 18928 189292018-06-20 Maya Rashish <coypu@sdf.org> 18930 18931 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD. 18932 189332018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 18934 18935 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New. 18936 * config/aarch64/aarch64.c (xgene1_tunings): Add 18937 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags. 18938 (aarch64_mode_valid_for_sched_fusion_p): 18939 Allow 16-byte modes. 18940 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p. 18941 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of 18942 128-bit modes. 18943 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>): 18944 New pattern. 18945 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise. 18946 * config/aarch64/iterators.md (VQ2): New mode iterator. 18947 189482018-06-20 Martin Liska <mliska@suse.cz> 18949 18950 * tree-switch-conversion.c (jump_table_cluster::can_be_handled): 18951 Change default ratio from 10 to 8. 18952 189532018-06-20 Martin Liska <mliska@suse.cz> 18954 18955 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables): 18956 New. 18957 (bit_test_cluster::find_bit_tests): Likewise. 18958 (switch_decision_tree::analyze_switch_statement): Find clusters. 18959 * tree-switch-conversion.h (struct jump_table_cluster): Document 18960 hierarchy. 18961 189622018-06-20 Martin Liska <mliska@suse.cz> 18963 18964 * tree-switch-conversion.c (switch_conversion::collect): 18965 Record m_uniq property. 18966 (switch_conversion::expand): Bail out for special conditions. 18967 (group_cluster::~group_cluster): New. 18968 (group_cluster::group_cluster): Likewise. 18969 (group_cluster::dump): Likewise. 18970 (jump_table_cluster::emit): New. 18971 (switch_decision_tree::fix_phi_operands_for_edges): New. 18972 (struct case_node): Remove struct. 18973 (jump_table_cluster::can_be_handled): New. 18974 (case_values_threshold): Moved to header. 18975 (reset_out_edges_aux): Likewise. 18976 (jump_table_cluster::is_beneficial): New. 18977 (bit_test_cluster::can_be_handled): Likewise. 18978 (add_case_node): Remove. 18979 (bit_test_cluster::is_beneficial): New. 18980 (case_bit_test::cmp): New. 18981 (bit_test_cluster::emit): New. 18982 (expand_switch_as_decision_tree_p): Remove. 18983 (bit_test_cluster::hoist_edge_and_branch_if_true): New. 18984 (fix_phi_operands_for_edge): Likewise. 18985 (switch_decision_tree::analyze_switch_statement): New. 18986 (compute_cases_per_edge): Move ... 18987 (switch_decision_tree::compute_cases_per_edge): ... here. 18988 (try_switch_expansion): Likewise. 18989 (switch_decision_tree::try_switch_expansion): Likewise. 18990 (record_phi_operand_mapping): Likewise. 18991 (switch_decision_tree::record_phi_operand_mapping): Likewise. 18992 (emit_case_decision_tree): Likewise. 18993 (switch_decision_tree::emit): Likewise. 18994 (balance_case_nodes): Likewise. 18995 (switch_decision_tree::balance_case_nodes): Likewise. 18996 (dump_case_nodes): Likewise. 18997 (switch_decision_tree::dump_case_nodes): Likewise. 18998 (emit_jump): Likewise. 18999 (switch_decision_tree::emit_jump): Likewise. 19000 (emit_cmp_and_jump_insns): Likewise. 19001 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise. 19002 (emit_case_nodes): Likewise. 19003 (switch_decision_tree::emit_case_nodes): Likewise. 19004 (conditional_probability): Remove. 19005 * tree-switch-conversion.h (enum cluster_type): New. 19006 (PRINT_CASE): New. 19007 (struct cluster): Likewise. 19008 (cluster::cluster): Likewise. 19009 (struct simple_cluster): Likewise. 19010 (simple_cluster::simple_cluster): Likewise. 19011 (struct group_cluster): Likewise. 19012 (struct jump_table_cluster): Likewise. 19013 (struct bit_test_cluster): Likewise. 19014 (struct min_cluster_item): Likewise. 19015 (struct case_tree_node): Likewise. 19016 (case_tree_node::case_tree_node): Likewise. 19017 (jump_table_cluster::case_values_threshold): Likewise. 19018 (struct case_bit_test): Likewise. 19019 (struct switch_decision_tree): Likewise. 19020 (struct switch_conversion): Likewise. 19021 (switch_decision_tree::reset_out_edges_aux): Likewise. 19022 190232018-06-20 Martin Liska <mliska@suse.cz> 19024 19025 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove. 19026 (hoist_edge_and_branch_if_true): Likewise. 19027 (expand_switch_using_bit_tests_p): Likewise. 19028 (struct case_bit_test): Likewise. 19029 (case_bit_test_cmp): Likewise. 19030 (emit_case_bit_tests): Likewise. 19031 (switch_conversion::switch_conversion): New class. 19032 (struct switch_conv_info): Remove old struct. 19033 (collect_switch_conv_info): More to ... 19034 (switch_conversion::collect): ... this. 19035 (check_range): Likewise. 19036 (switch_conversion::check_range): Likewise. 19037 (check_all_empty_except_final): Likewise. 19038 (switch_conversion::check_all_empty_except_final): Likewise. 19039 (check_final_bb): Likewise. 19040 (switch_conversion::check_final_bb): Likewise. 19041 (create_temp_arrays): Likewise. 19042 (switch_conversion::create_temp_arrays): Likewise. 19043 (free_temp_arrays): Likewise. 19044 (gather_default_values): Likewise. 19045 (switch_conversion::gather_default_values): Likewise. 19046 (build_constructors): Likewise. 19047 (switch_conversion::build_constructors): Likewise. 19048 (constructor_contains_same_values_p): Likewise. 19049 (switch_conversion::contains_same_values_p): Likewise. 19050 (array_value_type): Likewise. 19051 (switch_conversion::array_value_type): Likewise. 19052 (build_one_array): Likewise. 19053 (switch_conversion::build_one_array): Likewise. 19054 (build_arrays): Likewise. 19055 (switch_conversion::build_arrays): Likewise. 19056 (gen_def_assigns): Likewise. 19057 (switch_conversion::gen_def_assigns): Likewise. 19058 (prune_bbs): Likewise. 19059 (switch_conversion::prune_bbs): Likewise. 19060 (fix_phi_nodes): Likewise. 19061 (switch_conversion::fix_phi_nodes): Likewise. 19062 (gen_inbound_check): Likewise. 19063 (switch_conversion::gen_inbound_check): Likewise. 19064 (process_switch): Use the newly created class. 19065 (switch_conversion::expand): New. 19066 (switch_conversion::~switch_conversion): New. 19067 * tree-switch-conversion.h: New file. 19068 190692018-06-20 Richard Sandiford <richard.sandiford@arm.com> 19070 19071 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to 19072 tree-vect-patterns.c. 19073 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function. 19074 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument. 19075 (vect_recog_sad_pattern): Likewise. 19076 (vect_recog_widen_sum_pattern): Likewise. 19077 (vect_recog_pow_pattern): Likewise. Check for a null vectype. 19078 (vect_recog_widen_shift_pattern): Remove the type_in argument. 19079 (vect_recog_rotate_pattern): Likewise. 19080 (vect_recog_mult_pattern): Likewise. 19081 (vect_recog_vector_vector_shift_pattern): Likewise. 19082 (vect_recog_divmod_pattern): Likewise. 19083 (vect_recog_mixed_size_cond_pattern): Likewise. 19084 (vect_recog_bool_pattern): Likewise. 19085 (vect_recog_mask_conversion_pattern): Likewise. 19086 (vect_try_gather_scatter_pattern): Likewise. 19087 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype. 19088 (vect_recog_over_widening_pattern): Likewise. 19089 (vect_recog_gather_scatter_pattern): Likewise. 19090 (vect_recog_func_ptr): Move from tree-vectorizer.h 19091 (vect_vect_recog_func_ptrs): Move further down the file. 19092 (vect_recog_func): Likewise. Remove the third argument. 19093 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs. 19094 (vect_pattern_recog_1): Expect the pattern function to do any 19095 necessary target tests. Also expect it to provide a vector type. 19096 Remove the type_in handling. 19097 190982018-06-20 Richard Sandiford <richard.sandiford@arm.com> 19099 19100 * tree-vect-patterns.c (vect_pattern_detected): New function. 19101 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern) 19102 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern) 19103 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern 19104 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern) 19105 (vect_recog_mult_pattern, vect_recog_divmod_pattern) 19106 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern) 19107 (vect_recog_mask_conversion_pattern) 19108 (vect_try_gather_scatter_pattern): Likewise. 19109 191102018-06-20 Richard Sandiford <richard.sandiford@arm.com> 19111 19112 * tree-vect-patterns.c (vect_get_internal_def): New function. 19113 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern) 19114 (vect_recog_vector_vector_shift_pattern, check_bool_pattern) 19115 (search_type_for_mask_1): Use it. 19116 191172018-06-20 Richard Sandiford <richard.sandiford@arm.com> 19118 19119 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove 19120 redundant WIDEN_SUM_EXPR handling. 19121 (vect_recog_sad_pattern): Likewise. 19122 191232018-06-20 Richard Sandiford <richard.sandiford@arm.com> 19124 19125 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove 19126 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree. 19127 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR. 19128 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR. 19129 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR. 19130 191312018-06-20 Richard Sandiford <richard.sandiford@arm.com> 19132 19133 * tree-vect-stmts.c (vectorizable_call): Make sure that we 19134 use the stmt_vec_info of the original bb statement for the 19135 new zero assignment, even if the call is part of a pattern. 19136 191372018-06-20 Richard Sandiford <richard.sandiford@arm.com> 19138 19139 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq 19140 that the sequence is attached to the original statement rather 19141 than the pattern statement. 19142 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the 19143 PATTERN_DEF_SEQ from the original statement rather than 19144 the main pattern statement. 19145 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. 19146 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise. 19147 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ. 19148 191492018-06-20 Richard Sandiford <richard.sandiford@arm.com> 19150 19151 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern 19152 definition statements before the early exit for statements that aren't 19153 live or relevant. 19154 * tree-vect-loop.c (vect_transform_loop_stmt): New function, 19155 split out from... 19156 (vect_transform_loop): ...here. Process pattern definition 19157 statements without first checking whether the main pattern 19158 statement is live or relevant. 19159 191602018-06-19 Eric Botcazou <ebotcazou@adacore.com> 19161 19162 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at 19163 -O0 if the locus represent UNKNOWN_LOCATION but have different values. 19164 191652018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com> 19166 19167 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check 19168 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller. 19169 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions. 19170 (expand_block_compare): Change select_block_compare_mode call. 19171 (expand_strncmp_align_check): Use new functions, fix comment. 19172 (emit_final_str_compare_gpr): New function. 19173 (expand_strn_compare): Refactor and clean up code. 19174 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *. 19175 191762018-06-19 Tony Reix <tony.reix@atos.com> 19177 Damien Bergamini <damien.bergamini@atos.com> 19178 David Edelsohn <dje.gcc@gmail.com> 19179 19180 * collect2.c (static_obj): New variable. 19181 (static_libs): New variable. 19182 (is_in_list): Uncomment declaration. 19183 (main): Track AIX libraries linked statically. 19184 (is_in_list): Uncomment definition. 19185 (scan_prog_file): Don't add AIX shared libraries initializer 19186 to constructor list if linking statically. 19187 191882018-06-19 Max Filippov <jcmvbkbc@gmail.com> 19189 19190 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec 19191 constant. 19192 (allocate_stack, frame_blockage, *frame_blockage): New patterns. 19193 191942018-06-19 Jan Hubicka <hubicka@ucw.cz> 19195 19196 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from 19197 blocks. 19198 191992018-06-19 Martin Liska <mliska@suse.cz> 19200 19201 * config/i386/i386.c (ix86_can_inline_p): Do not use 19202 ipa_fn_summaries::get_create. 19203 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with 19204 get. 19205 (devirtualization_time_bonus): Likewise. 19206 (ipcp_propagate_stage): Likewise. 19207 * ipa-fnsummary.c (redirect_to_unreachable): Likewise. 19208 (edge_set_predicate): Likewise. 19209 (evaluate_conditions_for_known_args): Likewise. 19210 (evaluate_properties_for_edge): Likewise. 19211 (ipa_call_summary::reset): Tranform to ... 19212 (ipa_call_summary::~ipa_call_summary): ... this. 19213 (ipa_fn_summary::reset): Transform to ... 19214 (ipa_fn_summary::~ipa_fn_summary): ... this. 19215 (ipa_fn_summary_t::remove): Rename to ... 19216 (ipa_fn_summary_t::remove_callees): ... this. 19217 (ipa_fn_summary_t::duplicate): Use placement new 19218 instead of memory copy. 19219 (ipa_call_summary_t::duplicate): Likewise. 19220 (ipa_call_summary_t::remove): Remove. 19221 (dump_ipa_call_summary): Change get_create to get. 19222 (ipa_dump_fn_summary): Dump only when summary exists. 19223 (analyze_function_body): Use symbol_summary::get instead 19224 of get_create. 19225 (compute_fn_summary): Likewise. 19226 (estimate_edge_devirt_benefit): Likewise. 19227 (estimate_edge_size_and_time): Likewise. 19228 (inline_update_callee_summaries): Likewise. 19229 (remap_edge_change_prob): Likewise. 19230 (remap_edge_summaries): Likewise. 19231 (ipa_merge_fn_summary_after_inlining): Likewise. 19232 (write_ipa_call_summary): Likewise. 19233 (ipa_fn_summary_write): Likewise. 19234 (ipa_free_fn_summary): Likewise. 19235 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor. 19236 (struct ipa_call_summary): Likewise. 19237 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead 19238 of get_create. 19239 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise. 19240 (estimate_size_after_inlining): Likewise. 19241 (estimate_growth): Likewise. 19242 (growth_likely_positive): Likewise. 19243 * ipa-inline-transform.c (clone_inlined_nodes): Likewise. 19244 (inline_call): Likewise. 19245 * ipa-inline.c (caller_growth_limits): Likewise. 19246 (can_inline_edge_p): Likewise. 19247 (can_inline_edge_by_limits_p): Likewise. 19248 (compute_uninlined_call_time): Likewise. 19249 (compute_inlined_call_time): Likewise. 19250 (want_inline_small_function_p): Likewise. 19251 (edge_badness): Likewise. 19252 (update_caller_keys): Likewise. 19253 (update_callee_keys): Likewise. 19254 (inline_small_functions): Likewise. 19255 (inline_to_all_callers_1): Likewise. 19256 (dump_overall_stats): Likewise. 19257 (early_inline_small_functions): Likewise. 19258 (early_inliner): Likewise. 19259 * ipa-profile.c (ipa_propagate_frequency_1): Likewise. 19260 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise. 19261 * ipa-pure-const.c (malloc_candidate_p): Likewise. 19262 * ipa-split.c (execute_split_functions): Likewise. 19263 * symbol-summary.h: Likewise. 19264 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise. 19265 192662018-06-19 Richard Biener <rguenther@suse.de> 19267 19268 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ... 19269 (vectorize_loops): ... here. Fix dbgcnt handling. 19270 (try_vectorize_loop): Wrap try_vectorize_loop_1. 19271 192722018-06-19 Segher Boessenkool <segher@kernel.crashing.org> 19273 19274 PR target/86197 19275 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An 19276 ieee128 argument takes up only one (vector) register, not two (floating 19277 point) registers. 19278 192792018-06-19 Eric Botcazou <ebotcazou@adacore.com> 19280 19281 * gimplify.c (gimplify_init_constructor): Really never clear for an 19282 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set. 19283 192842018-06-19 Richard Biener <rguenther@suse.de> 19285 19286 PR tree-optimization/86179 19287 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up 19288 after failed recognition. 19289 192902018-06-18 Martin Sebor <msebor@redhat.com> 19291 19292 PR middle-end/85602 19293 * calls.c (maybe_warn_nonstring_arg): Handle strncat. 19294 * tree-ssa-strlen.c (is_strlen_related_p): Make extern. 19295 Handle integer subtraction. 19296 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments. 19297 * tree-ssa-strlen.h (is_strlen_related_p): Declare. 19298 192992018-06-18 David Malcolm <dmalcolm@redhat.com> 19300 19301 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd 19302 param from rtx to rtx_insn *. 19303 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn" 19304 param. 19305 (frv_ifcvt_modify_insn): Likwise. 19306 (frv_ifcvt_modify_final): Likwise for local "existing_insn", 19307 adding an as_a <rtx_insn *> cast. Likewise for local "insn". 19308 * config/mips/mips.c (r10k_insert_cache_barriers): Add an 19309 as_a <rtx_insn *> cast to local "unprotected_region" once 19310 it's been established that it's not NULL or pc_rtx. 19311 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen 19312 param "sethi" from rtx to rtx_insn *. 19313 (nds32_group_float_insns): Likewise for param "insn". 19314 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st 19315 param. 19316 (vax_output_int_subtract): Likewise. 19317 * config/vax/vax.c (vax_output_int_add): Likewise for param 19318 "insn". 19319 (vax_output_int_subtract): Likewise. 19320 * emit-rtl.c (set_insn_deleted): Likewise, removing cast. 19321 (emit_pattern_after): Likewise for param "after". 19322 (emit_insn_after): Likewise. 19323 (emit_jump_insn_after): Likewise. 19324 (emit_call_insn_after): Likewise. 19325 (emit_debug_insn_after): Likewise. 19326 (emit_pattern_before): Likewise for param "before". 19327 (emit_insn_before): Likewise. 19328 (emit_jump_insn_before): Likewise. 19329 * final.c (get_insn_template): Likewise for param "insn", removing 19330 a cast. 19331 * output.h (get_insn_template): Likewise for 2nd param. 19332 * rtl.h (emit_insn_before): Likewise. 19333 (emit_jump_insn_before): Likewise. 19334 (emit_debug_insn_before_noloc): Likewise. 19335 (emit_insn_after): Likewise. 19336 (emit_jump_insn_after): Likewise. 19337 (emit_call_insn_after): Likewise. 19338 (emit_debug_insn_after): Likewise. 19339 (set_insn_deleted): Likewise for param. 19340 193412018-06-18 Michael Meissner <meissner@linux.ibm.com> 19342 19343 PR target/85358 19344 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit 19345 floating point modes, so that IFmode is numerically greater than 19346 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE 19347 to declare the ordering. This prevents IFmode from being 19348 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0 19349 machine. Include rs6000-modes.h to share the fractional values 19350 between genmodes* and the rest of the compiler. 19351 (IFmode): Likewise. 19352 (KFmode): Likewise. 19353 (TFmode): Likewise. 19354 * config/rs6000/rs6000-modes.h: New file. 19355 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the 19356 meaning of rs6000_long_double_size so that 126..128 selects an 19357 appropriate 128-bit floating point type. 19358 (rs6000_option_override_internal): Likewise. 19359 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h. 19360 (TARGET_LONG_DOUBLE_128): Change the meaning of 19361 rs6000_long_double_size so that 126..128 selects an appropriate 19362 128-bit floating point type. 19363 (LONG_DOUBLE_TYPE_SIZE): Update comment. 19364 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the 19365 source and destination to match the standard usage. 19366 (truncifkf2): Likewise. 19367 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on 19368 ISA 2.07 to use an explicit clobber, instead of passing in a 19369 temporary. 19370 (copysign<mode>3_soft): Likewise. 19371 193722018-06-18 David Malcolm <dmalcolm@redhat.com> 19373 19374 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): 19375 Replace dump_printf_loc call with DUMP_VECT_SCOPE. 19376 (vect_slp_analyze_instance_dependence): Likewise. 19377 (vect_enhance_data_refs_alignment): Likewise. 19378 (vect_analyze_data_refs_alignment): Likewise. 19379 (vect_slp_analyze_and_verify_instance_alignment 19380 (vect_analyze_data_ref_accesses): Likewise. 19381 (vect_prune_runtime_alias_test_list): Likewise. 19382 (vect_analyze_data_refs): Likewise. 19383 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise. 19384 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise. 19385 (vect_analyze_scalar_cycles_1): Likewise. 19386 (vect_get_loop_niters): Likewise. 19387 (vect_analyze_loop_form_1): Likewise. 19388 (vect_update_vf_for_slp): Likewise. 19389 (vect_analyze_loop_operations): Likewise. 19390 (vect_analyze_loop): Likewise. 19391 (vectorizable_induction): Likewise. 19392 (vect_transform_loop): Likewise. 19393 * tree-vect-patterns.c (vect_pattern_recog): Likewise. 19394 * tree-vect-slp.c (vect_analyze_slp): Likewise. 19395 (vect_make_slp_decision): Likewise. 19396 (vect_detect_hybrid_slp): Likewise. 19397 (vect_slp_analyze_operations): Likewise. 19398 (vect_slp_bb): Likewise. 19399 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise. 19400 (vectorizable_bswap): Likewise. 19401 (vectorizable_call): Likewise. 19402 (vectorizable_simd_clone_call): Likewise. 19403 (vectorizable_conversion): Likewise. 19404 (vectorizable_assignment): Likewise. 19405 (vectorizable_shift): Likewise. 19406 (vectorizable_operation): Likewise. 19407 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro. 19408 194092018-06-18 Martin Sebor <msebor@redhat.com> 19410 19411 PR tree-optimization/81384 19412 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New. 19413 * builtins.c (expand_builtin_strnlen): New function. 19414 (expand_builtin): Call it. 19415 (fold_builtin_n): Avoid setting TREE_NO_WARNING. 19416 * builtins.def (BUILT_IN_STRNLEN): New. 19417 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN. 19418 Warn for bounds in excess of maximum object size. 19419 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing 19420 single-value ranges. Handle strnlen. 19421 (handle_builtin_strlen): Handle strnlen. 19422 (strlen_check_and_optimize_stmt): Same. 19423 * doc/extend.texi (Other Builtins): Document strnlen. 19424 194252018-06-18 Maya Rashish <coypu@sdf.org> 19426 19427 * config/alpha/openbsd.h (TARGET_DEFAULT): Define. 19428 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise. 19429 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise. 19430 19431 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from 19432 here to ... 19433 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here. 19434 194352018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org> 19436 19437 * tree.c (escaped_string::escape): Replace cast to char * by 19438 const_cast<char *> (unescaped). 19439 194402018-06-18 Nick Clifton <nickc@redhat.com> 19441 19442 PR 84195 19443 * tree.c (escaped_string): New class. Converts an unescaped 19444 string into its escaped equivalent. 19445 (warn_deprecated_use): Use the new class to convert the 19446 deprecation message, if present. 19447 (test_escaped_strings): New self test. 19448 (test_c_tests): Add test_escaped_strings. 19449 * doc/extend.texi (deprecated): Add a note that the 19450 deprecation message is affected by the -fmessage-length 19451 option, and that control characters will be escaped. 19452 (#pragma GCC error): Document this pragma. 19453 (#pragma GCC warning): Likewise. 19454 * doc/invoke.texi (-fmessage-length): Document this option's 19455 effect on the #warning and #error preprocessor directives and 19456 the deprecated attribute. 19457 194582018-06-18 Eric Botcazou <ebotcazou@adacore.com> 19459 19460 * tree.c (decl_value_expr_lookup): Revert latest change. 19461 (decl_value_expr_insert): Likewise. 19462 194632018-06-17 Eric Botcazou <ebotcazou@adacore.com> 19464 19465 * gimplify.c (nonlocal_vlas): Delete. 19466 (nonlocal_vla_vars): Likewise. 19467 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local 19468 referenced VLAs. 19469 (gimplify_body): Do not create and destroy nonlocal_vlas. 19470 * tree-nested.c: Include diagnostic.h. 19471 (use_pointer_in_frame): Tweak. 19472 (lookup_field_for_decl): Add assertion and declare the transformation. 19473 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an 19474 internal error when the reference is in a wrong context. Do not 19475 create a debug decl by default. 19476 (note_nonlocal_block_vlas): Delete. 19477 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it. 19478 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not 19479 create a debug decl by default. 19480 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the 19481 call is in a wrong context. 19482 (fixup_vla_decls): New function. 19483 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no 19484 debug variables were created. 19485 * tree.c (decl_value_expr_lookup): Add checking assertion. 19486 (decl_value_expr_insert): Likewise. 19487 194882018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org> 19489 19490 PR middle-end/82479 19491 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR. 19492 * tree-scalar-evolution.c (interpret_expr): Likewise. 19493 (expression_expensive_p): Likewise. 19494 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise. 19495 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New. 19496 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount. 19497 (ssa_defined_by_minus_one_stmt_p): New. 19498 194992018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org> 19500 19501 PR middle-end/64946 19502 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR. 19503 * config/i386/i386.c (ix86_add_stmt_cost): Likewise. 19504 * dojump.c (do_jump): Likewise. 19505 * expr.c (expand_expr_real_2): Check operand type's sign. 19506 * fold-const.c (const_unop): Handle ABSU_EXPR. 19507 (fold_abs_const): Likewise. 19508 * gimple-pretty-print.c (dump_unary_rhs): Likewise. 19509 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie. 19510 (strip_sign_op_1): Likesise. 19511 * match.pd: Add new pattern to generate ABSU_EXPR. 19512 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR. 19513 * tree-cfg.c (verify_gimple_assign_unary): Likewise. 19514 * tree-eh.c (operation_could_trap_helper_p): Likewise. 19515 * tree-inline.c (estimate_operator_cost): Likewise. 19516 * tree-pretty-print.c (dump_generic_node): Likewise. 19517 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise. 19518 * tree.def (ABSU_EXPR): New. 19519 195202018-06-16 Jakub Jelinek <jakub@redhat.com> 19521 19522 PR middle-end/86095 19523 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var, 19524 documented as preserved for backward compatibility only. 19525 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation. 19526 19527 PR rtl-optimization/86108 19528 * bb-reorder.c (create_forwarder_block): Renamed to ... 19529 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and 19530 jump from new landing pad to the second part. 19531 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad): 19532 Adjust callers. 19533 195342018-06-15 Jakub Jelinek <jakub@redhat.com> 19535 19536 PR middle-end/85878 19537 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P 19538 check from first store_expr, use to_mode instead of GET_MODE (to_rtx). 19539 Only call store_expr for halves if the mode is the same. 19540 19541 PR middle-end/86123 19542 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions. 19543 Fix up comment formatting. 19544 195452018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de> 19546 19547 * typed-splay-tree.h (typed_splay_tree::remove): New function. 19548 (typed_splay_tree::closure, 19549 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted. 19550 (typed_splay_tree::typed_splay_tree, 19551 typed_splay_tree::operator =): Declared private. 19552 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value, 19553 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL, 19554 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper, 19555 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right, 19556 typed_splay_tree::splay_tree_splay, 19557 typed_splay_tree::splay_tree_foreach_helper, 19558 typed_splay_tree::splay_tree_insert, 19559 typed_splay_tree::splay_tree_remove, 19560 typed_splay_tree::splay_tree_lookup, 19561 typed_splay_tree::splay_tree_predecessor, 19562 typed_splay_tree::splay_tree_successor, 19563 typed_splay_tree::splay_tree_min, 19564 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h. 19565 (typed_splay_tree::root, typed_splay_tree::comp, 19566 typed_splay_tree::delete_key, 19567 typed_splay_tree::delete_value): New data members. 19568 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for 19569 typed_splay_tree::remove. 19570 195712018-06-15 Matthew Fortune <matthew.fortune@mips.com> 19572 19573 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc, 19574 -mginv and -mno-ginv to the assembler. 19575 * config/mips/mips.opt (-mcrc): New option. 19576 (-mginv): Likewise. 19577 * doc/invoke.text (-mcrc): Document. 19578 (-mginv): Likewise. 19579 195802018-06-15 Nick Clifton <nickc@redhat.com> 19581 19582 PR 84195 19583 * tree.c (escaped_string): New class. Converts an unescaped 19584 string into its escaped equivalent. 19585 (warn_deprecated_use): Use the new class to convert the 19586 deprecation message, if present. 19587 (test_escaped_strings): New self test. 19588 (test_c_tests): Add test_escaped_strings. 19589 * doc/extend.texi (deprecated): Add a note that the 19590 deprecation message is affected by the -fmessage-length 19591 option, and that control characters will be escaped. 19592 (#pragma GCC error): Document this pragma. 19593 (#pragma GCC warning): Likewise. 19594 * doc/invoke.texi (-fmessage-length): Document this option's 19595 effect on the #warning and #error preprocessor directives and 19596 the deprecated attribute. 19597 195982018-06-15 Richard Biener <rguenther@suse.de> 19599 19600 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS 19601 here, also noting vector size used. 19602 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector 19603 size used in MSG_OPTIMIZED_LOCATIONS dump. 19604 (pass_slp_vectorize::execute): Adjust. 19605 196062018-06-15 Claudiu Zissulescu <claziss@synopsys.com> 19607 19608 PR target/85968 19609 * config/arc/arc.c (arc_return_address_register): Fix 19610 if-condition. 19611 196122018-06-15 Richard Biener <rguenther@suse.de> 19613 19614 PR middle-end/86159 19615 * tree-cfg.c (gimplify_build3): Do not strip sign conversions, 19616 leave useless conversion stripping to force_gimple_operand_gsi. 19617 (gimplify_build2): Likewise. 19618 (gimplify_build1): Likewise. 19619 196202018-06-15 Richard Biener <rguenther@suse.de> 19621 19622 PR middle-end/86076 19623 * tree-cfg.c (move_stmt_op): unshare invariant addresses 19624 before adjusting their block. 19625 196262018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de> 19627 19628 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom 19629 multilibs for *-*-rtems*. 19630 * config/riscv/t-rtems: New file. 19631 196322018-06-14 Jakub Jelinek <jakub@redhat.com> 19633 19634 PR middle-end/86122 19635 * match.pd ((A +- CST1) +- CST2): Punt if last resort 19636 unsigned_type_for returns NULL. 19637 19638 PR target/85945 19639 * lower-subreg.c (find_decomposable_subregs): Don't decompose float 19640 subregs of multi-word pseudos unless the float mode has word size. 19641 196422018-06-14 Richard Biener <rguenther@suse.de> 19643 19644 PR middle-end/86139 19645 * tree-vect-generic.c (build_word_mode_vector_type): Remove 19646 duplicate and harmful type_hash_canon. 19647 * tree.c (type_hash_canon): Assert we didn't find ourselves. 19648 196492018-06-14 Richard Biener <rguenther@suse.de> 19650 19651 PR ipa/86124 19652 * tree-ssa-struct-alias.c (create_variable_info_for): Handle 19653 NULL cgraph_node. 19654 196552018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de> 19656 19657 * config/rtems.h (STDINT_LONG32): Define. 19658 196592018-06-13 Matthew Fortune <matthew.fortune@mips.com> 19660 Prachi Godbole <prachi.godbole@imgtec.com> 19661 19662 * config/mips/mips-cpus.def: Define P6600. 19663 * config/mips/mips-tables.opt: Regenerate. 19664 * config/mips/mips.c (mips_ucbranch_type): New enum. 19665 (mips_rtx_cost_data): Add support for P6600. 19666 (mips_issue_rate): Likewise. 19667 (mips_multipass_dfa_lookahead): Likewise. 19668 (mips_avoid_hazard): Likewise. 19669 (mips_reorg_process_insns): Likewise. 19670 (mips_classify_branch_p6600): New function. 19671 * config/mips/mips.h (TUNE_P6600): New define. 19672 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600. 19673 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600. 19674 * config/mips/mips.md: Include p6600.md. 19675 (processor): Add p6600. 19676 * config/mips/p6600.md: New file. 19677 * doc/invoke.texi: Add p6600 to supported architectures. 19678 196792018-06-13 Martin Sebor <msebor@redhat.com> 19680 19681 PR tree-optimization/86114 19682 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS 19683 of integer types. 19684 * tree-ssa-strlen.c (maybe_set_strlen_range): Same. 19685 196862018-06-13 Richard Biener <rguenther@suse.de> 19687 19688 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern): 19689 Properly set vector type of the intermediate stmt. 19690 * tree-vect-stmts.c (vectorizable_operation): The destination 19691 var always has vectype_out type. 19692 196932018-06-13 Jeff Law <law@redhat.com> 19694 19695 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than 19696 integer 0 for argument to print_rtl_with_bb. 19697 (rl78_reorg): Likewise. 19698 196992018-06-13 David Malcolm <dmalcolm@redhat.com> 19700 19701 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label" 19702 from rtx to rtx_insn *. 19703 * config/bfin/bfin.c (hwloop_optimize): Likewise for local 19704 "label". 19705 (add_sched_insns_for_speculation): Likewise for local "target", 19706 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN. 19707 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels" 19708 from rtx_insn ** to rtx_code_label **. 19709 (reorg_emit_nops): Likewise. 19710 (c6x_reorg): Likewise for local "call_labels". 19711 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from 19712 rtx to rtx_insn *. 19713 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to 19714 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to 19715 the loops over LABEL_REFS. 19716 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of 19717 braf_label. 19718 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN. 19719 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *. 19720 (split_branches): Strengthen local "olabel" from rtx to 19721 rtx_insn *, adding a safe_as_a cast. 19722 * emit-rtl.c (next_real_insn): Strengthen param from "rtx" 19723 to "rtx_insn *". 19724 (add_insn_after): Likewise for first two params. 19725 (add_insn_before): Likewise. 19726 (remove_insn): Likewise for param. 19727 (emit_pattern_before_noloc): Likewise for second and third params. 19728 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL. 19729 (emit_call_insn_before_noloc): Likewise. 19730 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx" 19731 to "rtx_insn *". 19732 (emit_barrier_before): Likewise. 19733 (emit_label_before): Strengthen "label" param from "rtx" to 19734 "rtx_code_label *". Strengthen "before" param from "rtx" to 19735 "rtx_insn *". 19736 (emit_insn_after_1): Strengthen "after" param from "rtx" to 19737 "rtx_insn *". 19738 (emit_pattern_after_noloc): Likewise. 19739 (emit_insn_after_noloc): Likewise. 19740 (emit_jump_insn_after_noloc): Likewise. 19741 (emit_call_insn_after_noloc): Likewise. 19742 (emit_debug_insn_after_noloc): Likewise. 19743 (emit_barrier_after): Likewise. 19744 (emit_label_after): Likewise for both params. 19745 (emit_pattern_after_setloc): Likewise for "after" param. Convert 19746 "loc" param from "int" to "location_t". 19747 (emit_insn_after_setloc): Likewise. 19748 (emit_jump_insn_after_setloc): Likewise. 19749 (emit_call_insn_after_setloc): Likewise. 19750 (emit_debug_insn_after_setloc): Likewise. 19751 (emit_pattern_before_setloc): Likewise for "before" param. Convert 19752 "loc" param from "int" to "location_t". 19753 (emit_pattern_before): Convert NULL_RTX to NULL. 19754 (emit_insn_before_setloc): Convert "loc" param from "int" to 19755 "location_t". 19756 (emit_jump_insn_before_setloc): Likewise. 19757 (emit_call_insn_before_setloc): Likewise. 19758 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to 19759 rtx_insn *. Convert "loc" param from "int" to "location_t". 19760 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc, 19761 emit_call_insn_before_setloc, emit_debug_insn_before_setloc): 19762 Convert 3rd param from "int" to "location_t". 19763 (emit_barrier_before, emit_barrier_after, next_real_insn): 19764 Strengthen param from rtx to rtx_insn *. 19765 (emit_label_before): Strengthen 1st param from "rtx" to 19766 "rtx_code_label *". Strengthen 2nd param from "rtx" to 19767 "rtx_insn *". 19768 (emit_insn_after_noloc, emit_jump_insn_after_noloc, 19769 emit_call_insn_after_noloc, emit_debug_insn_after_noloc): 19770 Strengthen 2nd param from "rtx" to "rtx_insn *". 19771 (emit_insn_after_setloc, emit_jump_insn_after_setloc) 19772 emit_call_insn_after_setloc, emit_debug_insn_after_setloc): 19773 Likewise. Convert 3rd param from "int" to "location_t". 19774 (emit_label_after): Strengthen 1st param from "rtx" to 19775 "rtx_code_label *". 19776 (next_real_insn, remove_insn): Strengthen param from "rtx" to 19777 "rtx_insn *". 19778 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params 19779 from "rtx" to "rtx_insn *". 19780 197812018-06-13 Jan Hubicka <hubicka@gcc.gnu.org> 19782 19783 * cgraph.c (cgraph_node::get_untransformed_body): Dump function 19784 bodies streamed in with -Q. 19785 * dumpfile.c (dump_files): Add lto-stream-out dump file. 19786 * dumpfile.h (tree_dump_index): Add lto_stream_out. 19787 * gimple-streamer-out.c: Include gimple-pretty-print.h 19788 (output_bb): Dump stmts streamed. 19789 * lto-section-out.c: Include print-tree.h 19790 (lto_begin_section): Dump sections created. 19791 (lto_output_decl_index): Dump decl encoded. 19792 * lto-streamer-out.c: Include print-tree.h 19793 (create_output_block): Dump output block created. 19794 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed. 19795 (output_function): Dump function output. 19796 (output_constructor): Dump constructor streamed. 19797 (write_global_stream): Output indexes encoded. 19798 (produce_asm_for_decls): Dump streams encoded. 19799 * lto-streamer.c (streamer_dump_file): New global var. 19800 * lto-streamer.h (streamer_dump_file): Declare. 19801 * passes.c (ipa_write_summaries): Initialize streamer dump. 19802 * varpool.c (varpool_node::get_constructor): Dump constructors streamed 19803 in. 19804 198052018-06-13 Eric Botcazou <ebotcazou@adacore.com> 19806 19807 PR target/86048 19808 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative 19809 offsets for register save directives. Emit a second batch of save 19810 directives, if need be, when the function accesses prior frames. 19811 198122018-06-12 Claudiu Zissulescu <claziss@synopsys.com> 19813 19814 * config/arc/fpu.md (fmasf4): Force operand to register. 19815 (fnmasf4): Likewise. 19816 198172018-06-12 Claudiu Zissulescu <claziss@synopsys.com> 19818 19819 * config/arc/arc-protos.h (arc_pad_return): Remove. 19820 * config/arc/arc.c (machine_function): Remove force_short_suffix 19821 and size_reason. 19822 (arc_print_operand): Adjust printing of '&'. 19823 (arc_verify_short): Remove conditional printing of short suffix. 19824 (arc_final_prescan_insn): Remove reference to size_reason. 19825 (pad_return): New function. 19826 (arc_reorg): Call pad_return. 19827 (arc_pad_return): Remove. 19828 (arc_init_machine_status): Remove reference to force_short_suffix. 19829 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE. 19830 (attr length): When attribute iscompact is true force to 2 19831 regardless; in the case of maybe check if we want to force the 19832 instruction to have 4 bytes length. 19833 (nopv): Change it to generate 4 byte long nop as well. 19834 (blockage): New pattern. 19835 (simple_return): Remove call to arc_pad_return. 19836 (p_return_i): Likewise. 19837 198382018-06-12 Claudiu Zissulescu <claziss@synopsys.com> 19839 19840 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define. 19841 198422018-06-12 Claudiu Zissulescu <claziss@synopsys.com> 19843 19844 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all 19845 ARC cores. 19846 198472018-06-12 Claudiu Zissulescu <claziss@synopsys.com> 19848 19849 * config/arc/arc.c (atomic_exchangesi): EX instruction is default 19850 for ARC700 and ARCv2. 19851 198522018-06-13 Chenghua Xu <paul.hua.gm@gmail.com> 19853 19854 PR target/86076 19855 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for 19856 operands[2] instead of operands[1]. 19857 19858 198592018-06-12 Richard Sandiford <richard.sandiford@linaro.org> 19860 19861 * lra-constraints.c (simplify_operand_subreg): In the paradoxical 19862 case, check whether the outer register overlaps an unallocatable 19863 register, not just whether it fits the required class. 19864 198652018-06-12 Richard Sandiford <richard.sandiford@linaro.org> 19866 19867 * poly-int.h (can_div_trunc_p): Add new overload in which all values 19868 are poly_ints. 19869 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p. 19870 (memrefs_conflict_p): Likewise. 19871 (init_alias_analysis): Likewise. 19872 * cfgexpand.c (expand_debug_expr): Likewise. 19873 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise. 19874 * cse.c (fold_rtx): Likewise. 19875 * explow.c (adjust_stack, anti_adjust_stack): Likewise. 19876 * expr.c (emit_block_move_hints): Likewise. 19877 (clear_storage_hints, push_block, emit_push_insn): Likewise. 19878 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise. 19879 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets. 19880 (emit_group_store): Likewise. 19881 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64 19882 to read the PRE/POST_MODIFY increment. 19883 * calls.c (store_one_arg): Use strip_offset. 19884 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to 19885 poly_int_rtx_p. 19886 (set_noop_p): Use rtx_to_poly_int64 for the elements selected 19887 by a VEC_SELECT. 19888 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset. 19889 (simplify_binary_operation_1): Extend CONST_INT handling to 19890 poly_int_rtx_p. 19891 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather 19892 than a HOST_WIDE_INT. 19893 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to 19894 poly_int64. 19895 (adjust_mems, add_stores): Update accodingly. 19896 (vt_canonicalize_addr): Track polynomial offsets. 19897 (emit_note_insn_var_location): Likewise. 19898 (vt_add_function_parameter): Likewise. 19899 (vt_initialize): Likewise. 19900 199012018-06-12 Jeff Law <law@redhat.com> 19902 19903 * config.gcc (alpha*-*-freebsd*): Remove. 19904 * config/alpha/freebsd.h: Remove. 19905 199062018-06-12 David Malcolm <dmalcolm@redhat.com> 19907 19908 PR other/69968 19909 * spellcheck-tree.c (levenshtein_distance): Rename to... 19910 (get_edit_distance): ...this, and update for underlying renaming. 19911 * spellcheck-tree.h (levenshtein_distance): Rename to... 19912 (get_edit_distance): ...this. 19913 * spellcheck.c (levenshtein_distance): Rename to... 19914 (get_edit_distance): ...this. Convert from Levenshtein distance 19915 to Damerau-Levenshtein distance by supporting transpositions of 19916 adjacent characters. Rename "v1" to "v_next" and "v0" to 19917 "v_one_ago". 19918 (selftest::levenshtein_distance_unit_test_oneway): Rename to... 19919 (selftest::test_edit_distance_unit_test_oneway): ...this, and 19920 update for underlying renaming. 19921 (selftest::levenshtein_distance_unit_test): Rename to... 19922 (selftest::test_get_edit_distance_unit): ...this, and update for 19923 underlying renaming. 19924 (selftest::test_find_closest_string): Add example from PR 69968 19925 where transposition helps 19926 (selftest::test_metric_conditions): Update for renaming. 19927 (selftest::test_metric_conditions): Likewise. 19928 (selftest::spellcheck_c_tests): Likewise. 19929 * spellcheck.h (levenshtein_distance): Rename both overloads to... 19930 (get_edit_distance): ...this. 19931 (best_match::consider): Update for renaming. 19932 199332018-06-12 Martin Sebor <msebor@redhat.com> 19934 19935 PR tree-optimization/85259 19936 * builtins.c (compute_objsize): Handle constant offsets. 19937 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return 19938 true iff a warning has been issued. 19939 * gimple.h (gimple_nonartificial_location): New function. 19940 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call 19941 gimple_nonartificial_location and handle -Wno-system-headers. 19942 (handle_builtin_stxncpy): Same. 19943 199442018-06-12 Martin Sebor <msebor@redhat.com> 19945 19946 PR c/85931 19947 * fold-const.c (operand_equal_p): Handle SAVE_EXPR. 19948 199492018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com> 19950 19951 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add 19952 BUILTIN_VEC_XST entries for pointer to double and long long. 19953 199542018-06-12 H.J. Lu <hongjiu.lu@intel.com> 19955 19956 PR target/85990 19957 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET): 19958 Update comments. 19959 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET): 19960 Likewise. 19961 199622018-06-12 Martin Liska <mliska@suse.cz> 19963 19964 * doc/options.texi: Document IntegerRange. 19965 199662018-06-12 Martin Liska <mliska@suse.cz> 19967 19968 * config/i386/i386.opt: Make MPX-related options as Deprecated. 19969 * opt-functions.awk: Handle Deprecated flag. 19970 * opts-common.c (decode_cmdline_option): Handle cl_deprecated 19971 and report error. 19972 (read_cmdline_option): Report warning for a deprecated option. 19973 * opts.h (struct cl_option): Add new field cl_deprecated. 19974 (CL_ERR_DEPRECATED): New. 19975 199762018-06-12 Martin Liska <mliska@suse.cz> 19977 19978 * doc/options.texi: Document Deprecated option flag. 19979 199802018-06-12 Claudiu Zissulescu <claziss@synopsys.com> 19981 19982 * config/arc/arc-arch.h (arc_extras): New enum. 19983 (arc_cpu_t):Add field extra. 19984 (arc_cpu_types): Consider the extras. 19985 * config/arc/arc-cpus.def: Add extras info. 19986 * config/arc/arc-opts.h (processor_type): Consider extra field. 19987 * config/arc/arc.c (arc_override_options): Handle extra field. 19988 199892018-06-12 Claudiu Zissulescu <claziss@synopsys.com> 19990 19991 * config/arc/arc-arch.h: Update ARC_OPTX macro. 19992 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc 19993 field. 19994 * config/arc/arc.c (arc_init): Update pic warning. 19995 (irq_range): Update irq range parsing warnings. 19996 (arc_override_options): Update various warning messages. 19997 (arc_handle_aux_attribute): Likewise. 19998 199992018-06-12 Robert Suchanek <robert.suchanek@mips.com> 20000 20001 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint. 20002 200032018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com> 20004 20005 * doc/sourcebuild.texi: Document usage of line number 0 in verify 20006 compiler messages directives. 20007 200082018-06-12 Matthew Fortune <mfortune@gmail.com> 20009 20010 * config/mips/mips-cpus.def: New MIPS_CPU for i6500. 20011 * config/mips/mips-tables.opt: Regenerate. 20012 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as 20013 mips64r6. 20014 * doc/invoke.texi: Document -march=i6500. 20015 200162018-06-12 Prachi Godbole <prachi.godbole@imgtec.com> 20017 20018 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit. 20019 (i6400_gpmul): Add cpu_unit. 20020 (i6400_gpdiv): Likewise. 20021 (i6400_msa_add_d): Update reservations. 20022 (i6400_msa_int_add) Likewise. 20023 (i6400_msa_short_logic3) Likewise. 20024 (i6400_msa_short_logic2) Likewise. 20025 (i6400_msa_short_logic) Likewise. 20026 (i6400_msa_move) Likewise. 20027 (i6400_msa_cmp) Likewise. 20028 (i6400_msa_short_float2) Likewise. 20029 (i6400_msa_div_d) Likewise. 20030 (i6400_msa_long_logic1) Likewise. 20031 (i6400_msa_long_logic2) Likewise. 20032 (i6400_msa_mult) Likewise. 20033 (i6400_msa_long_float2) Likewise. 20034 (i6400_msa_long_float4) Likewise. 20035 (i6400_msa_long_float5) Likewise. 20036 (i6400_msa_long_float8) Likewise. 20037 (i6400_fpu_fadd): Include frint type. 20038 (i6400_fpu_store): New define_insn_reservation. 20039 (i6400_fpu_load): Likewise. 20040 (i6400_fpu_move): Likewise. 20041 (i6400_fpu_fcmp): Likewise. 20042 (i6400_fpu_fmadd): Likewise. 20043 (i6400_int_mult): Include imul3nc type and update reservation. 20044 (i6400_int_div): Include idiv3 type and update reservation. 20045 (i6400_int_load): Update to check type not move_type. 20046 (i6400_int_store): Likewise. 20047 (i6400_int_prefetch): Set zero latency. 20048 200492018-06-12 Eric Botcazou <ebotcazou@adacore.com> 20050 20051 * gcc.c: Document new %@{...} sequence. 20052 (LINK_COMMAND_SPEC): Use it for the -L switches. 20053 (cpp_unique_options): Use it for the -I switches. 20054 (at_file_argbuf): New global variable. 20055 (in_at_file): Likewise. 20056 (alloc_args): Create at_file_argbuf. 20057 (clear_args): Truncate at_file_argbuf. 20058 (store_arg): If in_at_file, push the argument onto at_file_argbuf. 20059 (open_at_file): New function. 20060 (close_at_file): Likewise. 20061 (create_at_file): Delete. 20062 (do_spec_1) <'i'>: Use open_at_file/close_at_file. 20063 <'o'>: Likewise. 20064 <'@'>: New case. 20065 (validate_switches_from_spec): Deal with %@{...} sequence. 20066 (validate_switches): Likewise. 20067 (driver::finalize): Call clear_args. 20068 200692018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk> 20070 20071 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0. 20072 200732018-06-11 Martin Sebor <msebor@redhat.com> 20074 20075 * doc/invoke.texi (-Wall): List -Wc++17-compat. 20076 (Wno-class-memaccess): Add @opindex. 20077 (Wno-templates, Wno-multiple-inheritance): Same. 20078 (Wno-virtual-inheritance, Wno-namespaces): Same. 20079 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same. 20080 (Wno-format-overflow, Wno-format-truncation): Same. 20081 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same. 20082 (Wno-alloc-size-larger-than, Wframe-larger-than): Same 20083 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same. 20084 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same. 20085 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same. 20086 (Wno-misspelled-isr): Same. 20087 200882018-06-11 Martin Sebor <msebor@redhat.com> 20089 20090 * PR tree-optimization/86083 20091 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p. 20092 200932018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com> 20094 20095 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment. 20096 200972018-06-11 Segher Boessenkool <segher@kernel.crashing.org> 20098 20099 PR target/85755 20100 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers 20101 on the correct operand. 20102 (*movdi_internal64): Ditto. 20103 201042018-06-11 Martin Liska <mliska@suse.cz> 20105 20106 PR tree-optimization/86089 20107 * tree-ssa-strlen.c (get_string_length): Move back removed hunk. 20108 201092018-06-11 Julia Koval <julia.koval@intel.com> 20110 20111 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128) 20112 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove. 20113 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove. 20114 201152018-06-11 Olivier Hainque <hainque@adacore.com> 20116 20117 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83 20118 for Ada with strict dwarf2. 20119 201202018-06-08 Peter Bergner <bergner@vnet.ibm.com> 20121 20122 PR target/85755 20123 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC 20124 addresses. 20125 201262018-06-08 Jan Hubicka <hubicka@ucw.cz> 20127 20128 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4. 20129 201302018-06-08 David Edelsohn <dje.gcc@gmail.com> 20131 20132 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if 20133 TARGET_ELF. 20134 201352018-06-08 Martin Liska <mliska@suse.cz> 20136 20137 * tree-cfg.h (debug_function): Fix argument type to match 20138 implementation. 20139 201402018-06-08 Martin Liska <mliska@suse.cz> 20141 20142 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility): 20143 Remove usage of MPX-related (and removed) fields. 20144 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise. 20145 201462018-06-08 David Malcolm <dmalcolm@redhat.com> 20147 20148 * cfg.c (debug): Use TDF_NONE rather than 0. 20149 * cfghooks.c (debug): Likewise. 20150 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP. 20151 (struct dump_option_value_info): Convert to... 20152 (struct kv_pair): ...this template type. 20153 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE 20154 rather than 0. 20155 (optinfo_verbosity_options): Likewise. 20156 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use 20157 OPTGROUP_NONE. 20158 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather 20159 than int for "optgroup_flags" param. 20160 (dump_generic_expr_loc): Use dump_flags_t rather than int for 20161 "dump_kind" param. 20162 (dump_dec): Likewise. 20163 (dump_finish): Use TDF_NONE rather than 0. 20164 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t 20165 rather than int for "optgroup_flags" param. Use TDF_NONE rather 20166 than 0. Update for change to option_ptr. 20167 (opt_info_switch_p_1): Convert "optgroup_flags" param from int * 20168 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than 20169 0. Update for changes to optinfo_verbosity_options and 20170 optgroup_options. 20171 (opt_info_switch_p): Convert optgroup_flags from int to 20172 optgroup_flags_t. 20173 (dump_basic_block): Use dump_flags_t rather than int 20174 for "dump_kind" param. 20175 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS, 20176 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID) 20177 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME, 20178 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB, 20179 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS, 20180 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG, 20181 TDF_NONE): Convert from macros to... 20182 (enum dump_flag): ...this new enum. 20183 (dump_flags_t): Update to use enum. 20184 (operator|, operator&, operator~, operator|=, operator&=): 20185 Implement for dump_flags_t. 20186 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE, 20187 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL): 20188 Convert from macros to... 20189 (enum optgroup_flag): ...this new enum. 20190 (optgroup_flags_t): New typedef. 20191 (operator|, operator|=): Implement for optgroup_flags_t. 20192 (struct dump_file_info): Convert field "alt_flags" to 20193 dump_flags_t. Convert field "optgroup_flags" to 20194 optgroup_flags_t. 20195 (dump_basic_block): Use dump_flags_t rather than int for param. 20196 (dump_generic_expr_loc): Likewise. 20197 (dump_dec): Likewise. 20198 (dump_register): Convert param "optgroup_flags" to 20199 optgroup_flags_t. 20200 (opt_info_enable_passes): Likewise. 20201 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather 20202 than 0. 20203 * gimple-pretty-print.c (debug): Likewise. 20204 * gimple-ssa-store-merging.c (bswap_replace): Likewise. 20205 (merged_store_group::apply_stores): Likewise. 20206 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise. 20207 * gimple.c (verify_gimple_pp): Likewise. 20208 * graphite-poly.c (print_pbb_body): Likewise. 20209 * passes.c (pass_manager::register_one_dump_file): Convert 20210 local "optgroup_flags" to optgroup_flags_t. 20211 * print-tree.c (print_node): Use TDF_NONE rather than 0. 20212 (debug): Likewise. 20213 (debug_body): Likewise. 20214 * tree-pass.h (struct pass_data): Convert field "optgroup_flags" 20215 to optgroup_flags_t. 20216 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather 20217 than 0. 20218 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise. 20219 (convert_mult_to_fma): Likewise. 20220 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise. 20221 * tree-ssa-sccvn.c (vn_eliminate): Likewise. 20222 * tree-vect-data-refs.c (dump_lower_bound): Convert param 20223 "dump_kind" to dump_flags_t. 20224 202252018-06-08 Segher Boessenkool <segher@kernel.crashing.org> 20226 20227 * config/rs6000/rs6000.c (min, max): Delete. 20228 202292018-06-08 Segher Boessenkool <segher@kernel.crashing.org> 20230 20231 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of 20232 -mabi=spe and -mabi=no-spe. 20233 202342018-06-08 Martin Liska <mliska@suse.cz> 20235 20236 * ipa-pure-const.c (propagate_pure_const): Use ::get at places 20237 where we expect an existing summary. 20238 202392018-06-08 Martin Liska <mliska@suse.cz> 20240 20241 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method. 20242 * ipa-inline.h (estimate_edge_growth): Likewise. 20243 202442018-06-08 Martin Liska <mliska@suse.cz> 20245 20246 * cgraph.c (function_version_hasher::hash): Use 20247 cgraph_node::get_uid (). 20248 (function_version_hasher::equal): 20249 * cgraph.h (cgraph_node::get_uid): New method. 20250 * ipa-inline.c (update_caller_keys): Use 20251 cgraph_node::get_uid (). 20252 (update_callee_keys): Likewise. 20253 * ipa-utils.c (searchc): Likewise. 20254 (ipa_reduced_postorder): Likewise. 20255 * lto-cgraph.c (input_node): Likewise. 20256 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise. 20257 * symbol-summary.h (symtab_insertion): Likewise. 20258 (symtab_removal): Likewise. 20259 (symtab_duplication): Likewise. 20260 * tree-pretty-print.c (dump_function_header): Likewise. 20261 * tree-sra.c (convert_callers_for_node): Likewise. 20262 202632018-06-08 Martin Liska <mliska@suse.cz> 20264 20265 * cgraph.c (symbol_table::create_edge): Always assign a new 20266 unique number. 20267 (symbol_table::free_edge): Do not recycle numbers. 20268 * cgraph.h (cgraph_edge::get): New method. 20269 * symbol-summary.h (symtab_removal): Use it. 20270 (symtab_duplication): Likewise. 20271 (call_summary::hashable_uid): Remove. 20272 202732018-06-08 Martin Liska <mliska@suse.cz> 20274 20275 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove. 20276 (initialize_growth_caches): Remove. 20277 (free_growth_caches): Likewise. 20278 (do_estimate_edge_time): Use edge_growth_cache. 20279 (do_estimate_edge_size): Likewise. 20280 (do_estimate_edge_hints): Likewise. 20281 * ipa-inline.c (reset_edge_caches): Likewise. 20282 (recursive_inlining): Likewise. 20283 (inline_small_functions): Likewise. 20284 * ipa-inline.h (initialize_growth_caches): Remove. 20285 (estimate_edge_size): Likewise. 20286 (estimate_edge_time): Likewise. 20287 (estimate_edge_hints): Likewise. 20288 (reset_edge_growth_cache): Likewise. 20289 * symbol-summary.h (call_summary::remove): New method. 20290 202912018-06-08 Martin Liska <mliska@suse.cz> 20292 20293 * ipa-cp.c (class edge_clone_summary): New summary. 20294 (grow_edge_clone_vectors): Remove. 20295 (ipcp_edge_duplication_hook): Remove. 20296 (class edge_clone_summary_t): New call_summary class. 20297 (ipcp_edge_removal_hook): Remove. 20298 (edge_clone_summary_t::duplicate): New function. 20299 (get_next_cgraph_edge_clone): Use edge_clone_summaries. 20300 (create_specialized_node): Likewise. 20301 (ipcp_driver): Initialize edge_clone_summaries and do not 20302 register hooks. 20303 203042018-06-08 Martin Liska <mliska@suse.cz> 20305 20306 * symbol-summary.h (get): New function. 20307 (call_summary::m_initialize_when_cloning): New class member. 20308 203092018-06-08 Martin Liska <mliska@suse.cz> 20310 20311 * cgraph.c (cgraph_node::remove): Do not recycle uid. 20312 * cgraph.h (symbol_table::release_symbol): Do not pass uid. 20313 (symbol_table::allocate_cgraph_symbol): Do not set uid. 20314 * passes.c (uid_hash_t): Record removed_nodes by their uids. 20315 (remove_cgraph_node_from_order): Use the removed_nodes set. 20316 (do_per_function_toporder): Likwise. 20317 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid 20318 instead of summary_uid. 20319 (symtab_removal): Likewise. 20320 (symtab_duplication): Likewise. 20321 203222018-06-08 Martin Liska <mliska@suse.cz> 20323 20324 * ipa-cp.c (ipcp_store_bits_results): Use 20325 ipcp_transformation_sum. 20326 (ipcp_store_vr_results): Likewise. 20327 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed 20328 to ... 20329 (ipcp_transformation_initialize): ... this. 20330 (ipa_set_node_agg_value_chain): 20331 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum. 20332 (write_ipcp_transformation_info): Likewise. 20333 (read_ipcp_transformation_info): Likewise. 20334 (ipcp_update_bits): Likewise. 20335 (ipcp_update_vr): Likewise. 20336 (ipcp_transform_function): Likewise. 20337 * ipa-prop.h: Rename ipcp_transformation_summary to 20338 ipcp_transformation. 20339 (class ipcp_transformation_t): New function summary. 20340 (ipcp_get_transformation_summary): Use ipcp_transformation_sum. 20341 (ipa_get_agg_replacements_for_node): Likewise. 20342 203432018-06-08 Martin Liska <mliska@suse.cz> 20344 20345 * ipa-pure-const.c (struct funct_state_d): Do it class instead 20346 of struct. 20347 (class funct_state_summary_t): New function_summary class. 20348 (has_function_state): Remove. 20349 (get_function_state): Likewise. 20350 (set_function_state): Likewise. 20351 (add_new_function): Likewise. 20352 (funct_state_summary_t::insert): New function. 20353 (duplicate_node_data): Remove. 20354 (remove_node_data): Remove. 20355 (funct_state_summary_t::duplicate): New function. 20356 (register_hooks): Create new funct_state_summaries. 20357 (pure_const_generate_summary): Use it. 20358 (pure_const_write_summary): Likewise. 20359 (pure_const_read_summary): Likewise. 20360 (propagate_pure_const): Likewise. 20361 (propagate_nothrow): Likewise. 20362 (dump_malloc_lattice): Likewise. 20363 (propagate_malloc): Likewise. 20364 (execute): Do not register hooks, just remove summary 20365 instead. 20366 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify 20367 constructor. 20368 203692018-06-08 Martin Liska <mliska@suse.cz> 20370 20371 * ipa-reference.c (remove_node_data): Remove. 20372 (duplicate_node_data): Likewise. 20373 (class ipa_ref_var_info_summary_t): New class. 20374 (class ipa_ref_opt_summary_t): Likewise. 20375 (get_reference_vars_info): Use ipa_ref_var_info_summaries. 20376 (get_reference_optimization_summary): Use 20377 ipa_ref_opt_sum_summaries. 20378 (set_reference_vars_info): Remove. 20379 (set_reference_optimization_summary): Likewise. 20380 (ipa_init): Create summaries. 20381 (init_function_info): Use function summary. 20382 (ipa_ref_opt_summary_t::duplicate): New function. 20383 (ipa_ref_opt_summary_t::remove): New function. 20384 (get_read_write_all_from_node): Fix GNU coding style. 20385 (propagate): Use function summary. 20386 (write_node_summary_p): Fix GNU coding style. 20387 (stream_out_bitmap): Likewise. 20388 (ipa_reference_read_optimization_summary): Use function summary. 20389 (ipa_reference_c_finalize): Do not release hooks. 20390 203912018-06-08 Martin Liska <mliska@suse.cz> 20392 20393 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method. 20394 (analyze_function_body): Extract multiple calls of get_create. 20395 * ipa-inline-analysis.c (simple_edge_hints): Likewise. 20396 * ipa-inline.c (recursive_inlining): Use ::get method. 20397 * ipa-inline.h (estimate_edge_growth): Likewise. 20398 203992018-06-08 Martin Liska <mliska@suse.cz> 20400 20401 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to 20402 HSA_INVALID. 20403 (hsa_function_summary::hsa_function_summary): Use the new enum 20404 value. 20405 (hsa_gpu_implementation_p): Use hsa_summaries::get. 20406 * hsa-gen.c (hsa_get_host_function): Likewise. 20407 (get_brig_function_name): Likewise. 20408 * ipa-hsa.c (process_hsa_functions): Likewise. 20409 (ipa_hsa_write_summary): Likewise. 20410 * symbol-summary.h (symtab_duplication): Use ::get function/ 20411 (get): New function. 20412 204132018-06-08 Martin Liska <mliska@suse.cz> 20414 20415 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead 20416 of get. 20417 * hsa-common.c (hsa_summary_t::link_functions): Likewise. 20418 (hsa_register_kernel): Likewise. 20419 * hsa-common.h (hsa_gpu_implementation_p): Likewise. 20420 * hsa-gen.c (hsa_get_host_function): Likewise. 20421 (get_brig_function_name): Likewise. 20422 (generate_hsa): Likewise. 20423 (pass_gen_hsail::execute): Likewise. 20424 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise. 20425 (devirtualization_time_bonus): Likewise. 20426 (ipcp_propagate_stage): Likewise. 20427 * ipa-fnsummary.c (redirect_to_unreachable): Likewise. 20428 (edge_set_predicate): Likewise. 20429 (evaluate_conditions_for_known_args): Likewise. 20430 (evaluate_properties_for_edge): Likewise. 20431 (ipa_fn_summary::reset): Likewise. 20432 (ipa_fn_summary_t::duplicate): Likewise. 20433 (dump_ipa_call_summary): Likewise. 20434 (ipa_dump_fn_summary): Likewise. 20435 (analyze_function_body): Likewise. 20436 (compute_fn_summary): Likewise. 20437 (estimate_edge_devirt_benefit): Likewise. 20438 (estimate_edge_size_and_time): Likewise. 20439 (estimate_calls_size_and_time): Likewise. 20440 (estimate_node_size_and_time): Likewise. 20441 (inline_update_callee_summaries): Likewise. 20442 (remap_edge_change_prob): Likewise. 20443 (remap_edge_summaries): Likewise. 20444 (ipa_merge_fn_summary_after_inlining): Likewise. 20445 (ipa_update_overall_fn_summary): Likewise. 20446 (read_ipa_call_summary): Likewise. 20447 (inline_read_section): Likewise. 20448 (write_ipa_call_summary): Likewise. 20449 (ipa_fn_summary_write): Likewise. 20450 (ipa_free_fn_summary): Likewise. 20451 * ipa-hsa.c (process_hsa_functions): Likewise. 20452 (ipa_hsa_write_summary): Likewise. 20453 (ipa_hsa_read_section): Likewise. 20454 * ipa-icf.c (sem_function::merge): Likewise. 20455 * ipa-inline-analysis.c (simple_edge_hints): Likewise. 20456 (do_estimate_edge_time): Likewise. 20457 (estimate_size_after_inlining): Likewise. 20458 (estimate_growth): Likewise. 20459 (growth_likely_positive): Likewise. 20460 * ipa-inline-transform.c (clone_inlined_nodes): Likewise. 20461 (inline_call): Likewise. 20462 * ipa-inline.c (caller_growth_limits): Likewise. 20463 (can_inline_edge_p): Likewise. 20464 (can_inline_edge_by_limits_p): Likewise. 20465 (compute_uninlined_call_time): Likewise. 20466 (compute_inlined_call_time): Likewise. 20467 (want_inline_small_function_p): Likewise. 20468 (edge_badness): Likewise. 20469 (update_caller_keys): Likewise. 20470 (update_callee_keys): Likewise. 20471 (recursive_inlining): Likewise. 20472 (inline_small_functions): Likewise. 20473 (inline_to_all_callers_1): Likewise. 20474 (dump_overall_stats): Likewise. 20475 (early_inline_small_functions): Likewise. 20476 (early_inliner): Likewise. 20477 * ipa-inline.h (estimate_edge_growth): Likewise. 20478 * ipa-profile.c (ipa_propagate_frequency_1): Likewise. 20479 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise. 20480 * ipa-prop.h (IPA_NODE_REF): Likewise. 20481 (IPA_EDGE_REF): Likewise. 20482 * ipa-pure-const.c (malloc_candidate_p): Likewise. 20483 (propagate_malloc): Likewise. 20484 * ipa-split.c (execute_split_functions): Likewise. 20485 * symbol-summary.h: Rename get to get_create. 20486 (get): Likewise. 20487 (get_create): Likewise. 20488 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise. 20489 204902018-06-08 Martin Liska <mliska@suse.cz> 20491 20492 * symbol-summary.h (release): Move definition out of class 20493 declaration. 20494 (symtab_removal): Likewise. 20495 (symtab_duplication): Likewise. 20496 204972018-06-08 Martin Liska <mliska@suse.cz> 20498 20499 * symbol-summary.h (function_summary): Move constructor 20500 implementation out of class declaration. 20501 (release): Likewise. 20502 (symtab_insertion): Likewise. 20503 (symtab_removal): Likewise. 20504 (symtab_duplication): Likewise. 20505 (get): Likewise. 20506 205072018-06-08 Martin Liska <mliska@suse.cz> 20508 20509 * Makefile.in: Remove support for MPX (macros, related functions, 20510 fields in cgraph_node, ...). 20511 * builtin-types.def (BT_BND): Likewise. 20512 (BT_FN_BND_CONST_PTR): Likewise. 20513 (BT_FN_CONST_PTR_BND): Likewise. 20514 (BT_FN_VOID_PTR_BND): Likewise. 20515 (BT_FN_BND_CONST_PTR_SIZE): Likewise. 20516 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise. 20517 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise. 20518 (expand_builtin_mempcpy_with_bounds): Likewise. 20519 (expand_builtin_memset_with_bounds): Likewise. 20520 (expand_builtin_memset_args): Likewise. 20521 (std_expand_builtin_va_start): Likewise. 20522 (expand_builtin): Likewise. 20523 (expand_builtin_with_bounds): Likewise. 20524 * builtins.def (DEF_BUILTIN_CHKP): Likewise. 20525 (DEF_LIB_BUILTIN_CHKP): Likewise. 20526 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise. 20527 (DEF_CHKP_BUILTIN): Likewise. 20528 (BUILT_IN_MEMCPY): Likewise. 20529 (BUILT_IN_MEMMOVE): Likewise. 20530 (BUILT_IN_MEMPCPY): Likewise. 20531 (BUILT_IN_MEMSET): Likewise. 20532 (BUILT_IN_STPCPY): Likewise. 20533 (BUILT_IN_STRCAT): Likewise. 20534 (BUILT_IN_STRCHR): Likewise. 20535 (BUILT_IN_STRCPY): Likewise. 20536 (BUILT_IN_STRLEN): Likewise. 20537 (BUILT_IN_MEMCPY_CHK): Likewise. 20538 (BUILT_IN_MEMMOVE_CHK): Likewise. 20539 (BUILT_IN_MEMPCPY_CHK): Likewise. 20540 (BUILT_IN_MEMSET_CHK): Likewise. 20541 (BUILT_IN_STPCPY_CHK): Likewise. 20542 (BUILT_IN_STRCAT_CHK): Likewise. 20543 (BUILT_IN_STRCPY_CHK): Likewise. 20544 * calls.c (store_bounds): Likewise. 20545 (emit_call_1): Likewise. 20546 (special_function_p): Likewise. 20547 (maybe_warn_nonstring_arg): Likewise. 20548 (initialize_argument_information): Likewise. 20549 (finalize_must_preallocate): Likewise. 20550 (compute_argument_addresses): Likewise. 20551 (expand_call): Likewise. 20552 * cfgexpand.c (expand_call_stmt): Likewise. 20553 (expand_return): Likewise. 20554 (expand_gimple_stmt_1): Likewise. 20555 (pass_expand::execute): Likewise. 20556 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise. 20557 (cgraph_node::remove): Likewise. 20558 (cgraph_node::dump): Likewise. 20559 (cgraph_node::verify_node): Likewise. 20560 * cgraph.h (chkp_function_instrumented_p): Likewise. 20561 (symtab_node::get_alias_target): Likewise. 20562 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise. 20563 (cgraph_local_p): Likewise. 20564 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise. 20565 (cgraph_edge::rebuild_references): Likewise. 20566 * cgraphunit.c (varpool_node::finalize_decl): Likewise. 20567 (walk_polymorphic_call_targets): Likewise. 20568 (cgraph_node::expand_thunk): Likewise. 20569 (symbol_table::output_weakrefs): Likewise. 20570 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise. 20571 (ix86_handle_option): Likewise. 20572 * config/i386/constraints.md: Likewise. 20573 * config/i386/i386-builtin-types.def (BND): Likewise. 20574 (VOID): Likewise. 20575 (PVOID): Likewise. 20576 (ULONG): Likewise. 20577 * config/i386/i386-builtin.def (BDESC_END): Likewise. 20578 (BDESC_FIRST): Likewise. 20579 (BDESC): Likewise. 20580 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise. 20581 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise. 20582 * config/i386/i386.c (enum reg_class): Likewise. 20583 (ix86_target_string): Likewise. 20584 (ix86_option_override_internal): Likewise. 20585 (ix86_conditional_register_usage): Likewise. 20586 (ix86_valid_target_attribute_inner_p): Likewise. 20587 (ix86_set_indirect_branch_type): Likewise. 20588 (ix86_set_current_function): Likewise. 20589 (ix86_function_arg_regno_p): Likewise. 20590 (init_cumulative_args): Likewise. 20591 (ix86_function_arg_advance): Likewise. 20592 (ix86_function_arg): Likewise. 20593 (ix86_pass_by_reference): Likewise. 20594 (ix86_function_value_regno_p): Likewise. 20595 (ix86_function_value_1): Likewise. 20596 (ix86_function_value_bounds): Likewise. 20597 (ix86_return_in_memory): Likewise. 20598 (ix86_setup_incoming_vararg_bounds): Likewise. 20599 (ix86_va_start): Likewise. 20600 (indirect_thunk_need_prefix): Likewise. 20601 (print_reg): Likewise. 20602 (ix86_print_operand): Likewise. 20603 (ix86_expand_call): Likewise. 20604 (ix86_output_function_return): Likewise. 20605 (reg_encoded_number): Likewise. 20606 (BDESC_VERIFYS): Likewise. 20607 (ix86_init_mpx_builtins): Likewise. 20608 (ix86_init_builtins): Likewise. 20609 (ix86_emit_cmove): Likewise. 20610 (ix86_emit_move_max): Likewise. 20611 (ix86_expand_builtin): Likewise. 20612 (ix86_builtin_mpx_function): Likewise. 20613 (ix86_get_arg_address_for_bt): Likewise. 20614 (ix86_load_bounds): Likewise. 20615 (ix86_store_bounds): Likewise. 20616 (ix86_load_returned_bounds): Likewise. 20617 (ix86_store_returned_bounds): Likewise. 20618 (ix86_class_likely_spilled_p): Likewise. 20619 (ix86_hard_regno_mode_ok): Likewise. 20620 (x86_order_regs_for_local_alloc): Likewise. 20621 (ix86_mitigate_rop): Likewise. 20622 (ix86_bnd_prefixed_insn_p): Likewise. 20623 (ix86_mpx_bound_mode): Likewise. 20624 (ix86_make_bounds_constant): Likewise. 20625 (ix86_initialize_bounds): Likewise. 20626 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise. 20627 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise. 20628 (TARGET_LOAD_RETURNED_BOUNDS): Likewise. 20629 (TARGET_STORE_RETURNED_BOUNDS): Likewise. 20630 (TARGET_CHKP_BOUND_MODE): Likewise. 20631 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise. 20632 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise. 20633 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise. 20634 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise. 20635 * config/i386/i386.h (TARGET_MPX): Likewise. 20636 (TARGET_MPX_P): Likewise. 20637 (VALID_BND_REG_MODE): Likewise. 20638 (FIRST_BND_REG): Likewise. 20639 (LAST_BND_REG): Likewise. 20640 (enum reg_class): Likewise. 20641 (BND_REG_P): Likewise. 20642 (BND_REGNO_P): Likewise. 20643 (BNDmode): Likewise. 20644 (ADJUST_INSN_LENGTH): Likewise. 20645 * config/i386/i386.md: Likewise. 20646 * config/i386/i386.opt: Likewise. 20647 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise. 20648 (defined): Likewise. 20649 (LINK_MPX): Likewise. 20650 (MPX_SPEC): Likewise. 20651 (LIBMPX_SPEC): Likewise. 20652 (LIBMPXWRAPPERS_SPEC): Likewise. 20653 (CHKP_SPEC): Likewise. 20654 * config/i386/predicates.md: Likewise. 20655 * dbxout.c (dbxout_type): Likewise. 20656 * doc/extend.texi: Likewise. 20657 * doc/invoke.texi: Likewise. 20658 * doc/md.texi: Likewise. 20659 * doc/tm.texi: Likewise. 20660 * doc/tm.texi.in: Likewise. 20661 * dwarf2out.c (is_base_type): Likewise. 20662 (gen_formal_types_die): Likewise. 20663 (gen_subprogram_die): Likewise. 20664 (gen_type_die_with_usage): Likewise. 20665 (gen_decl_die): Likewise. 20666 (dwarf2out_late_global_decl): Likewise. 20667 * expr.c (expand_assignment): Likewise. 20668 (emit_storent_insn): Likewise. 20669 (store_expr_with_bounds): Likewise. 20670 (store_expr): Likewise. 20671 (expand_expr_real_1): Likewise. 20672 * expr.h (store_expr_with_bounds): Likewise. 20673 * function.c (use_register_for_decl): Likewise. 20674 (struct bounds_parm_data): Likewise. 20675 (assign_parms_augmented_arg_list): Likewise. 20676 (assign_parm_find_entry_rtl): Likewise. 20677 (assign_parm_is_stack_parm): Likewise. 20678 (assign_parm_load_bounds): Likewise. 20679 (assign_bounds): Likewise. 20680 (assign_parms): Likewise. 20681 (expand_function_start): Likewise. 20682 * gcc.c (CHKP_SPEC): Likewise. 20683 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise. 20684 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise. 20685 (wrestrict_dom_walker::check_call): Likewise. 20686 * gimple.c (gimple_build_call_from_tree): Likewise. 20687 * gimple.h (enum gf_mask): Likewise. 20688 (gimple_call_with_bounds_p): Likewise. 20689 (gimple_call_set_with_bounds): Likewise. 20690 * gimplify.c (gimplify_init_constructor): Likewise. 20691 * ipa-cp.c (initialize_node_lattices): Likewise. 20692 (propagate_constants_across_call): Likewise. 20693 (find_more_scalar_values_for_callers_subset): Likewise. 20694 * ipa-hsa.c (process_hsa_functions): Likewise. 20695 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise. 20696 * ipa-icf.c (sem_function::merge): Likewise. 20697 * ipa-inline.c (early_inliner): Likewise. 20698 * ipa-pure-const.c (warn_function_noreturn): Likewise. 20699 (warn_function_cold): Likewise. 20700 (propagate_pure_const): Likewise. 20701 * ipa-ref.h (enum GTY): Likewise. 20702 * ipa-split.c (find_retbnd): Likewise. 20703 (consider_split): Likewise. 20704 (split_function): Likewise. 20705 * ipa-visibility.c (cgraph_externally_visible_p): Likewise. 20706 * ipa.c (walk_polymorphic_call_targets): Likewise. 20707 (symbol_table::remove_unreachable_nodes): Likewise. 20708 (process_references): Likewise. 20709 (cgraph_build_static_cdtor_1): Likewise. 20710 * lto-cgraph.c (lto_output_node): Likewise. 20711 (output_refs): Likewise. 20712 (compute_ltrans_boundary): Likewise. 20713 (input_overwrite_node): Likewise. 20714 (input_node): Likewise. 20715 (input_cgraph_1): Likewise. 20716 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise. 20717 * passes.c (pass_manager::execute_early_local_passes): Likewise. 20718 (class pass_chkp_instrumentation_passes): Likewise. 20719 (make_pass_chkp_instrumentation_passes): Likewise. 20720 * passes.def: Likewise. 20721 * rtl.h (struct GTY): Likewise. 20722 (CALL_EXPR_WITH_BOUNDS_P): Likewise. 20723 * stor-layout.c (layout_type): Likewise. 20724 * symtab.c: Likewise. 20725 * target.def: Likewise. 20726 * targhooks.c (default_chkp_bound_type): Likewise. 20727 (default_chkp_bound_mode): Likewise. 20728 (default_builtin_chkp_function): Likewise. 20729 (default_chkp_function_value_bounds): Likewise. 20730 (default_chkp_make_bounds_constant): Likewise. 20731 (default_chkp_initialize_bounds): Likewise. 20732 * targhooks.h (default_chkp_bound_type): Likewise. 20733 (default_chkp_bound_mode): Likewise. 20734 (default_builtin_chkp_function): Likewise. 20735 (default_chkp_function_value_bounds): Likewise. 20736 (default_chkp_make_bounds_constant): Likewise. 20737 (default_chkp_initialize_bounds): Likewise. 20738 * toplev.c (compile_file): Likewise. 20739 (process_options): Likewise. 20740 * tree-core.h (DEF_BUILTIN): Likewise. 20741 (DEF_BUILTIN_CHKP): Likewise. 20742 * tree-inline.c (declare_return_variable): Likewise. 20743 (remap_gimple_stmt): Likewise. 20744 (copy_bb): Likewise. 20745 (initialize_inlined_parameters): Likewise. 20746 (expand_call_inline): Likewise. 20747 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise. 20748 (make_pass_ipa_chkp_early_produce_thunks): Likewise. 20749 (make_pass_ipa_chkp_produce_thunks): Likewise. 20750 (make_pass_chkp): Likewise. 20751 (make_pass_chkp_opt): Likewise. 20752 (make_pass_chkp_instrumentation_passes): Likewise. 20753 * tree-pretty-print.c (dump_generic_node): Likewise. 20754 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise. 20755 * tree-ssa-dce.c (propagate_necessity): Likewise. 20756 (eliminate_unnecessary_stmts): Likewise. 20757 * tree-ssa-pre.c (create_expression_by_pieces): Likewise. 20758 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise. 20759 * tree-ssa-sccvn.h: Likewise. 20760 * tree-ssa-strlen.c (get_string_length): Likewise. 20761 (valid_builtin_call): Likewise. 20762 (adjust_last_stmt): Likewise. 20763 (handle_builtin_strchr): Likewise. 20764 (handle_builtin_strcpy): Likewise. 20765 (handle_builtin_stxncpy): Likewise. 20766 (handle_builtin_memcpy): Likewise. 20767 (handle_builtin_strcat): Likewise. 20768 (strlen_check_and_optimize_stmt): Likewise. 20769 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise. 20770 * tree-streamer-in.c: Likewise. 20771 * tree-streamer.c (record_common_node): Likewise. 20772 * tree.c (tree_code_size): Likewise. 20773 (wide_int_to_tree_1): Likewise. 20774 (type_contains_placeholder_1): Likewise. 20775 (build_common_tree_nodes): Likewise. 20776 * tree.def (POINTER_BOUNDS_TYPE): Likewise. 20777 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise. 20778 (POINTER_BOUNDS_P): Likewise. 20779 (BOUNDED_TYPE_P): Likewise. 20780 (BOUNDED_P): Likewise. 20781 (CALL_WITH_BOUNDS_P): Likewise. 20782 (pointer_bounds_type_node): Likewise. 20783 * value-prof.c (gimple_ic): Likewise. 20784 * var-tracking.c (vt_add_function_parameters): Likewise. 20785 * varasm.c (make_decl_rtl): Likewise. 20786 (assemble_start_function): Likewise. 20787 (output_constant): Likewise. 20788 (maybe_assemble_visibility): Likewise. 20789 * varpool.c (ctor_for_folding): Likewise. 20790 * chkp-builtins.def: Remove. 20791 * ipa-chkp.c: Remove. 20792 * ipa-chkp.h: Remove. 20793 * rtl-chkp.c: Remove. 20794 * rtl-chkp.h: Remove. 20795 * tree-chkp-opt.c: Remove. 20796 * tree-chkp.c: Remove. 20797 * tree-chkp.h: Remove. 20798 207992018-06-07 Carl Love <cel@us.ibm.com> 20800 20801 * config/rs6000/vsx.md (vextract_fp_from_shorth, 20802 vextract_fp_from_shortl): Add BE support. 20803 208042018-06-07 Paul Koning <ni1d@arrl.net> 20805 20806 * compare-elim.c (try_merge_compare): Don't merge compare if 20807 address contains a side effect. 20808 (try_eliminate_compare): Likewise. 20809 208102018-06-07 Olga Makhotina <olga.makhotina@intel.com> 20811 20812 * config.gcc: Support "tremont". 20813 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont". 20814 * config/i386/i386-c.c (ix86_target_macros_internal): Handle 20815 PROCESSOR_TREMONT. 20816 * config/i386/i386.c (m_TREMONT): Define. 20817 (processor_target_table): Add "tremont". 20818 (PTA_TREMONT): Define. 20819 (ix86_lea_outperforms): Add TARGET_TREMONT. 20820 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT. 20821 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT 20822 and M_INTEL_GOLDMONT_PLUS. 20823 (fold_builtin_cpu): Add "tremont". 20824 (ix86_add_stmt_cost): Add TARGET_TREMONT. 20825 (ix86_option_override_internal): Add "tremont". 20826 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT. 20827 (processor_type): Add PROCESSOR_TREMONT. 20828 * config/i386/x86-tune.def: Add m_TREMONT. 20829 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type. 20830 208312018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com> 20832 20833 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the 20834 symbol defined for msp430i* devices to be lower case. 20835 208362018-06-07 Richard Biener <rguenther@suse.de> 20837 20838 * graphite-sese-to-poly.c (extract_affine): Avoid unneded 20839 wrapping. Properly wrap the result of a BIT_NOT_EXPR. 20840 Properly wrap signed arithmetic if overflow wraps. 20841 208422018-06-07 Jakub Jelinek <jakub@redhat.com> 20843 20844 PR tree-optimization/69615 20845 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs 20846 of a cast from a same precision integral SSA_NAME in a bb dominated 20847 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit 20848 cast to utype if rhs2 has already a compatible type. 20849 208502018-06-07 Richard Biener <rguenther@suse.de> 20851 20852 PR tree-optimization/85935 20853 * graphite-scop-detection.c (find_params_in_bb): Analyze 20854 condition operands with respect to the correct loop. Assert 20855 the analysis doesn't fail. 20856 208572018-06-04 Carl Love <cel@us.ibm.com> 20858 20859 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index 20860 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3 20861 as it is slightly cheaper. 20862 (first_match_or_eos_index_<mode>): 20863 Calculate index using natural element order. 20864 (first_match_index_<mode>): 20865 Calculate index using natural element order. 20866 (first_match_or_eos_index_<mode>): 20867 Calculate index using natural order. 20868 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>. 20869 for BE and LE modes. 20870 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB, 20871 P9V_BUILTIN_VCLZLSBB_V16QI. 20872 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode 20873 specific. 20874 208752018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org> 20876 20877 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust 20878 indentation and line wrap for many prototypes. Add missing 20879 @smallexample directives around block of prototypes for vec_xl and 20880 vec_xst. 20881 208822018-06-05 Michael Meissner <meissner@linux.ibm.com> 20883 20884 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to 20885 track if we pass or return IEEE 128-bit floating point. 20886 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate 20887 C++ mangling that is compatible with GCC 8.1. 20888 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook. 20889 (init_cumulative_args): Note if we pass or return IEEE 128-bit 20890 floating point types. 20891 (rs6000_function_arg_advance_1): Likewise. 20892 (rs6000_mangle_type): Optionally generate mangled names that match 20893 what GCC 8.1 generated for IEEE 128-bit floating point types. 20894 (rs6000_globalize_decl_name): If we have an external function that 20895 passes or returns IEEE 128-bit types, generate a weak reference 20896 from the mangled name used in GCC 8.1 to the current mangled 20897 name. 20898 (rs6000_init_builtins): Make __ibm128 use the long double type if 20899 long double is IBM extended double. Make __float128 use the long 20900 double type if long double is IEEE 128-bit. 20901 20902 PR target/85657 20903 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper 20904 macro for __ibm128 built-in functions. 20905 (PACK_IF): Add __ibm128 pack/unpack functions. 20906 (UNPACK_IF): Likewise. 20907 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not 20908 enable long double built-in functions if long double is IEEE 20909 128-bit floating point. 20910 (rs6000_invalid_builtin): Update long double built-in function 20911 error message. 20912 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in 20913 functions, adjust the built-in function to use the long double 20914 built-in function if __ibm128 and long double are the same type. 20915 * doc/extend.texi (PowerPC builtins): Update documention for 20916 __builtin_{,un}pack_longdouble. Add documentation for 20917 __builtin_{,un}pack_ibm128. 20918 209192018-06-06 Jim Wilson <jimw@sifive.com> 20920 20921 * config/riscv/riscv.c (enum riscv_privilege_levels): New. 20922 (struct machine_function): New field interrupt_mode. 20923 (riscv_handle_type_attribute): New function. Add forward declaration. 20924 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute. 20925 (riscv_expand_epilogue): Check interrupt_mode field. 20926 (riscv_set_current_function): Check interrupt attribute args and 20927 set interrupt_mode field. 20928 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New. 20929 (riscv_sret, riscv_uret): New. 20930 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document 20931 new arguments to interrupt attribute. 20932 209332018-06-06 Peter Bergner <bergner@vnet.ibm.com> 20934 20935 PR target/63177 20936 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9. 20937 Don't handle -mcpu=power8 if -mpower9-vector is also used. 20938 209392018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org> 20940 20941 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD, 20942 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses. 20943 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove 20944 several redundant entries. 20945 209462018-06-06 David Malcolm <dmalcolm@redhat.com> 20947 20948 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return 20949 type from "rtx" to "rtx_insn *". 20950 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise 20951 for local "call_insn", removing cast. 20952 (ix86_expand_call): Likewise, introducing a "call_insn" local. 20953 209542018-06-06 Eric Botcazou <ebotcazou@adacore.com> 20955 20956 PR tree-optimization/86066 20957 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR 20958 for BIT_INSERT_EXPR stores. 20959 209602018-06-06 Richard Biener <rguenther@suse.de> 20961 20962 PR tree-optimization/86062 20963 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary 20964 component refs ontop 20965 of to be offsetted base. 20966 209672018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com> 20968 20969 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers 20970 to be static and remove check on interrupt attribute name. 20971 209722018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org> 20973 20974 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove 20975 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes. 20976 209772018-06-05 Steve Ellcey <sellcey@cavium.com> 20978 20979 PR target/79924 20980 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove 20981 second argument. 20982 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd): 20983 Remove second argument, change how error is called. 20984 (aarch64_layout_arg): Remove second argument from 20985 aarch64_err_no_fpadvsimd call. 20986 (aarch64_init_cumulative_args): Ditto. 20987 (aarch64_gimplify_va_arg_expr): Ditto. 20988 * config/aarch64/aarch64.md (mov<mode>): Ditto. 20989 209902018-06-05 Uros Bizjak <ubizjak@gmail.com> 20991 20992 * config/i386/i386.md (simple_return_indirect_internal): New expander. 20993 (*simple_return_indirect_internal<mode>): Rename from 20994 simple_return_indirect_internal. Use W mode iterator. 20995 (rstorssp): New expander. 20996 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator. 20997 (clrssbsy): New expander. 20998 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator. 20999 210002018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com> 21001 21002 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to 21003 __typeof__. 21004 (cmse_check_pointed_object): Likewise. 21005 210062018-06-05 Martin Liska <mliska@suse.cz> 21007 21008 PR gcov-profile/47618 21009 * doc/invoke.texi: Document how -fprofile-dir format 21010 is extended. 21011 210122018-06-05 Richard Biener <rguenther@suse.de> 21013 21014 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge 21015 removal pretend DOM info isn't available so we do not update 21016 it and only remove edges, not dominated blocks. Actually free 21017 DOM info in case we removed something. Remove unreachable blocks. 21018 (mfb_keep_latches): Work with either DOM info or marked backedges. 21019 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks 21020 first. Mark backedges if DOM info isn't available. 21021 (Re-)compute DOM info after cleanup_control_flow_pre. 21022 210232018-06-05 Richard Biener <rguenther@suse.de> 21024 21025 * tree-cfg.c (struct locus_discrim_map): Store line, not location. 21026 (locus_discrim_hasher::hash): Adjust. 21027 (locus_discrim_hasher::equal): Likewise. 21028 (next_discriminator_for_locus): Work on line directly. 21029 (same_line_p): Pass in expanded locus1 as well. 21030 (assign_discriminators): Avoid redundant location expansions. 21031 210322018-06-05 Richard Biener <rguenther@suse.de> 21033 21034 PR tree-optimization/86046 21035 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P 21036 if required after clearing TREE_ADDRESSABLE. 21037 210382018-06-05 Richard Biener <rguenther@suse.de> 21039 21040 PR tree-optimization/86047 21041 * tree-ssa-loop.c (for_each_index): Glob handling of all 21042 decls and constants and really handle all of them. 21043 210442018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 21045 21046 PR target/81497 21047 * config/arm/arm-builtins.c (arm_type_qualifiers): Add 21048 qualifier_void_pointer and qualifier_const_void_pointer. 21049 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above. 21050 (arm_init_builtins): Handle the above. 21051 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc, 21052 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2, 21053 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for 21054 void intrinsics. 21055 210562018-06-05 Martin Liska <mliska@suse.cz> 21057 21058 * auto-profile.c (read_autofdo_file): Do not use 21059 gcov_ctr_summary struct. 21060 (afdo_callsite_hot_enough_for_early_inline): Likewise. 21061 * coverage.c (struct counts_entry): Likewise. 21062 (read_counts_file): Read just single summary entry. 21063 (get_coverage_counts): Use gcov_summary struct. 21064 * coverage.h (get_coverage_counts): Likewise. 21065 * gcov-dump.c (dump_working_sets): Likewise. 21066 (tag_summary): Dump just single summary. 21067 * gcov-io.c (gcov_write_summary): Write just histogram 21068 summary. 21069 (gcov_read_summary): Read just single summary. 21070 (compute_working_sets): Use gcov_summary struct. 21071 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage 21072 of GCOV_COUNTERS_SUMMABLE. 21073 (GCOV_COUNTERS_SUMMABLE): Remove. 21074 (GCOV_FIRST_VALUE_COUNTER): Replace with 21075 GCOV_COUNTER_V_INTERVAL. 21076 (struct gcov_ctr_summary): Remove. 21077 (struct gcov_summary): Directly use fields of former 21078 gcov_ctr_summary. 21079 (compute_working_sets): Use gcov_summary struct. 21080 * gcov.c (read_count_file): Do not use ctrs fields. 21081 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary 21082 struct. 21083 * lto-streamer.h (struct GTY): Make profile_info gcov_summary 21084 struct. 21085 * profile.c: Likewise. 21086 * profile.h: Likewise. 21087 210882018-06-05 Martin Liska <mliska@suse.cz> 21089 21090 PR gcov-profile/84846 21091 * gcov.c (output_lines): Print working directory only 21092 in intermediate format. 21093 210942018-06-05 Andreas Krebbel <krebbel@linux.ibm.com> 21095 21096 * config/s390/s390-builtin-types.def: Add void function type. 21097 * config/s390/s390-builtins.def: Use the function type for the 21098 tbeginc builtin. 21099 211002018-06-04 Jim Wilson <jimw@sifive.com> 21101 21102 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg 21103 to int. 21104 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue 21105 and maybe_eh_return. Change regno to unsigned int. Use new args to 21106 handle EH_RETURN_DATA_REGNO registers properly. 21107 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg. 21108 (riscv_expand_epilogue): Update comment. Change argument name and 21109 type. Update code to use new name and type. Pass new args to 21110 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when 21111 EXCEPTION_RETURN. 21112 * config/riscv/riscv.md (NORMAL_RETURN): New. 21113 (SIBCALL_RETURN, EXCEPTION_RETURN): New. 21114 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg. 21115 (eh_return): Call gen_eh_return_internal and emit barrier. 21116 (eh_return_internal): Call riscv_expand_epilogue. 21117 211182018-06-04 Eric Botcazou <ebotcazou@adacore.com> 21119 21120 * gimple-ssa-store-merging.c (struct merged_store_group): Move up 21121 bit_insertion field and declare can_be_merged_into method. 21122 (merged_store_group::can_be_merged_into): New method. 21123 (imm_store_chain_info::coalesce_immediate): Call it to decide whether 21124 consecutive non-overlapping stores can be merged. Turn MEM_REF stores 21125 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store. 21126 211272018-06-04 Richard Biener <rguenther@suse.de> 21128 21129 PR tree-optimization/85955 21130 * builtins.c (fold_builtin_sincos): Convert pointers to 21131 destination to appropriate type before dereferencing. 21132 211332018-06-04 Segher Boessenkool <segher@kernel.crashing.org> 21134 21135 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode. 21136 211372018-06-04 Richard Sandiford <richard.sandiford@linaro.org> 21138 21139 * expr.c (expand_expr_real_1): Force the operand into memory if 21140 its TYPE_MODE is BLKmode and if there is no integer mode for 21141 the number of bits being extracted. 21142 211432018-06-04 Jakub Jelinek <jakub@redhat.com> 21144 21145 PR target/85832 21146 PR target/86036 21147 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): 21148 Use vptestnm rather than vptestm in (=Yc,v,C) variant. 21149 211502018-06-04 Richard Biener <rguenther@suse.de> 21151 21152 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into... 21153 (cleanup_tree_cfg_noloop): ... single caller. Do 21154 start_recording_case_labels later. 21155 211562018-06-04 Sebastian Peryt <sebastian.peryt@intel.com> 21157 21158 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED 21159 to _IMMINTRIN_H_INCLUDED. 21160 * config/i386/pconfigintrin.h: Ditto. 21161 * config/i386/waitpkgintrin.h: Ditto. 21162 * config/i386/immintrin.h: Add includes for sgxintrin.h, 21163 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h. 21164 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h, 21165 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h, 21166 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h, 21167 waitpkgintrin.h and cldemoteintrin.h. 21168 211692018-06-04 Richard Biener <rguenther@suse.de> 21170 21171 PR tree-optimization/86038 21172 * tracer.c (find_best_successor): Check probability for 21173 being initialized, bail out if not. 21174 211752018-06-04 Richard Earnshaw <rearnsha@arm.com> 21176 21177 PR target/86003 21178 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list 21179 of bits to ignore when comparing architectures. 21180 211812018-06-04 Jakub Jelinek <jakub@redhat.com> 21182 21183 PR tree-optimization/69615 21184 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the 21185 maximum or minimum of the type, try to merge it also as if 21186 range1 is + [-, x - 1] or + [x + 1, -]. 21187 21188 PR c++/86025 21189 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE. 21190 211912018-06-03 Eric Botcazou <ebotcazou@adacore.com> 21192 21193 PR tree-optimization/86034 21194 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to 21195 the unsigned bitfield type in a bit insertion sequence if it does not 21196 have a larger precision than the bitfield size. 21197 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR. 21198 211992018-06-03 Kito Cheng <kito.cheng@gmail.com> 21200 21201 * config/nds32/nds32-peephole2.md: Add new patterns for code size. 21202 212032018-06-03 Chung-Ju Wu <jasonwucj@gmail.com> 21204 21205 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J. 21206 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J. 21207 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking. 21208 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J). 21209 212102018-06-03 Chung-Ju Wu <jasonwucj@gmail.com> 21211 21212 * common/config/nds32/nds32-common.c (nds32_option_optimization_table): 21213 Disable -fdelete-null-pointer-checks for ELF toolchain. 21214 212152018-06-02 Chung-Ju Wu <jasonwucj@gmail.com> 21216 Kito Cheng <kito.cheng@gmail.com> 21217 21218 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt. 21219 (nds32le-*-*, nds32be-*-*): Integrate checking process. 21220 (nds32*-*-*): Add glibc and uclibc conditions. 21221 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New. 21222 (TARGET_EXCEPT_UNWIND_INFO): Define. 21223 * config/nds32/elf.h: New file. 21224 * config/nds32/linux.h: New file. 21225 * config/nds32/nds32-elf.opt: New file. 21226 * config/nds32/nds32-linux.opt: New file. 21227 * config/nds32/nds32-fp-as-gp.c 21228 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI. 21229 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider 21230 TARGET_LINUX_ABI. 21231 (nds32_asm_file_end): Ditto. 21232 (nds32_print_operand): Ditto. 21233 (nds32_insert_attributes): Ditto. 21234 (nds32_init_libfuncs): New function. 21235 (TARGET_HAVE_TLS): Define. 21236 (TARGET_INIT_LIBFUNCS): Define. 21237 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax 21238 spec content. 21239 (TARGET_ELF): Apply different mcmodel setting. 21240 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has 21241 been migrated into elf.h and linux.h files. 21242 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI. 21243 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI. 21244 (mcmodel): The content has been migrated into nds32-elf.opt and 21245 nds32-linux.opt files. 21246 * config/nds32/t-elf: New file. 21247 * config/nds32/t-linux: New file. 21248 212492018-06-02 Chung-Ju Wu <jasonwucj@gmail.com> 21250 Shiva Chen <shiva0217@gmail.com> 21251 21252 * config/nds32/constants.md (unspec_volatile_element): Add 21253 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END. 21254 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp 21255 optimization. 21256 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare. 21257 (make_pass_nds32_fp_as_gp): Declare. 21258 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one 21259 optmization pass. 21260 (nds32_asm_function_end_prologue): Remove unused asm output. 21261 (nds32_asm_function_begin_epilogue): Remove unused asm output. 21262 (nds32_asm_file_start): Output necessary fp_as_gp information. 21263 (nds32_option_override): Adjust register usage. 21264 (nds32_expand_prologue): Consider fp_as_gp situation. 21265 (nds32_expand_prologue_v3push): Consider fp_as_gp situation. 21266 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p. 21267 (epilogue): Ditto. 21268 (return): Ditto. 21269 (simple_return): Ditto. 21270 (omit_fp_begin): Output special directive for fp_as_gp. 21271 (omit_fp_end): Output special directive for fp_as_gp. 21272 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp, 21273 mforbid-fp-as-gp): New options. 21274 212752018-06-01 Mark Wielaard <mark@klomp.org> 21276 21277 * dwarf2out.c (dwarf2out_finish): Remove generation of 21278 DW_AT_loclists_base. 21279 212802018-06-01 Eric Botcazou <ebotcazou@adacore.com> 21281 21282 * gimple-ssa-store-merging.c: Include gimple-fold.h. 21283 (struct store_immediate_info): Document BIT_INSERT_EXPR stores. 21284 (struct merged_store_group): Add bit_insertion field. 21285 (dump_char_array): Use standard hexadecimal format. 21286 (merged_store_group::merged_store_group): Set bit_insertion to false. 21287 (merged_store_group::apply_stores): Use optimal buffer size. Deal 21288 with BIT_INSERT_EXPR stores. Move up code updating the mask and 21289 also print the mask in the dump file. 21290 (pass_store_merging::gate): Minor tweak. 21291 (imm_store_chain_info::coalesce_immediate): Fix wrong association 21292 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR 21293 stores with INTEGER_CST stores. 21294 (count_multiple_uses) <BIT_INSERT_EXPR>: New case. 21295 (imm_store_chain_info::output_merged_store): Add try_bitpos variable 21296 and use it throughout. Generate bit insertion sequences if need be. 21297 (pass_store_merging::process_store): Remove redundant condition. 21298 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR. 21299 213002018-06-01 Segher Boessenkool <segher@kernel.crashing.org> 21301 21302 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of 21303 the 128-bit floating point types. Fix function comment. 21304 213052018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 21306 21307 * config/aarch64/aarch64-simd.md 21308 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler 21309 mnemonics. 21310 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler 21311 mnemonics. 21312 213132018-06-01 Richard Sandiford <richard.sandiford@linaro.org> 21314 21315 PR tree-optimization/85989 21316 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member 21317 variable. 21318 (backprop::intersect_uses): Check it when deciding whether this 21319 is a backedge reference. 21320 (backprop::process_block): Add each phi to m_visited_phis 21321 after visiting it, then clear it at the end. 21322 213232018-06-01 Richard Biener <rguenther@suse.de> 21324 21325 * tree-vectorizer.h (vect_dr_stmt): New function. 21326 (vect_get_load_cost): Adjust. 21327 (vect_get_store_cost): Likewise. 21328 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): 21329 Use vect_dr_stmt instead of DR_SMTT. 21330 (vect_record_base_alignments): Likewise. 21331 (vect_calculate_target_alignment): Likewise. 21332 (vect_compute_data_ref_alignment): Likewise and make static. 21333 (vect_update_misalignment_for_peel): Likewise. 21334 (vect_verify_datarefs_alignment): Likewise. 21335 (vector_alignment_reachable_p): Likewise. 21336 (vect_get_data_access_cost): Likewise. Pass down 21337 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR. 21338 (vect_get_peeling_costs_all_drs): Likewise. 21339 (vect_peeling_hash_get_lowest_cost): Likewise. 21340 (vect_enhance_data_refs_alignment): Likewise. 21341 (vect_find_same_alignment_drs): Likewise. 21342 (vect_analyze_data_refs_alignment): Likewise. 21343 (vect_analyze_group_access_1): Likewise. 21344 (vect_analyze_group_access): Likewise. 21345 (vect_analyze_data_ref_access): Likewise. 21346 (vect_analyze_data_ref_accesses): Likewise. 21347 (vect_vfa_segment_size): Likewise. 21348 (vect_small_gap_p): Likewise. 21349 (vectorizable_with_step_bound_p): Likewise. 21350 (vect_prune_runtime_alias_test_list): Likewise. 21351 (vect_analyze_data_refs): Likewise. 21352 (vect_supportable_dr_alignment): Likewise. 21353 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise. 21354 (vect_gen_prolog_loop_niters): Likewise. 21355 * tree-vect-loop.c (vect_analyze_loop_2): Likewise. 21356 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not 21357 modify DR_STMT. 21358 (vect_recog_mask_conversion_pattern): Likewise. 21359 (vect_try_gather_scatter_pattern): Likewise. 21360 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info 21361 to vect_get_store_cost. 21362 (vect_get_store_cost): Get stmt_info instead of DR. 21363 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost. 21364 (vect_get_load_cost): Get stmt_info instead of DR. 21365 213662018-06-01 Richard Biener <rguenther@suse.de> 21367 21368 PR middle-end/86017 21369 * gimple-fold.c (var_decl_component_p): Also allow offsetted 21370 vars wrapped in MEM_REFs. 21371 213722018-06-01 Richard Sandiford <richard.sandiford@linaro.org> 21373 21374 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class): 21375 Fix subreg tests so that we only return a choice between 21376 GENERAL_REGS and FP_REGS if the original classes included both. 21377 213782018-06-01 Richard Biener <rguenther@suse.de> 21379 21380 PR ipa/85960 21381 * tree-ssa-structalias.c (get_function_part_constraint): 21382 Handle NULL fi->decl. 21383 (find_func_aliases_for_call): Properly handle indirect 21384 fi from direct call. 21385 (find_func_clobbers): Likewise. 21386 (ipa_pta_execute): Likewise. 21387 (create_variable_info_for): For functions that are ifunc_resolver 21388 resolve to a varinfo that contains the result of the resolver call. 21389 (associate_varinfo_to_alias): Do not treat ifunc resolvers as 21390 aliases. 21391 213922018-05-31 Michael Collison <michael.collison@arm.com> 21393 21394 * config/aarch64/aarch64.md: 21395 (*fix_to_zero_extenddfdi2): New pattern. 21396 * gcc.target/aarch64/fix_extend1.c: New testcase. 21397 213982018-05-31 Qing Zhao <qing.zhao@oracle.com> 21399 21400 PR middle-end/78809 21401 PR middle-end/83026 21402 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ 21403 and BUILT_IN_STRNCMP_EQ. 21404 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and 21405 BUILT_IN_STRNCMP_EQ. 21406 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the 21407 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ. 21408 (gimple_fold_builtin): Likewise. 21409 * tree-ssa-strlen.c (compute_string_length): New function. 21410 (determine_min_obsize): New function. 21411 (handle_builtin_string_cmp): New function to handle calls to 21412 string compare functions. 21413 (strlen_optimize_stmt): Add handling to builtin string compare 21414 calls. 21415 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call): 21416 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ. 21417 * tree.c (build_common_builtin_nodes): Add new defines of 21418 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ. 21419 214202018-05-31 Jakub Jelinek <jakub@redhat.com> 21421 21422 PR target/85984 21423 * bb-reorder.c (pass_partition_blocks::gate): Return false for 21424 functions with naked attribute. 21425 214262018-05-31 Uros Bizjak <ubizjak@gmail.com> 21427 21428 * config/i386/sse.md (avx_vec_concat<mode>): 21429 Substitute concat_tg_mode mode attribute with xtg_mode. 21430 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto. 21431 (concat_tg_mode): Remove mode attribute. 21432 214332018-05-31 Martin Sebor <msebor@redhat.com> 21434 21435 PR c/82063 21436 * calls.c (alloc_max_size): Correct a logic error/typo. 21437 Treat excessive arguments as infinite. Warn for invalid arguments. 21438 * doc/invoke.texi (-Walloc-size-larger-than): Update. 21439 214402018-05-31 H.J. Lu <hongjiu.lu@intel.com> 21441 21442 PR target/85829 21443 * config/i386/x86-tune.def: Re-enable partial_reg_dependency 21444 and movx for Haswell. 21445 214462018-05-31 Chung-Lin Tang <cltang@codesourcery.com> 21447 Cesar Philippidis <cesar@codesourcery.com> 21448 21449 PR middle-end/85879 21450 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true' 21451 when emitting error on private/firstprivate reductions. 21452 * omp-low.c (lower_omp_target): Avoid reference-type processing 21453 on pointers for firstprivate clause. 21454 214552018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org> 21456 21457 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New. 21458 (st1x2): Likewise. 21459 (st1x3): Likewise. 21460 * config/aarch64/aarch64-simd.md 21461 (aarch64_ld1x3<VALLDIF:mode>): New pattern. 21462 (aarch64_ld1_x3_<mode>): Likewise 21463 (aarch64_st1x2<VALLDIF:mode>): Likewise 21464 (aarch64_st1_x2_<mode>): Likewise 21465 (aarch64_st1x3<VALLDIF:mode>): Likewise 21466 (aarch64_st1_x3_<mode>): Likewise 21467 * config/aarch64/arm_neon.h (vld1_u8_x3): New function. 21468 (vld1_s8_x3): Likewise. 21469 (vld1_u16_x3): Likewise. 21470 (vld1_s16_x3): Likewise. 21471 (vld1_u32_x3): Likewise. 21472 (vld1_s32_x3): Likewise. 21473 (vld1_u64_x3): Likewise. 21474 (vld1_s64_x3): Likewise. 21475 (vld1_f16_x3): Likewise. 21476 (vld1_f32_x3): Likewise. 21477 (vld1_f64_x3): Likewise. 21478 (vld1_p8_x3): Likewise. 21479 (vld1_p16_x3): Likewise. 21480 (vld1_p64_x3): Likewise. 21481 (vld1q_u8_x3): Likewise. 21482 (vld1q_s8_x3): Likewise. 21483 (vld1q_u16_x3): Likewise. 21484 (vld1q_s16_x3): Likewise. 21485 (vld1q_u32_x3): Likewise. 21486 (vld1q_s32_x3): Likewise. 21487 (vld1q_u64_x3): Likewise. 21488 (vld1q_s64_x3): Likewise. 21489 (vld1q_f16_x3): Likewise. 21490 (vld1q_f32_x3): Likewise. 21491 (vld1q_f64_x3): Likewise. 21492 (vld1q_p8_x3): Likewise. 21493 (vld1q_p16_x3): Likewise. 21494 (vld1q_p64_x3): Likewise. 21495 (vst1_s64_x2): Likewise. 21496 (vst1_u64_x2): Likewise. 21497 (vst1_f64_x2): Likewise. 21498 (vst1_s8_x2): Likewise. 21499 (vst1_p8_x2): Likewise. 21500 (vst1_s16_x2): Likewise. 21501 (vst1_p16_x2): Likewise. 21502 (vst1_s32_x2): Likewise. 21503 (vst1_u8_x2): Likewise. 21504 (vst1_u16_x2): Likewise. 21505 (vst1_u32_x2): Likewise. 21506 (vst1_f16_x2): Likewise. 21507 (vst1_f32_x2): Likewise. 21508 (vst1_p64_x2): Likewise. 21509 (vst1q_s8_x2): Likewise. 21510 (vst1q_p8_x2): Likewise. 21511 (vst1q_s16_x2): Likewise. 21512 (vst1q_p16_x2): Likewise. 21513 (vst1q_s32_x2): Likewise. 21514 (vst1q_s64_x2): Likewise. 21515 (vst1q_u8_x2): Likewise. 21516 (vst1q_u16_x2): Likewise. 21517 (vst1q_u32_x2): Likewise. 21518 (vst1q_u64_x2): Likewise. 21519 (vst1q_f16_x2): Likewise. 21520 (vst1q_f32_x2): Likewise. 21521 (vst1q_f64_x2): Likewise. 21522 (vst1q_p64_x2): Likewise. 21523 (vst1_s64_x3): Likewise. 21524 (vst1_u64_x3): Likewise. 21525 (vst1_f64_x3): Likewise. 21526 (vst1_s8_x3): Likewise. 21527 (vst1_p8_x3): Likewise. 21528 (vst1_s16_x3): Likewise. 21529 (vst1_p16_x3): Likewise. 21530 (vst1_s32_x3): Likewise. 21531 (vst1_u8_x3): Likewise. 21532 (vst1_u16_x3): Likewise. 21533 (vst1_u32_x3): Likewise. 21534 (vst1_f16_x3): Likewise. 21535 (vst1_f32_x3): Likewise. 21536 (vst1_p64_x3): Likewise. 21537 (vst1q_s8_x3): Likewise. 21538 (vst1q_p8_x3): Likewise. 21539 (vst1q_s16_x3): Likewise. 21540 (vst1q_p16_x3): Likewise. 21541 (vst1q_s32_x3): Likewise. 21542 (vst1q_s64_x3): Likewise. 21543 (vst1q_u8_x3): Likewise. 21544 (vst1q_u16_x3): Likewise. 21545 (vst1q_u32_x3): Likewise. 21546 (vst1q_u64_x3): Likewise. 21547 (vst1q_f16_x3): Likewise. 21548 (vst1q_f32_x3): Likewise. 21549 (vst1q_f64_x3): Likewise. 21550 (vst1q_p64_x3): Likewise. 21551 215522018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com> 21553 21554 * config/msp430/msp430.c (msp430_output_labelref): Prepend 21555 user_label_prefix to name. 21556 21557 * tree-core.h: Update comment about the format of NAME string 21558 passed to handler in attribute_spec. 21559 21560 * config/msp430/msp430.md: Remove erroneous subreg expression from 21561 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on 21562 zero_extend{q,h}isi2. 21563 215642018-05-30 Borislav Petkov <bp@suse.de> 21565 21566 * doc/extend.texi: Document some architecture specific 21567 constraints and sort entries. 21568 215692018-05-30 Martin Sebor <msebor@redhat.com> 21570 21571 PR middle-end/85369 21572 * builtins.c (expand_builtin_stpcpy_1): New function. 21573 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg 21574 only if the former succeeds. 21575 215762018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org> 21577 21578 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4 21579 in saphira. 21580 215812018-05-30 Jan Hubicka <hubicka@ucw.cz> 21582 21583 * doc/invoke.texi (-flinker-output): Document 21584 215852018-05-30 Jan Hubicka <hubicka@ucw.cz> 21586 21587 * passes.c (ipa_write_summaries): Only modify statements if body 21588 is in memory. 21589 * cgraphunit.c (ipa_passes): Also produce intermeidate code when 21590 incrementally linking. 21591 (ipa_passes): Likewise. 21592 * lto-cgraph.c (lto_output_node): When incrementally linking do not 21593 pass down resolution info. 21594 * common.opt (flag_incremental_link): Update info. 21595 * gcc.c (plugin specs): Turn flinker-output=* to 21596 -plugin-opt=-linker-output-known 21597 * toplev.c (compile_file): Also cut compilation when doing incremental 21598 link. 21599 * flag-types. (enum lto_partition_model): Add 21600 LTO_LINKER_OUTPUT_NOLTOREL. 21601 (invoke.texi): Add -flinker-output docs. 21602 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental 21603 link same way as WPA; do not stream in dead initializers. 21604 21605 * dwarf2out.c (dwarf2out_die_ref_for_decl, 21606 darf2out_register_external_decl): Support incremental link. 21607 216082018-05-30 Jan Hubicka <hubicka@ucw.cz> 21609 21610 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_. 21611 216122018-05-30 Jan Hubicka <hubicka@ucw.cz> 21613 21614 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass 21615 it down to simple_object_copy_lto_debug_sections. 21616 (run_gcc): Determine incremental LTO link time and configure 21617 lto1 into non-wpa mode, disable renaming of debug sections. 21618 216192018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org> 21620 21621 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove 21622 descriptions of various incorrectly documented functions. 21623 216242018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com> 21625 21626 Revert: 21627 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make 21628 address check not strict. 21629 216302018-05-30 Richard Biener <rguenther@suse.de> 21631 21632 PR tree-optimization/85964 21633 * tracer.c (better_p): Drop initialized count check, we only 21634 call the function with initialized counts now. 21635 (find_best_successor): Do find a best edge if one 21636 has uninitialized count. 21637 (find_best_predecessor): Likewise. Do BB frequency check only 21638 if count is initialized. 21639 216402017-05-30 Jackson Woodruff <jackson.woodruff@arm.com> 21641 21642 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New. 21643 (aarch64_ldrstr_offset_compare): New. 21644 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all 21645 load/store orderings. 21646 (aarch64_gen_adjusted_ldpstp): Likewise. 21647 216482018-05-30 Wilco Dijkstra <wdijkstr@arm.com> 21649 21650 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class): 21651 Check for subset of GENERAL_REGS and FP_REGS. 21652 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of 21653 r=w alternative. 21654 216552018-05-30 Richard Sandiford <richard.sandiford@linaro.org> 21656 21657 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p 21658 and wi::to_poly_offset. Add the current offset and then check 21659 whether the sum fits, rather than using an unchecked addition of 21660 a checked term. Check for a shwi rather than a uhwi. 21661 * expr.c (get_bit_range): Use tree_to_poly_uint64. 21662 (store_constructor): Use poly_int_tree_p. 21663 (expand_expr_real_1): Likewise. 21664 * function.c (assign_temp): Likewise. 21665 * fold-const.c (const_binop): Use poly_int_tree_p and 21666 wi::to_poly_offset. 21667 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact 21668 division. 21669 * ipa-icf-gimple.c (func_checker::compare_operand): Use 21670 to_poly_offset for MEM offsets. 21671 * ipa-icf.c (sem_variable::equals): Likewise. 21672 * stor-layout.c (compute_record_mode): Use poly_int_tree_p. 21673 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use 21674 wi::to_poly_offset for BIT_FIELD_REF offsets. 21675 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and 21676 wi::to_poly_offset. 21677 * var-tracking.c (emit_note_insn_var_location): Use 21678 tree_to_poly_uint64. 21679 216802018-05-29 Jim Wilson <jimw@sifive.com> 21681 21682 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo. 21683 216842018-05-29 Uros Bizjak <ubizjak@gmail.com> 21685 21686 PR target/85950 21687 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2): 21688 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q} 21689 sequence. 21690 (sse4_1_round<mode>2): Use nonimmediate_operand 21691 for operand 1 predicate. 21692 216932018-05-29 Martin Sebor <msebor@redhat.com> 21694 Richard Biener <rguenther@suse.de> 21695 21696 PR testsuite/85888 21697 * calls.c (get_size_range): Call determine_value_range instead 21698 of get_value_range.. 21699 * tree-vrp.h (determine_value_range): Declared new function. 21700 * tree-vrp.c (determine_value_range_1, determine_value_range): New. 21701 217022018-05-29 Richard Biener <rguenther@suse.de> 21703 21704 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make 21705 sure to use non-pattern stmts for get_earlier_stmt arguments. 21706 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get 21707 called on pattern stmts. 21708 (get_later_stmt): Likewise. 21709 217102018-05-29 Martin Liska <mliska@suse.cz> 21711 21712 PR gcov-profile/85759 21713 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR 21714 env variables. 21715 217162018-05-29 Jakub Jelinek <jakub@redhat.com> 21717 21718 * tree-cfg.c (verify_gimple_assign_unary): Add checking for 21719 VEC_UNPACK_*_EXPR. 21720 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for 21721 VEC_PACK_*_EXPR. 21722 21723 PR target/85918 21724 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR, 21725 VEC_PACK_FLOAT_EXPR): New tree codes. 21726 * tree-pretty-print.c (op_code_prio): Handle 21727 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR. 21728 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR, 21729 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR. 21730 * tree-inline.c (estimate_operator_cost): Likewise. 21731 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR. 21732 * fold-const.c (const_binop): Likewise. 21733 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and 21734 VEC_UNPACK_FIX_TRUNC_LO_EXPR. 21735 * tree-cfg.c (verify_gimple_assign_unary): Likewise. 21736 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR. 21737 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR, 21738 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR. 21739 * expr.c (expand_expr_real_2): Likewise. 21740 * optabs.def (vec_packs_float_optab, vec_packu_float_optab, 21741 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab, 21742 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New 21743 optabs. 21744 * optabs.c (expand_widen_pattern_expr): For 21745 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use 21746 sign from result type rather than operand's type. 21747 (expand_binop_directly): For vec_packu_float_optab and 21748 vec_packs_float_optab allow result type to be different from operand's 21749 type. 21750 * optabs-tree.c (optab_for_tree_code): Handle 21751 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and 21752 VEC_PACK_FLOAT_EXPR. Formatting fixes. 21753 * tree-vect-generic.c (expand_vector_operations_1): Handle 21754 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and 21755 VEC_PACK_FLOAT_EXPR. 21756 * tree-vect-stmts.c (supportable_widening_operation): Handle 21757 FIX_TRUNC_EXPR. 21758 (supportable_narrowing_operation): Handle FLOAT_EXPR. 21759 * config/i386/i386.md (fixprefix, floatprefix): New code attributes. 21760 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ... 21761 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix. 21762 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New 21763 mode attributes. 21764 (vec_pack<floatprefix>_float_<mode>): New expander. 21765 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode 21766 attributes. 21767 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>, 21768 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders. 21769 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m}, 21770 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m}, 21771 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}): 21772 Document. 21773 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR, 21774 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description. 21775 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR, 21776 VEC_PACK_FLOAT_EXPR): Document. 21777 217782018-05-29 Richard Biener <rguenther@suse.de> 21779 21780 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos 21781 member. 21782 (stmt_vec_info_vec): Make pointer. 21783 (init_stmt_vec_info_vec): Remove. 21784 (free_stmt_vec_info_vec): Likewise. 21785 (set_stmt_vec_info_vec): New function. 21786 (free_stmt_vec_infos): Likewise. 21787 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection. 21788 (set_vinfo_for_stmt): Likewise. 21789 (get_earlier_stmt): Likewise. 21790 (get_later_stmt): Likewise. 21791 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer. 21792 (vec_info::vec_info): Allocate stmt_vec_infos and set the global. 21793 (vec_info::~vec_info): Free stmt_vec_infos. 21794 (vectorize_loops): Set the global stmt_vec_info_vec to NULL. 21795 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls. 21796 (pass_slp_vectorize::execute): Likewise. 21797 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove. 21798 (free_stmt_vec_info_vec): Likewise. 21799 (set_stmt_vec_info_vec): New function. 21800 (free_stmt_vec_infos): Likewise. 21801 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set 21802 the global stmt_vec_info_vec. 21803 * tree-parloops.c (gather_scalar_reductions): Use 21804 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local 21805 vector. 21806 218072018-05-29 Richard Biener <rguenther@suse.de> 21808 21809 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE. 21810 218112018-05-29 Martin Liska <mliska@suse.cz> 21812 David Malcolm <dmalcolm@redhat.com> 21813 21814 * vec.c (test_reverse): New. 21815 (vec_c_tests): Add new test. 21816 * vec.h (vl_ptr>::reverse): New function. 21817 218182018-05-29 Gerald Pfeifer <gerald@pfeifer.com> 21819 21820 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported. 21821 21822 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5 21823 and later. 21824 218252018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de> 21826 21827 * tree-dump.c (dump_node): Use splay_tree_delete_pointers. 21828 218292018-05-28 Richard Biener <rguenther@suse.de> 21830 21831 PR tree-optimization/85933 21832 * tree-vect-data-refs.c (vect_record_base_alignments): Only 21833 look at stmts marked as vectorizable. 21834 218352018-05-28 Richard Biener <rguenther@suse.de> 21836 21837 PR tree-optimization/85934 21838 * tree-vect-generic.c (expand_vector_operations_1): Hoist 21839 vector boolean check before scalar optimization. 21840 218412018-05-28 Jakub Jelinek <jakub@redhat.com> 21842 21843 * doc/invoke.texi (ARM Options): Use @item instead of @itemx 21844 for armv5te. 21845 218462018-05-28 Mark Wielaard <mark@klomp.org> 21847 21848 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2 21849 if it is an expression containing a minus sign. 21850 218512018-05-27 John David Anglin <danglin@gcc.gnu.org> 21852 21853 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0. 21854 218552018-05-27 Paul Koning <ni1d@arrl.net> 21856 21857 * config/pdp11/pdp11.md (truncsihi2): Remove. 21858 218592018-05-27 Monk Chiang <sh.chiang04@gmail.com> 21860 Chung-Ju Wu <jasonwucj@gmail.com> 21861 21862 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns 21863 implementation. 21864 (unaligned_store_dw): Ditto. 21865 * config/nds32/nds32-memory-manipulation.c 21866 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation. 21867 (nds32_gen_dup_4_byte_to_word_value): Rename to ... 21868 (nds32_gen_dup_4_byte_to_word_value_aux): ... this. 21869 (emit_setmem_word_loop): Rename to ... 21870 (emit_setmem_doubleword_loop): ... this. 21871 (nds32_gen_dup_4_byte_to_word_value): New function. 21872 (nds32_gen_dup_8_byte_to_double_word_value): New function. 21873 (nds32_expand_setmem_loop): Refine implementation. 21874 (nds32_expand_setmem_loop_v3m): Ditto. 21875 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New 21876 pattern. 21877 218782018-05-27 Chung-Ju Wu <jasonwucj@gmail.com> 21879 21880 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns. 21881 218822018-05-27 Chung-Ju Wu <jasonwucj@gmail.com> 21883 21884 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue". 21885 (nds32_init_machine_status): Initialize machine->attr_naked_p and 21886 machine->attr_no_prologue_p. 21887 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes. 21888 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes. 21889 (nds32_expand_epilogue): Consider attr_naked_p. 21890 (nds32_expand_epilogue_v3pop): Likewise. 21891 (nds32_can_use_return_insn): Likewise. 21892 * config/nds32/nds32.h (machine_function): Add attr_naked_p and 21893 attr_no_prologue_p fields. 21894 * config/nds32/nds32.opt (mret-in-naked-func): New option. 21895 218962018-05-27 Jakub Jelinek <jakub@redhat.com> 21897 21898 PR target/85918 21899 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code 21900 attributes. 21901 * config/i386/sse.md 21902 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>): 21903 Rename to ... 21904 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>): 21905 ... this. 21906 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>): 21907 Rename to ... 21908 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>): 21909 ... this. 21910 (*<floatsuffix>floatv2div2sf2): Rename to ... 21911 (*float<floatunssuffix>v2div2sf2): ... this. 21912 (<floatsuffix>floatv2div2sf2_mask): Rename to ... 21913 (float<floatunssuffix>v2div2sf2_mask): ... this. 21914 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ... 21915 (*float<floatunssuffix>v2div2sf2_mask_1): ... this. 21916 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename 21917 to ... 21918 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>): 21919 ... this. 21920 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>): 21921 Rename to ... 21922 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>): 21923 ... this. 21924 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>): 21925 Rename to ... 21926 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>): 21927 ... this. 21928 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ... 21929 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this. 21930 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of 21931 gen_ufix_truncv8dfv8si2. 21932 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask, 21933 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask, 21934 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask, 21935 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask, 21936 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask, 21937 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask, 21938 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask): 21939 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names. 21940 219412018-05-24 H.J. Lu <hongjiu.lu@intel.com> 21942 21943 PR target/85900 21944 PR target/85345 21945 * varasm.c (assemble_alias): Lookup ifunc attribute on error. 21946 219472018-05-25 Jim Wilson <jimw@sifive.com> 21948 21949 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New. 21950 * config/riscv/riscv.c (struct machine_function): Add 21951 interrupt_handler_p and attribute_checked_p fields. 21952 (riscv_attribute_table): Add interrupt. 21953 (riscv_interrupt_type_p): New. 21954 (riscv_save_reg_p): Save extra regs for interrupt handler. 21955 (riscv_use_save_libcall): Return false for interrupt handler. 21956 (riscv_first_stack_step): Add forward declaration. 21957 (riscv_compute_frame_info): New local interrupt_save_t1. Set it 21958 for interrupt handler with large frame. Use it for saved reg list. 21959 (riscv_expand_prologue): Move flag_stack_usage_info support to 21960 eliminate duplication. 21961 (riscv_expand_epilogue): Generate mret for interrupt handler. 21962 (riscv_epilogue_uses): New. 21963 (riscv_can_use_return_insn): Return false for interrupt handler. 21964 (riscv_function_ok_for_sibcall): Likewise. 21965 (riscv_set_current_function): Add interrupt handler support. 21966 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses. 21967 * config/riscv/riscv.md (UNSPECV_MRET): New. 21968 (GP_REGNUM): New. 21969 (riscv_frflags, riscv_fsflags): Use tab after opcode. 21970 (riscv_mret): New. 21971 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New. 21972 219732018-05-25 Bill Schmidt <wschmidt@linux.ibm.com> 21974 21975 PR tree-optimization/85712 21976 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if 21977 this candidate has already been replaced in-situ by a copy. 21978 219792018-05-25 Jason Merrill <jason@redhat.com> 21980 21981 PR c++/80485 - inline function non-zero address. 21982 * symtab.c (nonzero_address): Check DECL_COMDAT. 21983 219842018-05-25 Uros Bizjak <ubizjak@gmail.com> 21985 21986 PR target/83628 21987 * config/alpha/alpha.md (ashlsi3): New insn pattern. 21988 (*ashlsi_se): Rename from *ashldi_se. Define as sign 21989 extension of SImode operation. Use const123_operand predicate. 21990 (*saddsi_1): Remove. 21991 (*saddl_se_1): Ditto. 21992 (*ssubsi_1): Ditto. 21993 (*ssubl_se_1): Ditto. 21994 * config/alpha/predicates.md (const123_operand): New predicate. 21995 * config/alpha/constraints.md (P): Use IN_RANGE. 21996 219972018-05-25 Richard Biener <rguenther@suse.de> 21998 21999 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter, 22000 defaulted to true. 22001 (ref_maybe_used_by_stmt_p): Likewise. 22002 (stmt_may_clobber_ref_p): Likewise. 22003 (stmt_may_clobber_ref_p_1): Likewise. 22004 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter 22005 and pass it along. 22006 (ref_maybe_used_by_stmt_p): Likewise. 22007 (stmt_may_clobber_ref_p): Likewise. 22008 (stmt_may_clobber_ref_p_1): Likewise. 22009 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use 22010 the alias oracle to disambiguate DRs with stmts DR analysis 22011 couldn't handle. 22012 (vect_analyze_data_refs): Do not give up on not analyzable 22013 DRs for BB vectorization. Remove code truncating the dataref 22014 vector. 22015 220162018-05-25 Jakub Jelinek <jakub@redhat.com> 22017 22018 PR target/85832 22019 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): 22020 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW 22021 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator. 22022 220232018-05-25 Richard Biener <rguenther@suse.de> 22024 22025 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New 22026 function, combining stmt data ref gathering and fatal analysis 22027 parts. 22028 (vect_analyze_data_refs): Remove now redudnant code and simplify. 22029 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from 22030 vect_analyze_loop_2 and use vect_find_stmt_data_reference. 22031 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference. 22032 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare. 22033 220342018-05-25 Bin Cheng <bin.cheng@arm.com> 22035 22036 PR tree-optimization/85720 22037 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge 22038 SCC if all partitions are builtins. 22039 (version_loop_by_alias_check): New parameter. Generate cancelable 22040 runtime alias check if all partitions are builtins. 22041 (distribute_loop): Update call to above function. 22042 220432018-05-25 Bin Cheng <bin.cheng@arm.com> 22044 22045 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files. 22046 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c. 22047 (parm_default_def_partition_arg): Ditto. 22048 (set_parm_default_def_partition): Ditto. 22049 (get_parm_default_def_partitions): Ditto and make it static. 22050 (get_undefined_value_partitions): Ditto and make it static. 22051 (remove_ssa_form): Refactor call to init_var_map here. 22052 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range 22053 computation for loop region. 22054 (coalesce_partitions, compute_optimized_partition_bases): Ditto. 22055 (register_default_def): Delete. 22056 (for_all_parms, create_default_def): Move to tree-outof-ssa.c. 22057 (parm_default_def_partition_arg): Ditto. 22058 (set_parm_default_def_partition): Ditto. 22059 (get_parm_default_def_partitions): Ditto and make it static. 22060 (get_undefined_value_partitions): Ditto and make it static. 22061 (coalesce_with_default, coalesce_with_default): Update comment. 22062 (create_coalesce_list_for_region): New func factored out from 22063 create_outofssa_var_map. 22064 (populate_coalesce_list_for_outofssa): New func factored out from 22065 create_outofssa_var_map and coalesce_ssa_name. 22066 (create_outofssa_var_map): Delete. 22067 (coalesce_ssa_name): Refactor to support live range computation. 22068 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl. 22069 (get_parm_default_def_partitions): Delete. 22070 (get_undefined_value_partitions): Ditto. 22071 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range 22072 computation for loop region. 22073 (new_tree_live_info, loe_visit_block): Ditto. 22074 (live_worklist, set_var_live_on_entry): Ditto. 22075 (calculate_live_on_exit, verify_live_on_entry): Ditto. 22076 * tree-ssa-live.h (struct _var_map): New fields. 22077 (init_var_map): Change decl. 22078 (region_contains_p): New. 22079 220802018-05-25 Bin Cheng <bin.cheng@arm.com> 22081 22082 * tree-ssa-live.h (live_merge_and_clear): Delete. 22083 220842018-05-25 Richard Biener <rguenther@suse.de> 22085 22086 PR c++/85912 22087 * tree-dump.c (dequeue_and_dump): Remove access to removed 22088 operand 2 of a SWITCH_EXPR. 22089 220902018-05-25 Richard Sandiford <richard.sandiford@linaro.org> 22091 22092 * doc/sourcebuild.texi (vect_double_cond_arith): Include 22093 multiplication and division. 22094 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m}) 22095 (cond_udiv@var{m}, cond_umod@var{m}): Document. 22096 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab) 22097 (cond_udiv_optab, cond_umod_optab): New optabs. 22098 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD) 22099 (IFN_COND_RDIV): New internal functions. 22100 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR, 22101 TRUNC_MOD_EXPR and RDIV_EXPR. 22102 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them. 22103 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV): 22104 New unspecs. 22105 (SVE_INT_BINARY): Include mult. 22106 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV. 22107 (optab, sve_int_op): Handle mult. 22108 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and 22109 UNSPEC_COND_DIV. 22110 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern 22111 for SVE_INT_BINARY_SD. 22112 221132018-05-25 Richard Sandiford <richard.sandiford@linaro.org> 22114 22115 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator. 22116 (optab, sve_int_op): Handle div and udiv. 22117 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander 22118 for SVE_INT_BINARY_SD. 22119 (*<optab><mode>3): New insn for the same. 22120 221212018-05-25 Richard Sandiford <richard.sandiford@linaro.org> 22122 22123 * tree-vect-patterns.c: Include predict.h. 22124 (vect_recog_divmod_pattern): Restrict check for division support 22125 to when optimizing for size. 22126 221272018-05-25 Richard Sandiford <richard.sandiford@linaro.org> 22128 22129 * doc/sourcebuild.texi (vect_double_cond_arith: Document. 22130 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4. 22131 (gimple_match_op::gimple_match_op): Add an overload for 4 operands. 22132 (gimple_match_op::set_op): Likewise. 22133 (gimple_resimplify4): Declare. 22134 * genmatch.c (get_operand_type): Handle CFN_COND_* functions. 22135 (expr::gen_transform): Likewise. 22136 (decision_tree::gen): Generate a simplification routine for 4 operands. 22137 * gimple-match-head.c (gimple_simplify): Add an overload for 22138 4 operands. In the top-level function, handle up to 4 call 22139 arguments and call gimple_resimplify4. 22140 (gimple_resimplify4): New function. 22141 (build_call_internal): Pass a fourth operand. 22142 (maybe_push_to_seq): Likewise. 22143 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists. 22144 Fold VEC_COND_EXPRs of an operation and a default value into 22145 an IFN_COND_* function if possible. 22146 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN): 22147 New unspecs. 22148 (SVE_COND_FP_BINARY): Include them. 22149 (optab, sve_fp_op): Handle them. 22150 (SVE_INT_BINARY_REV): New code iterator. 22151 (SVE_COND_FP_BINARY_REV): New int iterator. 22152 (commutative): New int attribute. 22153 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op): 22154 Declare. 22155 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New 22156 function. 22157 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it. 22158 (*cond_<optab><mode>): New patterns for reversed operands. 22159 221602018-05-25 Richard Biener <rguenther@suse.de> 22161 22162 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove. 22163 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info. 22164 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info. 22165 (STMT_VINFO_GROUPED_ACCESS): Adjust. 22166 * tree-vect-data-refs.c (everywhere): Adjust users. 22167 * tree-vect-loop.c (everywhere): Likewise. 22168 * tree-vect-slp.c (everywhere): Likewise. 22169 * tree-vect-stmts.c (everywhere): Likewise. 22170 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise. 22171 221722018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE> 22173 22174 * configure.ac (gcc_cv_as_section_has_e): Move to common section. 22175 Rename to... 22176 (gcc_cv_as_section_exclude): ... this. 22177 Try Solaris as #exclude syntax. 22178 * configure: Regenerate. 22179 * config.in: Regenerate. 22180 * config/i386/i386.c (i386_solaris_elf_named_section): Handle 22181 SECTION_EXCLUDE. 22182 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section) 22183 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE. 22184 22185 * varasm.c (default_elf_asm_named_section): Don't check if 22186 HAVE_GAS_SECTION_EXCLUDE is defined. 22187 221882018-05-25 Richard Sandiford <richard.sandiford@linaro.org> 22189 22190 * doc/md.texi: Update the documentation of the cond_* optabs 22191 to mention the new final operand. Fix GET_MODE_NUNITS call. 22192 Describe the scalar case too. 22193 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left. 22194 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands 22195 instead of 2. 22196 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3. 22197 (get_conditional_internal_fn): Update comment. 22198 * tree-vect-loop.c (vectorizable_reduction): Pass the original 22199 accumulator value as a final argument to conditional functions. 22200 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into 22201 a define_expand and add an "else" operand. Assert for now that 22202 the else operand is equal to operand 2. Use SVE_INT_BINARY and 22203 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP. 22204 (*cond_<optab><mode>): New patterns. 22205 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX) 22206 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR) 22207 (UNSPEC_COND_EOR): Delete. 22208 (optab): Remove associated mappings. 22209 (SVE_INT_BINARY): New code iterator. 22210 (sve_int_op): Remove int attribute and add "minus" to the code 22211 attribute. 22212 (SVE_COND_INT_OP): Delete. 22213 (SVE_COND_FP_OP): Rename to... 22214 (SVE_COND_FP_BINARY): ...this. 22215 222162018-05-25 Richard Sandiford <richard.sandiford@linaro.org> 22217 22218 * optabs.c (can_reuse_operands_p): New function. 22219 (maybe_legitimize_operands): Try to reuse the results for 22220 earlier operands. 22221 222222018-05-24 Uros Bizjak <ubizjak@gmail.com> 22223 22224 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>): 22225 Add {q} suffix to insn mnemonic. 22226 222272018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com> 22228 22229 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define. 22230 (msp430_warn_func_return): New. 22231 222322018-05-24 Roger Sayle <roger@nextmovesoftware.com> 22233 22234 * fold-const.c (tree_nonzero_bits): New function. 22235 * fold-const.h (tree_nonzero_bits): Likewise. 22236 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and 22237 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc. 22238 222392018-05-24 H.J. Lu <hongjiu.lu@intel.com> 22240 22241 PR target/85900 22242 PR target/85345 22243 * varasm.c (assemble_alias): Check ifunc_resolver only on 22244 FUNCTION_DECL. 22245 222462018-05-24 Uros Bizjak <ubizjak@gmail.com> 22247 22248 PR target/85903 22249 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo 22250 when memory input operand is handled. 22251 222522018-05-24 Luis Machado <luis.machado@linaro.org> 22253 22254 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static 22255 global. 22256 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table. 22257 222582018-05-24 Richard Sandiford <richard.sandiford@linaro.org> 22259 22260 * match.pd: Delay FMA folds until after vectorization. 22261 222622018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com> 22263 22264 PR target/83009 22265 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make 22266 address check not strict. 22267 222682018-05-24 Richard Sandiford <richard.sandiford@linaro.org> 22269 22270 * gimple-match.h (gimple_match_op): New class. 22271 (mprts_hook): Replace parameters with a gimple_match_op *. 22272 (maybe_build_generic_op): Likewise. 22273 (gimple_simplified_result_is_gimple_val): Replace parameters with 22274 a const gimple_match_op *. 22275 (gimple_simplify): Replace code_helper * and tree * parameters with 22276 a gimple_match_op * parameter. 22277 (gimple_resimplify1): Replace code_helper *, tree and tree * 22278 parameters with a gimple_match_op * parameter. 22279 (gimple_resimplify2): Likewise. 22280 (gimple_resimplify3): Likewise. 22281 (maybe_push_res_to_seq): Replace code_helper, tree and tree * 22282 parameters with a gimple_match_op * parameter. 22283 * gimple-match-head.c (gimple_simplify): Change prototypes of 22284 auto-generated functions to take a gimple_match_op * instead of 22285 separate code_helper * and tree * parameters. Make the same 22286 change in the top-level overload and update calls to the 22287 gimple_resimplify routines. Update calls to the auto-generated 22288 functions and to maybe_push_res_to_seq in the publicly-facing 22289 operation-specific gimple_simplify overloads. 22290 (gimple_match_op::MAX_NUM_OPS): Define. 22291 (gimple_resimplify1): Replace rcode and ops with a single res_op 22292 parameter. Update call to gimple_simplify. 22293 (gimple_resimplify2): Likewise. 22294 (gimple_resimplify3): Likewise. 22295 (mprts_hook): Replace parameters with a gimple_match_op *. 22296 (maybe_build_generic_op): Likewise. 22297 (build_call_internal): Replace type, nargs and ops with 22298 a gimple_match_op *. 22299 (maybe_push_res_to_seq): Replace res_code, type and ops parameters 22300 with a single gimple_match_op *. Update calls to mprts_hook, 22301 build_call_internal and gimple_simplified_result_is_gimple_val. 22302 Factor out code that is common to the tree_code and combined_fn cases. 22303 * genmatch.c (expr::gen_transform): Replace tem_code and 22304 tem_ops with a gimple_match_op called tem_op. Update calls 22305 to the gimple_resimplify functions and maybe_push_res_to_seq. 22306 (dt_simplify::gen_1): Manipulate res_op instead of res_code and 22307 res_ops. Update call to the gimple_resimplify functions. 22308 (dt_simplify::gen): Pass res_op instead of res_code and res_ops. 22309 (decision_tree::gen): Make the functions take a gimple_match_op * 22310 called res_op instead of separate res_code and res_ops parameters. 22311 Update call accordingly. 22312 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode 22313 and ops with a single res_op parameter. Update calls to 22314 maybe_build_generic_op and maybe_push_res_to_seq. 22315 (fold_stmt_1): Update calls to gimple_simplify and 22316 replace_stmt_with_simplification. 22317 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify 22318 and gimple_simplified_result_is_gimple_val. 22319 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to 22320 gimple_simplify. 22321 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters 22322 with a gimple_match_op *. 22323 (vn_nary_build_or_lookup): Likewise. Update call to 22324 vn_nary_build_or_lookup_1. 22325 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a 22326 gimple_match_op *. Update calls to the gimple_resimplify routines 22327 and to gimple_simplified_result_is_gimple_val. 22328 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1. 22329 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3. 22330 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup. 22331 (visit_nary_op): Likewise. 22332 (visit_reference_op_load): Likewise. 22333 223342018-05-23 Luis Machado <luis.machado@linaro.org> 22335 22336 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type 22337 modifier for printing the step amount. 22338 223392018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com> 22340 22341 PR target/78849 22342 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN 22343 types. 22344 223452018-05-23 Segher Boessenkool <segher@kernel.crashing.org> 22346 22347 * doc/sourcebuild.texi (Endianness): New subsubsection. 22348 223492018-05-23 Luis Machado <luis.machado@linaro.org> 22350 22351 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune) 22352 <prefetch_dynamic_strides>: New const bool field. 22353 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include 22354 prefetch_dynamic_strides. 22355 (exynosm1_prefetch_tune): Likewise. 22356 (thunderxt88_prefetch_tune): Likewise. 22357 (thunderx_prefetch_tune): Likewise. 22358 (thunderx2t99_prefetch_tune): Likewise. 22359 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to 22360 false. 22361 (aarch64_override_options_internal): Update to set 22362 PARAM_PREFETCH_DYNAMIC_STRIDES. 22363 * doc/invoke.texi (prefetch-dynamic-strides): Document new option. 22364 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New. 22365 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define. 22366 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for 22367 prefetch-dynamic-strides setting. 22368 223692018-05-23 Luis Machado <luis.machado@linaro.org> 22370 22371 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune) 22372 <minimum_stride>: New const int field. 22373 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include 22374 minimum_stride field defaulting to -1. 22375 (exynosm1_prefetch_tune): Likewise. 22376 (thunderxt88_prefetch_tune): Likewise. 22377 (thunderx_prefetch_tune): Likewise. 22378 (thunderx2t99_prefetch_tune): Likewise. 22379 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048. 22380 <default_opt_level>: Set to 3. 22381 (aarch64_override_options_internal): Update to set 22382 PARAM_PREFETCH_MINIMUM_STRIDE. 22383 * doc/invoke.texi (prefetch-minimum-stride): Document new option. 22384 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New. 22385 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define. 22386 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if 22387 stride is constant and is below the minimum stride threshold. 22388 223892018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 22390 22391 * config/arm/arm-cpus.in (mode26): Delete. 22392 (armv4): Delete mode26 reference. 22393 * config/arm/arm.c (arm_configure_build_target): Delete use of 22394 isa_bit_mode26. 22395 223962018-05-23 Uros Bizjak <ubizjak@gmail.com> 22397 22398 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512): 22399 New insn pattern. 22400 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH. 22401 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm 22402 for non-SSE modes. 22403 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F. 22404 (floatunsdidf2): Ditto. 22405 224062018-05-23 Uros Bizjak <ubizjak@gmail.com> 22407 22408 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern. 22409 (fixuns_trunc<mode>si2_avx512f): Ditto. 22410 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto. 22411 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH. 22412 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets. 22413 224142018-05-23 Alexander Monakov <amonakov@ispras.ru> 22415 22416 PR rtl-optimization/79985 22417 * df-scan.c (df_insn_refs_collect): Remove special case for 22418 global registers and asm statements. 22419 224202018-05-23 Alexander Monakov <amonakov@ispras.ru> 22421 22422 * extend.texi (Global Register Variables): Rewrite the bullet list. 22423 Note that the register is available for allocation. Note that access 22424 via inline asm must use constraints. Add note about async-signal 22425 handlers. Remove paragraph about automagic register selection. 22426 224272018-05-23 Richard Biener <rguenther@suse.de> 22428 22429 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction 22430 of fixed offset from memset VN. 22431 224322018-05-23 Bill Schmidt <wschmidt@linux.ibm.com> 22433 22434 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add 22435 first_interp field. 22436 (alloc_cand_and_find_basis): Initialize first_interp field. 22437 (slsr_process_mul): Modify first_interp field. 22438 (slsr_process_add): Likewise. 22439 (slsr_process_cast): Modify first_interp field for each new 22440 interpretation. 22441 (slsr_process_copy): Likewise. 22442 (dump_candidate): Dump first_interp field. 22443 (replace_mult_candidate): Process all interpretations, not just 22444 subsequent ones. 22445 (replace_rhs_if_not_dup): Likewise. 22446 (replace_one_candidate): Likewise. 22447 224482018-05-23 Wilco Dijkstra <wdijkstr@arm.com> 22449 22450 * config/aarch64/aarch64.c (aarch64_use_frame_pointer): 22451 Add new boolean. 22452 (aarch64_needs_frame_chain): New function. 22453 (aarch64_parse_override_string): Set aarch64_use_frame_pointer. 22454 224552018-05-23 Sudakshina Das <sudi.das@arm.com> 22456 22457 PR target/84882 22458 * common/config/aarch64/aarch64-common.c (aarch64_handle_option): 22459 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags. 22460 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative. 22461 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg 22462 as true for strict-align. 22463 (aarch64_can_inline_p): Perform checks even when callee has no 22464 attributes to check for strict alignment. 22465 * doc/extend.texi (AArch64 Function Attributes): Document 22466 no-strict-align. 22467 * doc/invoke.texi: (AArch64 Options): Likewise. 22468 224692018-05-23 Richard Sandiford <richard.sandiford@linaro.org> 22470 22471 PR tree-optimization/85853 22472 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out 22473 the handling of the root of the node to... 22474 (vect_slp_analyze_node_operations_1): ...this new function, 22475 and run the whole thing with the child nodes' def types 22476 set according to their SLP node's def type. 22477 224782018-05-23 Richard Biener <rguenther@suse.de> 22479 22480 PR middle-end/85874 22481 * tree-data-ref.c (create_runtime_alias_checks): Defer 22482 and ignore overflow warnings. 22483 224842018-05-23 Yury Gribov <tetra2005@gmail.com> 22485 22486 PR tree-optimization/85822 22487 * tree-vrp.c (is_masked_range_test): Fix handling of negative 22488 constants. 22489 224902018-05-23 Richard Biener <rguenther@suse.de> 22491 22492 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary 22493 memset constants via native_interpret_expr. 22494 224952018-05-22 H.J. Lu <hongjiu.lu@intel.com> 22496 22497 PR target/85345 22498 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc 22499 attribute. 22500 (cgraph_node::create_alias): Likewise. 22501 (cgraph_node::get_availability): Check ifunc_resolver instead 22502 of looking up ifunc attribute. 22503 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise. 22504 * varasm.c (do_assemble_alias): Likewise. 22505 (assemble_alias): Likewise. 22506 (default_binds_local_p_3): Likewise. 22507 * cgraph.h (cgraph_node): Add ifunc_resolver. 22508 (cgraph_node::only_called_directly_or_aliased_p): Return false 22509 for IFUNC resolver. 22510 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc 22511 attribute. 22512 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver 22513 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)). 22514 (symtab_node::binds_to_current_def_p): Check ifunc_resolver 22515 instead of looking up ifunc attribute. 22516 225172018-05-22 Luis Machado <luis.machado@linaro.org> 22518 22519 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern. 22520 225212018-05-22 Martin Sebor <msebor@redhat.com> 22522 22523 PR middle-end/85359 22524 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg 22525 only when expasion succeeds. 22526 (expand_builtin_strcmp): Same. 22527 (expand_builtin_strncmp): Same. 22528 225292018-05-22 Martin Sebor <msebor@redhat.com> 22530 22531 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment. 22532 225332018-05-22 Jackson Woodruff <jackson.woodruff@arm.com> 22534 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 22535 22536 * config/aarch64/aarch64-ldpstp.md: Replace uses of 22537 aarch64_mem_pair_operand with memory_operand and delete operand swapping 22538 code. 22539 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp): 22540 Add check for legitimate_address. 22541 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate. 22542 (aarch64_swap_ldrstr_operands): New. 22543 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands): 22544 Define prototype. 22545 225462018-05-22 Jackson Woodruff <jackson.woodruff@arm.com> 22547 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 22548 22549 * config/aarch64/aarch64.md: New patterns to generate stp 22550 and ldp. 22551 (store_pair_sw, store_pair_dw): New patterns to generate stp for 22552 single words and double words. 22553 (load_pair_sw, load_pair_dw): Likewise. 22554 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di): 22555 Delete. 22556 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di): 22557 Delete. 22558 * config/aarch64/aarch64-ldpstp.md: Modify peephole 22559 for different mode ldpstp and add peephole for merged zero stores. 22560 Likewise for loads. 22561 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp): 22562 Add size check. 22563 (aarch64_gen_store_pair): Rename calls to match new patterns. 22564 (aarch64_gen_load_pair): Rename calls to match new patterns. 22565 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to... 22566 (load_pair<DREG:mode><DREG2:mode>): ... This. 22567 (store_pair<mode>): Rename to... 22568 (vec_store_pair<DREG:mode><DREG2:mode>): ... This. 22569 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX): 22570 New mode iterators. 22571 (V_INT_EQUIV): Handle SImode. 22572 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero): 22573 New predicate. 22574 225752018-05-22 Martin Sebor <msebor@redhat.com> 22576 22577 PR c/85623 22578 * calls.c (maybe_warn_nonstring_arg): Use string length to set 22579 or ajust the presumed bound on an operation to avoid unnecessary 22580 warnings. 22581 225822018-05-22 Martin Sebor <msebor@redhat.com> 22583 22584 PR tree-optimization/85826 22585 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid 22586 assuming that a DECL necesarily has a constant size. 22587 225882018-05-22 Richard Sandiford <richard.sandiford@linaro.org> 22589 22590 PR middle-end/85862 22591 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs. 22592 225932018-05-22 Richard Biener <rguenther@suse.de> 22594 22595 PR tree-optimization/85834 22596 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle 22597 non-constant and non-zero memset arguments. 22598 225992018-05-22 Martin Liska <mliska@suse.cz> 22600 22601 PR ipa/85607 22602 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types. 22603 226042018-05-22 Richard Biener <rguenther@suse.de> 22605 22606 PR tree-optimization/85863 22607 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant 22608 comparisons when vectype is specified. 22609 (vectorizable_condition): Do not specify vectype for 22610 vect_is_simple_cond when SLP vectorizing. 22611 226122018-05-21 Michael Meissner <meissner@linux.ibm.com> 22613 22614 PR target/85657 22615 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not 22616 define __ibm128 as long double. 22617 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128 22618 as a distinct type when IEEE 128-bit support is enabled. 22619 (init_float128_ieee): Fix up conversions between IFmode and IEEE 22620 128-bit types to use the correct functions. 22621 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to 22622 convert between 128-bit floating point types that have different 22623 modes but the same representation, instead of using gen_lowpart to 22624 makean alias. 22625 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and 22626 KFmode. 22627 (IFKF_reg): New attributes to give the register constraints for 22628 IFmode and KFmode. 22629 (extend<mode>tf2_internal): New insns to mark an explicit 22630 conversion between 128-bit floating point types that have a 22631 different mode but share the same representation. 22632 226332018-05-21 Richard Sandiford <richard.sandiford@linaro.org> 22634 22635 PR tree-optimization/85814 22636 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with 22637 a null return from get_strinfo when unsharing the next 22638 strinfo in the chain. 22639 226402018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com> 22641 22642 PR gcc/84923 22643 * varasm.c (weak_finish): Clean up weak_decls. 22644 226452018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 22646 22647 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL, 22648 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2, 22649 UNSPEC_UADALP values. 22650 * config/aarch64/iterators.md (ABAL): New int iterator. 22651 (ABDL2): Likewise. 22652 (ADALP): Likewise. 22653 (sur): Add mappings for the above. 22654 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3): 22655 New define_insn. 22656 (aarch64_<sur>abal<mode>_4): Likewise. 22657 (aarch64_<sur>adalp<mode>_3): Likewise. 22658 (<sur>sadv16qi): New define_expand. 22659 226602018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com> 22661 22662 * config/i386/i386.md (*movsf_internal): AVX falsedep fix. 22663 (*movdf_internal): Ditto. 22664 (*rcpsf2_sse): Ditto. 22665 (*rsqrtsf2_sse): Ditto. 22666 (*sqrt<mode>2_sse): Ditto. 22667 226682018-05-21 Tamar Christina <tamar.christina@arm.com> 22669 22670 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to 22671 eor3q<mode>4. 22672 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4. 22673 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32, 22674 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8, 22675 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32, 22676 vbcaxq_s64): New. 22677 * config/aarch64/arm_neon.h: Likewise. 22678 * config/aarch64/iterators.md (VQ_I): New. 22679 226802018-05-21 Alexey Brodkin <abrodkin@synopsys.com> 22681 22682 * config.gcc: Add arc/t-multilib-linux to tmake_file for 22683 arc*-*-linux*. 22684 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and 22685 MULTILIB_DIRNAMES 22686 226872018-05-20 Chung-Ju Wu <jasonwucj@gmail.com> 22688 22689 * config/nds32/constraints.md (S): New constraint. 22690 * config/nds32/nds32.md (call_internal): Use constraint S. 22691 (call_value_internal): Likewise. 22692 (sibcall_internal): Likewise. 22693 (sibcall_value_internal): Likewise. 22694 226952018-05-20 Kito Cheng <kito.cheng@gmail.com> 22696 Chung-Ju Wu <jasonwucj@gmail.com> 22697 22698 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu 22699 into consideration. 22700 227012018-05-20 Kito Cheng <kito.cheng@gmail.com> 22702 Chung-Ju Wu <jasonwucj@gmail.com> 22703 22704 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure. 22705 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation. 22706 (nds32_rtx_costs_impl): Simplify. 22707 (nds32_address_cost_impl): Simplify. 22708 (nds32_init_rtx_costs): New function. 22709 (nds32_rtx_costs_speed_prefer): Likewise. 22710 (nds32_rtx_costs_size_prefer): Likewise. 22711 (nds32_address_cost_speed_prefer): Likewise. 22712 (nds32_address_cost_speed_fwprop): Likewise. 22713 (nds32_address_cost_size_prefer): Likewise. 22714 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare. 22715 * config/nds32/nds32.c (nds32_option_override): Use 22716 nds32_init_rtx_costs function. 22717 227182018-05-20 Chung-Ju Wu <jasonwucj@gmail.com> 22719 22720 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model. 22721 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define. 22722 (TARGET_PIPELINE_N8): Likewise. 22723 (TARGET_PIPELINE_N10): Likewise. 22724 (TARGET_PIPELINE_N13): Likewise. 22725 (TARGET_PIPELINE_GRAYWOLF): Likewise. 22726 227272018-05-19 Monk Chiang <sh.chiang04@gmail.com> 22728 22729 * config/nds32/nds32-fpu.md: Update copyright year. 22730 227312018-05-19 Chung-Ju Wu <jasonwucj@gmail.com> 22732 22733 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule. 22734 227352018-05-19 Chung-Ju Wu <jasonwucj@gmail.com> 22736 22737 * config/nds32/nds32.c 22738 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable. 22739 * config/nds32/nds32.opt (minline-asm-r15): New option. 22740 227412018-05-19 Chung-Ju Wu <jasonwucj@gmail.com> 22742 22743 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add 22744 MASK_HW_ABS. 22745 * config/nds32/nds32.md (abssi2): New pattern. 22746 227472018-05-19 Uros Bizjak <ubizjak@gmail.com> 22748 22749 * config/i386/i386.md (rex64namesuffix): New mode attribute. 22750 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>): 22751 Merge insn pattern from sse_cvtsi2ss<round_name> and 22752 sse_cvtsi2ssq<round_name> using SWI48 mode iterator. 22753 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern 22754 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name> 22755 using SWI48 mode iterator. 22756 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from 22757 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator. 22758 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn 22759 pattern from sse_cvttss2si<round_saeonly_name> 22760 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator. 22761 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern 22762 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name> 22763 using SWI48 mode iterator. 22764 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge 22765 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and 22766 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator. 22767 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern 22768 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name> 22769 using SWI48 mode iterator. 22770 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge 22771 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and 22772 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator. 22773 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from 22774 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using 22775 SWI48 mode iterator. 22776 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from 22777 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator. 22778 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn 22779 pattern from sse_cvttsd2si<round_saeonly_name> 22780 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator. 22781 227822018-05-19 Chung-Ju Wu <jasonwucj@gmail.com> 22783 22784 * config/nds32/nds32-md-auxiliary.c 22785 (nds32_valid_smw_lwm_base_p): Refine. 22786 (nds32_output_smw_single_word): Refine. 22787 (nds32_output_smw_double_word): New. 22788 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New. 22789 227902018-05-19 Chung-Ju Wu <jasonwucj@gmail.com> 22791 22792 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine. 22793 (nds32_output_stack_pop): Refine. 22794 (nds32_expand_unaligned_load): Refine. 22795 (nds32_expand_unaligned_store): Refine. 22796 227972018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com> 22798 Chung-Ju Wu <jasonwucj@gmail.com> 22799 22800 * config/nds32/constants.md: Add TP_REGNUM constant. 22801 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF, 22802 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and 22803 UNSPEC_ADD32. 22804 * config/nds32/nds32-doubleword.md: Consider flag_pic. 22805 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases. 22806 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New. 22807 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS 22808 and PIC code generation. 22809 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC 22810 code generation. 22811 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax 22812 optimization. 22813 * config/nds32/nds32.md: Support TLS and PIC. 22814 * config/nds32/nds32.c: Support TLS and PIC. 22815 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type. 22816 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New 22817 predicate. 22818 228192018-05-19 Chung-Ju Wu <jasonwucj@gmail.com> 22820 22821 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine 22822 mode with E_ prefix. 22823 228242018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com> 22825 Chung-Ju Wu <jasonwucj@gmail.com> 22826 22827 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT. 22828 * config/nds32/nds32-md-auxiliary.c 22829 (symbolic_reference_mentioned_p): New. 22830 (nds32_legitimize_ict_address): New. 22831 (nds32_expand_ict_move): New. 22832 (nds32_indirect_call_referenced_p): New. 22833 (nds32_symbol_binds_local_p): Delete. 22834 (nds32_long_call_p): Modify. 22835 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type. 22836 * config/nds32/nds32-protos.h 22837 (symbolic_reference_mentioned_p): Declare. 22838 (nds32_legitimize_ict_address): Declare. 22839 (nds32_expand_ict_move): Declare. 22840 (nds32_indirect_call_referenced_p): Declare. 22841 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New. 22842 (nds32_relax_group): Use nds32_ict_const_p as condition. 22843 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call". 22844 (nds32_asm_file_start): Output ict_model directive in asm code. 22845 (nds32_legitimate_address_p): Consider indirect call. 22846 (nds32_print_operand): Consider indirect call. 22847 (nds32_print_operand_address): Consider indirect call. 22848 (nds32_insert_attributes): Handle "indirect_call" attribute. 22849 (TARGET_LEGITIMATE_ADDRESS_P): Define. 22850 (TARGET_LEGITIMATE_CONSTANT_P): Define. 22851 (TARGET_CANNOT_FORCE_CONST_MEM): Define. 22852 (TARGET_DELEGITIMIZE_ADDRESS): Define. 22853 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define. 22854 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define. 22855 (TARGET_ICT_MODEL_SMALL): Define. 22856 (TARGET_ICT_MODEL_LARGE): Define. 22857 * config/nds32/nds32.md (movsi): Consider ict model. 22858 (call, call_value): Consider ict model. 22859 (sibcall, sibcall_value): Consider ict model. 22860 * config/nds32/nds32.opt (mict-model): New option. 22861 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict 22862 model. 22863 228642018-05-18 Kito Cheng <kito.cheng@gmail.com> 22865 Monk Chiang <sh.chiang04@gmail.com> 22866 Jim Wilson <jimw@sifive.com> 22867 22868 * common/config/riscv/riscv-common.c (riscv_parse_arch_string): 22869 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i. 22870 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e. 22871 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define 22872 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI. 22873 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E. 22874 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE, 22875 compute save_libcall_adjustment properly. 22876 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E. 22877 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E. 22878 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E. 22879 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE. 22880 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise. 22881 (ABI_SPEC): Handle mabi=ilp32e. 22882 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E. 22883 (RVE): Add RVE mask. 22884 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info. 22885 <-march>: Add rv32e as an example. 22886 228872018-05-18 Marc Glisse <marc.glisse@inria.fr> 22888 22889 PR c++/82899 22890 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument. 22891 (intra_create_variable_infos): Handle C++ constructors. 22892 228932018-05-18 Martin Liska <mliska@suse.cz> 22894 22895 * passes.def: Remove a redundant pass. 22896 228972018-05-18 Eric Botcazou <ebotcazou@adacore.com> 22898 22899 PR bootstrap/85838 22900 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0]. 22901 229022018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 22903 22904 * config/arm/arm-cpus.in (armv3m, mode32): Delete features. 22905 (ARMv4): Update. 22906 (ARMv2, ARMv3, ARMv3m): Delete fgroups. 22907 (ARMv6m): Update. 22908 (armv2, armv2a, armv3, armv3m): Delete architectures. 22909 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620, 22910 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720, 22911 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi): 22912 Delete cpus. 22913 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m. 22914 (*mulsidi3adddi): Likewise. 22915 (mulsidi3): Likewise. 22916 (*mulsidi3_nov6): Likewise. 22917 (umulsidi3): Likewise. 22918 (umulsidi3_nov6): Likewise. 22919 (umaddsidi4): Likewise. 22920 (*umulsidi3adddi): Likewise. 22921 (smulsi3_highpart): Likewise. 22922 (*smulsi3_highpart_nov6): Likewise. 22923 (umulsi3_highpart): Likewise. 22924 (*umulsi3_highpart_nov6): Likewise. 22925 * config/arm/arm.h (arm_arch3m): Delete. 22926 * config/arm/arm.c (arm_arch3m): Delete. 22927 (arm_option_override_internal): Update armv3-related comment. 22928 (arm_configure_build_target): Delete use of isa_bit_mode32. 22929 (arm_option_reconfigure_globals): Delete set of arm_ach3m. 22930 (arm_rtx_costs_internal): Delete check of arm_arch3m. 22931 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m. 22932 (mulsa3): Likewise. 22933 (mulusa3): Likewise. 22934 * config/arm/arm-protos.h (arm_arch3m): Delete. 22935 * config/arm/arm-tables.opt: Regenerate. 22936 * config/arm/arm-tune.md: Likewise. 22937 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of 22938 deleted architectures. 22939 229402018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 22941 22942 * config/arm/arm-cpus.in (armv5, armv5e): Delete features. 22943 (armv5t, armv5te): New features. 22944 (ARMv5, ARMv5e): Delete fgroups. 22945 (ARMv5t, ARMv5te): Adjust for above changes. 22946 (ARMv6m): Likewise. 22947 (armv5, armv5e): Delete arches. 22948 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of 22949 arm_arch5. 22950 (*call_reg_arm): Likewise. 22951 (*call_value_reg_armv5): Likewise. 22952 (*call_value_reg_arm): Likewise. 22953 (*call_symbol): Likewise. 22954 (*call_value_symbol): Likewise. 22955 (*sibcall_insn): Likewise. 22956 (*sibcall_value_insn): Likewise. 22957 (clzsi2): Likewise. 22958 (prefetch): Likewise. 22959 (define_split and define_peephole2 dependent on arm_arch5): 22960 Likewise. 22961 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of 22962 arm_arch5e. 22963 (TARGET_ARM_QBIT): Likewise. 22964 (TARGET_DSP_MULTIPLY): Likewise. 22965 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E. 22966 (arm_arch5, arm_arch5e): Delete. 22967 (arm_arch5t, arm_arch5te): Declare. 22968 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete. 22969 (arm_arch5t): Declare. 22970 (arm_option_reconfigure_globals): Update for the above. 22971 (arm_options_perform_arch_sanity_checks): Update comment, replace 22972 use of arm_arch5 with arm_arch5t. 22973 (use_return_insn): Likewise. 22974 (arm_emit_call_insn): Likewise. 22975 (output_return_instruction): Likewise. 22976 (arm_final_prescan_insn): Likewise. 22977 (arm_coproc_builtin_available): Likewise. 22978 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and 22979 arm_arch5e with arm_arch5t and arm_arch5te. 22980 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete. 22981 (arm_arch5t, arm_arch5te): Declare. 22982 * config/arm/arm-tables.opt: Regenerate. 22983 * config/arm/t-arm-elf: Remove references to armv5, armv5e. 22984 * config/arm/t-multilib: Likewise. 22985 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t 22986 instead of arm_arch5. 22987 (*call_reg_thumb1): Likewise. 22988 (*call_value_reg_thumb1_v5): Likewise. 22989 (*call_value_reg_thumb1): Likewise. 22990 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now 22991 unreachable path. 22992 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e. 22993 229942018-05-18 Martin Liska <mliska@suse.cz> 22995 22996 PR gcov-profile/84846 22997 * doc/gcov.texi: Document -t option of gcov tool. 22998 229992018-05-18 Martin Liska <mliska@suse.cz> 23000 23001 PR gcov-profile/84846 23002 * gcov.c (print_usage): Add new -t option. 23003 (process_args): Handle the option. 23004 (generate_results): Use stdout as output when requested by 23005 the option. 23006 230072018-05-18 Martin Liska <mliska@suse.cz> 23008 23009 PR gcov-profile/84846 23010 * coverage.c (coverage_init): Write PWD to .gcno file. 23011 * doc/gcov.texi: Document how working directory is printed. 23012 * gcov-dump.c (dump_gcov_file): Print PWD. 23013 * gcov.c (output_intermediate_file): Likewise. 23014 (read_graph_file): Read PWD string. 23015 (output_lines): Print PWD. 23016 230172018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org> 23018 23019 PR middle-end/85817 23020 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop 23021 for retval and return false if all args to phi are zero. 23022 230232018-05-18 Richard Biener <rguenther@suse.de> 23024 23025 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges 23026 method. 23027 (evrp_dom_walker::before_dom_children): Call it. 23028 230292018-05-18 Richard Biener <rguenther@suse.de> 23030 23031 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine 23032 results when processing array refs with variable index. 23033 230342018-05-18 Toon Moene <toon@moene.org> 23035 23036 * doc/invoke.texi: Move -floop-unroll-and-jam documentation 23037 directly after that of -floop-interchange. Indicate that both 23038 options are enabled by default when specifying -O3. 23039 230402018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 23041 23042 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode 23043 iterator. Delete separate integer-mode vec_set<mode> expander. 23044 (aarch64_simd_vec_setv2di): Delete. 23045 (vec_setv2di): Delete. 23046 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name. 23047 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for 23048 the "w, r" alternative. 23049 230502018-05-18 Martin Liska <mliska@suse.cz> 23051 23052 * passes.def: Add pass_lower_switch and pass_lower_switch_O0. 23053 * tree-pass.h (make_pass_lower_switch_O0): New function. 23054 * tree-switch-conversion.c (node_has_low_bound): Remove. 23055 (node_has_high_bound): Likewise. 23056 (node_is_bounded): Likewise. 23057 (class pass_lower_switch): Make it a template type and create 23058 two instances. 23059 (pass_lower_switch::execute): Add template argument. 23060 (make_pass_lower_switch): New function. 23061 (make_pass_lower_switch_O0): New function. 23062 (do_jump_if_equal): Remove. 23063 (emit_case_nodes): Simplify to just handle all 3 cases and leave 23064 all the hard work to tree optimization passes. 23065 230662018-05-18 Martin Liska <mliska@suse.cz> 23067 23068 * dbgcnt.c (limit_low): Renamed from limit. 23069 (limit_high): New variable. 23070 (dbg_cnt_is_enabled): Check for upper limit. 23071 (dbg_cnt): Adjust dumping. 23072 (dbg_cnt_set_limit_by_index): Add new argument for high 23073 value. 23074 (dbg_cnt_set_limit_by_name): Likewise. 23075 (dbg_cnt_process_single_pair): Parse new format. 23076 (dbg_cnt_process_opt): Use strtok. 23077 (dbg_cnt_list_all_counters): Remove 'value' and add 23078 'limit_high'. 23079 * doc/invoke.texi: Document changes. 23080 230812018-05-18 Richard Sandiford <richard.sandiford@linaro.org> 23082 23083 * doc/sourcebuild.texi (scalar_all_fma): Document. 23084 * tree.def (FMA_EXPR): Delete. 23085 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions. 23086 * internal-fn.c (ternary_direct): New macro. 23087 (expand_ternary_optab_fn): Likewise. 23088 (direct_ternary_optab_supported_p): Likewise. 23089 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h. 23090 * builtins.c (fold_builtin_fma): Delete. 23091 (fold_builtin_3): Don't call it. 23092 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling. 23093 * expr.c (expand_expr_real_2): Likewise. 23094 * fold-const.c (operand_equal_p): Likewise. 23095 (fold_ternary_loc): Likewise. 23096 * gimple-pretty-print.c (dump_ternary_rhs): Likewise. 23097 * gimple.c (DEFTREECODE): Likewise. 23098 * gimplify.c (gimplify_expr): Likewise. 23099 * optabs-tree.c (optab_for_tree_code): Likewise. 23100 * tree-cfg.c (verify_gimple_assign_ternary): Likewise. 23101 * tree-eh.c (operation_could_trap_p): Likewise. 23102 (stmt_could_throw_1_p): Likewise. 23103 * tree-inline.c (estimate_operator_cost): Likewise. 23104 * tree-pretty-print.c (dump_generic_node): Likewise. 23105 (op_code_prio): Likewise. 23106 * tree-ssa-loop-im.c (stmt_cost): Likewise. 23107 * tree-ssa-operands.c (get_expr_operands): Likewise. 23108 * tree.c (commutative_ternary_tree_code, add_expr): Likewise. 23109 * fold-const-call.h (fold_fma): Delete. 23110 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS, 23111 CFN_FNMA and CFN_FNMS. 23112 (fold_fma): Delete. 23113 * genmatch.c (combined_fn): New enum. 23114 (commutative_ternary_tree_code): Remove FMA_EXPR handling. 23115 (commutative_op): New function. 23116 (commutate): Use it. Handle more than 2 operands. 23117 (dt_operand::gen_gimple_expr): Use commutative_op. 23118 (parser::parse_expr): Allow :c to be used with non-binary 23119 operators if the commutative operand is known. 23120 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle 23121 CFN_FMS, CFN_FNMA and CFN_FNMS. 23122 (backprop::process_assign_use): Remove FMA_EXPR handling. 23123 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise. 23124 (gen_hsa_fma): New function. 23125 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS, 23126 IFN_FNMA and IFN_FNMS. 23127 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS. 23128 * gimple-fold.h (follow_all_ssa_edges): Declare. 23129 * gimple-fold.c (follow_all_ssa_edges): New function. 23130 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the 23131 gimple_build interface and use follow_all_ssa_edges to fold the result. 23132 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p 23133 instead of checking for optabs directly. 23134 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls 23135 rather than FMA_EXPRs. 23136 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a 23137 call to IFN_FMA instead of an FMA_EXPR. 23138 231392018-05-17 Jim Wilson <jimw@sifive.com> 23140 23141 * expr.c (do_tablejump): When converting index to Pmode, if we have a 23142 sign extended promoted subreg, and the range does not have the sign bit 23143 set, then do a sign extend. 23144 23145 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode 23146 test, check for sign extended subreg and/or constant operands, and 23147 do a sign extend in that case. 23148 231492018-05-17 Steve Ellcey <sellcey@cavium.com> 23150 23151 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete. 23152 (thunderx2t99_multiple): Delete psuedo-units from used cpus. 23153 Add untyped. 23154 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg. 23155 Change logics_shift_reg to logics_shift_imm. 23156 (thunderx2t99_fp_loadpair_basic): Delete. 23157 (thunderx2t99_fp_storepair_basic): Delete. 23158 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types. 23159 (thunderx2t99_asimd_polynomial): Delete. 23160 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q 23161 and neon_fp_mul_d_scalar_q. 23162 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types. 23163 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q. 23164 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types. 23165 (thunderx2t99_asimd_lut): Add missing tbl types. 23166 (thunderx2t99_asimd_ext): Delete. 23167 (thunderx2t99_asimd_load1_1_mult): Delete. 23168 (thunderx2t99_asimd_load1_2_mult): Delete. 23169 (thunderx2t99_asimd_load1_ldp): New. 23170 (thunderx2t99_asimd_load1): New. 23171 (thunderx2t99_asimd_load2): Add missing *load2* types. 23172 (thunderx2t99_asimd_load3): New. 23173 (thunderx2t99_asimd_load4): New. 23174 (thunderx2t99_asimd_store1_1_mult): Delete. 23175 (thunderx2t99_asimd_store1_2_mult): Delete. 23176 (thunderx2t99_asimd_store2_mult): Delete. 23177 (thunderx2t99_asimd_store2_onelane): Delete. 23178 (thunderx2t99_asimd_store_stp): New. 23179 (thunderx2t99_asimd_store1): New. 23180 (thunderx2t99_asimd_store2): New. 23181 (thunderx2t99_asimd_store3): New. 23182 (thunderx2t99_asimd_store4): New. 23183 231842018-05-17 Jerome Lambourg <lambourg@adacore.com> 23185 23186 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove 23187 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__. 23188 231892018-05-17 Pat Haugen <pthaugen@us.ibm.com> 23190 Segher Boessenkool <segher@kernel.crashing.org> 23191 23192 PR target/85698 23193 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest 23194 operand. 23195 231962018-05-17 Richard Biener <rguenther@suse.de> 23197 23198 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment 23199 for pruning loop and prune defs feeding only already visited PHIs. 23200 232012018-05-17 Richard Biener <rguenther@suse.de> 23202 23203 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling. 23204 232052018-05-17 Bin Cheng <bin.cheng@arm.com> 23206 Richard Biener <rguenther@suse.de> 23207 23208 PR tree-optimization/85793 23209 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load 23210 for VMAT_ELEMENTWISE. 23211 232122018-05-17 Richard Sandiford <richard.sandiford@linaro.org> 23213 23214 * internal-fn.h (lookup_internal_fn): Declare 23215 * internal-fn.c (lookup_internal_fn): New function. 23216 * gimple.c (gimple_build_call_from_tree): Handle calls to 23217 internal functions. 23218 * gimple-pretty-print.c (dump_gimple_call): Print "." before 23219 internal function names. 23220 * tree-pretty-print.c (dump_generic_node): Likewise. 23221 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise. 23222 232232018-05-17 Richard Sandiford <richard.sandiford@linaro.org> 23224 23225 * gimple-fold.h (gimple_build): Make the function forms take 23226 combined_fn rather than built_in_function. 23227 (gimple_simplify): Likewise. 23228 * gimple-match-head.c (gimple_simplify): Likewise. 23229 * gimple-fold.c (gimple_build): Likewise. 23230 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build 23231 rather than gimple_build_call_internal. 23232 (get_initial_defs_for_reduction): Likewise. 23233 (vect_create_epilog_for_reduction): Likewise. 23234 (vectorizable_live_operation): Likewise. 23235 232362018-05-17 Martin Liska <mliska@suse.cz> 23237 23238 * gimple-ssa-sprintf.c (format_directive): Do not use 23239 space in between 'G_' and '('. 23240 232412018-05-17 Jakub Jelinek <jakub@redhat.com> 23242 23243 PR target/85323 23244 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts 23245 even if the mask is not all ones. 23246 23247 PR target/85323 23248 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by 23249 vector. 23250 (ix86_gimple_fold_builtin): Likewise. 23251 23252 PR target/85323 23253 * config/i386/i386.c: Include tree-vector-builder.h. 23254 (ix86_vector_shift_count): New function. 23255 (ix86_fold_builtin): Fold shift builtins by scalar count. 23256 (ix86_gimple_fold_builtin): Likewise. 23257 23258 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8, 23259 _mm512_setzero): New intrinsics. 23260 232612018-05-17 James Greenhalgh <james.greenhalgh@arm.com> 23262 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 23263 23264 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify 23265 code generation for cases where splatting a value is not useful. 23266 * simplify-rtx.c (simplify_ternary_operation): Simplify 23267 vec_merge across a vec_duplicate and a paradoxical subreg forming 23268 a vector mode to a vec_concat. 23269 232702018-05-17 Olga Makhotina <olga.makhotina@intel.com> 23271 23272 * config.gcc: Support "goldmont-plus". 23273 * config/i386/driver-i386.c (host_detect_local_cpu): Detect 23274 "goldmont-plus". 23275 * config/i386/i386-c.c (ix86_target_macros_internal): Handle 23276 PROCESSOR_GOLDMONT_PLUS. 23277 * config/i386/i386.c (m_GOLDMONT_PLUS): Define. 23278 (processor_target_table): Add "goldmont-plus". 23279 (PTA_GOLDMONT_PLUS): Define. 23280 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS. 23281 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS. 23282 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS. 23283 (fold_builtin_cpu): Add "goldmont-plus". 23284 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS. 23285 (ix86_option_override_internal): Add "goldmont-plus". 23286 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS. 23287 (processor_type): Add PROCESSOR_GOLDMONT_PLUS. 23288 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS. 23289 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type. 23290 232912018-05-17 Richard Biener <rguenther@suse.de> 23292 23293 PR tree-optimization/85757 23294 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and 23295 remove defs that only feed that PHI from further processing. 23296 232972018-05-16 Jim Wilson <jimw@sifive.com> 23298 23299 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend 23300 asterisk to name. 23301 (<optab>di3_mask, <optab>di3_mask_1): Likewise. 23302 233032018-05-16 Mark Wielaard <mark@klomp.org> 23304 23305 * dwarf2out.c (count_index_strings): New function. 23306 (output_indirect_strings): Call count_index_strings and generate 23307 header for dwarf_version >= 5. 23308 233092018-05-16 Mark Wielaard <mark@klomp.org> 23310 23311 * dwarf2out.c (dwarf_FORM): New function. 23312 (set_indirect_string): Use dwarf_FORM. 23313 (reset_indirect_string): Likewise. 23314 (size_of_die): Likewise. 23315 (value_format): Likewise. 23316 (output_die): Likewise. 23317 (add_skeleton_AT_string): Likewise. 23318 (output_macinfo_op): Likewise. 23319 (index_string): Likewise. 23320 (output_index_string_offset): Likewise. 23321 (output_index_string): Likewise. 23322 (count_index_strings): Likewise. 23323 233242018-05-16 Carl Love <cel@us.ibm.com> 23325 23326 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions 23327 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer. 23328 233292018-05-16 Martin Jambor <mjambor@suse.cz> 23330 23331 * ipa-prop.c (ipa_free_all_edge_args): Remove. 23332 * ipa-prop.h (ipa_free_all_edge_args): Likewise. 23333 233342018-05-16 Wilco Dijkstra <wdijkstr@arm.com> 23335 23336 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern. 23337 (fnma<mode>4): Likewise. 23338 (fms<mode>4): Likewise. 23339 (fnms<mode>4): Likewise. 23340 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand. 23341 (aarch64_fnma<mode>4): Likewise. 23342 (aarch64_fms<mode>4): Likewise. 23343 (aarch64_fnms<mode>4): Likewise. 23344 (aarch64_fnmadd<mode>4): Likewise. 23345 233462018-05-16 Jason Merrill <jason@redhat.com> 23347 23348 * tree.c (warn_deprecated_use): Return bool. Simplify logic. 23349 233502018-05-16 Richard Biener <rguenther@suse.de> 23351 23352 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member. 23353 (dump_stmt_cost): Declare. 23354 (add_stmt_cost): Dump cost we add. 23355 (add_stmt_costs): New function. 23356 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost): 23357 No longer exported. 23358 (vect_analyze_stmt): Adjust prototype. 23359 (vectorizable_condition): Likewise. 23360 (vectorizable_live_operation): Likewise. 23361 (vectorizable_reduction): Likewise. 23362 (vectorizable_induction): Likewise. 23363 * tree-vect-loop.c (vect_analyze_loop_operations): Create local 23364 cost vector to pass to vectorizable_ and record afterwards. 23365 (vect_model_reduction_cost): Take cost vector argument and adjust. 23366 (vect_model_induction_cost): Likewise. 23367 (vectorizable_reduction): Likewise. 23368 (vectorizable_induction): Likewise. 23369 (vectorizable_live_operation): Likewise. 23370 * tree-vect-slp.c (vect_create_new_slp_node): Initialize 23371 SLP_TREE_NUMBER_OF_VEC_STMTS. 23372 (vect_analyze_slp_cost_1): Remove. 23373 (vect_analyze_slp_cost): Likewise. 23374 (vect_slp_analyze_node_operations): Take visited args and 23375 a target cost vector. Avoid processing already visited stmt sets. 23376 (vect_slp_analyze_operations): Use a local cost vector to gather 23377 costs and register those of non-discarded instances. 23378 (vect_bb_vectorization_profitable_p): Use add_stmt_costs. 23379 (vect_schedule_slp_instance): Remove copying of 23380 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not 23381 zero. 23382 * tree-vect-stmts.c (record_stmt_cost): Remove path directly 23383 adding cost. Record cost entry location. 23384 (vect_prologue_cost_for_slp_op): Function to compute cost of 23385 a constant or invariant generated for SLP vect in the prologue, 23386 split out from vect_analyze_slp_cost_1. 23387 (vect_model_simple_cost): Make static. Adjust for SLP costing. 23388 (vect_model_promotion_demotion_cost): Likewise. 23389 (vect_model_store_cost): Likewise, make static. 23390 (vect_model_load_cost): Likewise. 23391 (vectorizable_bswap): Add cost vector arg and adjust. 23392 (vectorizable_call): Likewise. 23393 (vectorizable_simd_clone_call): Likewise. 23394 (vectorizable_conversion): Likewise. 23395 (vectorizable_assignment): Likewise. 23396 (vectorizable_shift): Likewise. 23397 (vectorizable_operation): Likewise. 23398 (vectorizable_store): Likewise. 23399 (vectorizable_load): Likewise. 23400 (vectorizable_condition): Likewise. 23401 (vectorizable_comparison): Likewise. 23402 (can_vectorize_live_stmts): Likewise. 23403 (vect_analyze_stmt): Likewise. 23404 (vect_transform_stmt): Adjust calls to vectorizable_*. 23405 * tree-vectorizer.c: Include gimple-pretty-print.h. 23406 (dump_stmt_cost): New function. 23407 234082018-05-16 Richard Biener <rguenther@suse.de> 23409 23410 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param. 23411 * doc/invoke.texi (dse-max-alias-queries-per-store): Document. 23412 * tree-ssa-dse.c: Include tree-ssa-loop.h. 23413 (check_name): New callback. 23414 (dse_classify_store): Track cycles via a visited bitmap of PHI 23415 defs and simplify handling of in-loop and across loop dead stores 23416 and properly fail for loop-variant refs. Handle byte-tracking with 23417 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for 23418 limiting the walk. 23419 234202018-05-16 Richard Sandiford <richard.sandiford@linaro.org> 23421 23422 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare. 23423 (vect_get_mask_type_for_stmt): Likewise. 23424 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function, 23425 split out from... 23426 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt 23427 to determine the statement's vector type and the vector type that 23428 should be used for calculating nunits. Deal with cases in which 23429 the type has to be deferred. 23430 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt 23431 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE. 23432 * tree-vect-loop.c (vect_determine_vf_for_stmt_1) 23433 (vect_determine_vf_for_stmt): New functions, split out from... 23434 (vect_determine_vectorization_factor): ...here. 23435 * tree-vect-stmts.c (vect_get_vector_types_for_stmt) 23436 (vect_get_mask_type_for_stmt): New functions, split out from 23437 vect_determine_vectorization_factor. 23438 234392018-05-16 Richard Biener <rguenther@suse.de> 23440 23441 * tree-cfg.c (verify_gimple_assign_ternary): Properly 23442 verify the [VEC_]COND_EXPR embedded comparison. 23443 234442018-05-15 Martin Sebor <msebor@redhat.com> 23445 23446 PR tree-optimization/85753 23447 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle 23448 RECORD_TYPE in addition to ARRAY_TYPE. 23449 234502018-05-15 Martin Sebor <msebor@redhat.com> 23451 23452 PR middle-end/85643 23453 * calls.c (get_attr_nonstring_decl): Handle MEM_REF. 23454 234552018-05-15 Richard Biener <rguenther@suse.de> 23456 23457 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter, 23458 add by_clobber_p one. Change algorithm to collect all defs 23459 representing uses we need to walk and try reducing them to 23460 a single one before failing. 23461 (dse_dom_walker::dse_optimize_stmt): Adjust. 23462 234632018-05-13 Mark Wielaard <mark@klomp.org> 23464 23465 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx. 23466 (size_of_loc_descr): Likewise. 23467 (output_loc_operands): Likewise. 23468 (output_loc_operands_raw): Likewise. 23469 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx. 23470 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx. 23471 (hash_loc_operands): Likewise. 23472 (compare_loc_operands): Likewise. 23473 234742018-05-14 Mark Wielaard <mark@klomp.org> 23475 23476 * dwarf2out.c (count_index_addrs): New function. 23477 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length. 23478 234792018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org> 23480 23481 PR tree-optimization/83648 23482 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL 23483 return value as malloc candidate. 23484 234852018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org> 23486 23487 PR ipa/85734 23488 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite 23489 param as true in call to suggest_attribute. 23490 234912018-05-14 Segher Boessenkool <segher@kernel.crashing.org> 23492 23493 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for 23494 -mreadonly-in-sdata. 23495 234962018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 23497 23498 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused): 23499 New pattern. 23500 (aarch64_crypto_aesd_fused): Likewise. 23501 235022018-05-14 Wilco Dijkstra <wdijkstr@arm.com> 23503 23504 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives. 23505 (movsi_aarch64): Likewise. 23506 (load_pairsi): Likewise. 23507 (load_pairdi): Likewise. 23508 (store_pairsi): Likewise. 23509 (store_pairdi): Likewise. 23510 (load_pairsf): Likewise. 23511 (load_pairdf): Likewise. 23512 (store_pairsf): Likewise. 23513 (store_pairdf): Likewise. 23514 (zero_extend): Likewise. 23515 (trunc): Swap alternatives. 23516 (fcvt_target): Add '?' to prefer w over r. 23517 235182018-05-14 Jakub Jelinek <jakub@redhat.com> 23519 23520 PR target/85756 23521 * config/i386/i386.md: Disallow non-commutative arithmetics in 23522 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim 23523 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS 23524 in the peephole2 before it. 23525 235262018-05-14 Sebastian Peryt <sebastian.peryt@intel.com> 23527 23528 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET, 23529 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines. 23530 (ix86_handle_option): Handle -mcldemote. 23531 * config.gcc: New header. 23532 * config/i386/cldemoteintrin.h: New file. 23533 * config/i386/cpuid.h (bit_CLDEMOTE): New bit. 23534 * config/i386/driver-i386.c (host_detect_local_cpu): Detect 23535 -mcldemote. 23536 * config/i386/i386-c.c (ix86_target_macros_internal): Handle 23537 OPTION_MASK_ISA_CLDEMOTE. 23538 * config/i386/i386.c (ix86_target_string): Add -mcldemote. 23539 (ix86_valid_target_attribute_inner_p): Ditto. 23540 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE. 23541 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote. 23542 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE. 23543 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New. 23544 * config/i386/i386.md (UNSPECV_CLDEMOTE): New. 23545 (cldemote): New. 23546 * config/i386/i386.opt: Add -mcldemote. 23547 * config/i386/x86intrin.h: New header. 23548 * doc/invoke.texi: Add -mcldemote. 23549 235502018-05-14 Richard Biener <rguenther@suse.de> 23551 23552 * doc/match-and-simplify.texi: Adjust :s documentation. 23553 235542018-05-14 Alexander Monakov <amonakov@ispras.ru> 23555 23556 * sort.cc (REORDER_23): Pass the type for the temporaries instead of 23557 intended memcpy size. 23558 (REORDER_45): Likewise. 23559 235602018-05-13 Alexander Monakov <amonakov@ispras.ru> 23561 23562 * sort.cc: New file. 23563 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort. 23564 * vec.c (qsort_chk): Use gcc_qsort. 23565 * Makefile.in (OBJS-libcommon): Add sort.o. 23566 (build/sort.o): New target. Use it... 23567 (BUILD_RTL): ... here, and... 23568 (build/gencfn-macros): ... here, and... 23569 (build/genmatch): ... here. 23570 235712018-05-13 Kito Cheng <kito.cheng@gmail.com> 23572 Chung-Ju Wu <jasonwucj@gmail.com> 23573 23574 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu. 23575 * config/nds32/nds32-graywolf.md: New file. 23576 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF. 23577 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15 23578 pipeline. 23579 * config/nds32/nds32-protos.h: More declarations for n15 pipeline. 23580 * config/nds32/nds32-utils.c: More implementations for n15 pipeline. 23581 * config/nds32/nds32.md (pipeline_model): Add graywolf. 23582 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus. 23583 * config/nds32/pipelines.md: Include n15 settings. 23584 235852018-05-13 Kito Cheng <kito.cheng@gmail.com> 23586 Chung-Ju Wu <jasonwucj@gmail.com> 23587 23588 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu. 23589 * config/nds32/nds32-n13.md: New file. 23590 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13. 23591 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13 23592 pipeline. 23593 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline. 23594 * config/nds32/nds32.md (pipeline_model): Add n13. 23595 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus. 23596 * config/nds32/pipelines.md: Include n13 settings. 23597 235982018-05-13 Kito Cheng <kito.cheng@gmail.com> 23599 Chung-Ju Wu <jasonwucj@gmail.com> 23600 23601 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu. 23602 * config/nds32/nds32-n10.md: New file. 23603 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10. 23604 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10 23605 pipeline. 23606 * config/nds32/nds32-protos.h: More declarations for n10 pipeline. 23607 * config/nds32/nds32-utils.c: More implementations for n10 pipeline. 23608 * config/nds32/nds32.md (pipeline_model): Add n10. 23609 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus. 23610 * config/nds32/pipelines.md: Include n10 settings. 23611 236122018-05-13 Monk Chiang <sh.chiang04@gmail.com> 23613 Kito Cheng <kito.cheng@gmail.com> 23614 Chung-Ju Wu <jasonwucj@gmail.com> 23615 23616 * config.gcc (nds32be-*-*): Handle --with-ext-dsp. 23617 * config/nds32/constants.md (unspec_element, unspec_volatile_element): 23618 Add enum values for DSP extension instructions. 23619 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi): 23620 New constraints. 23621 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus, 23622 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max): 23623 New code iterators. 23624 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes. 23625 * config/nds32/nds32-dspext.md: New file for DSP implementation. 23626 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension. 23627 * config/nds32/nds32-intrinsic.md: Likewise. 23628 * config/nds32/nds32_intrinsic.h: Likewise. 23629 * config/nds32/nds32-md-auxiliary.c: Likewise. 23630 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension. 23631 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New. 23632 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New. 23633 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New. 23634 * config/nds32/nds32-protos.h: New declarations for DSP extension. 23635 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case 23636 TYPE_DMAC in switch statement. 23637 * config/nds32/nds32.c: New checking and implementation for DSP 23638 extension instructions. 23639 * config/nds32/nds32.h: Likewise. 23640 * config/nds32/nds32.md: Likewise. 23641 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options. 23642 * config/nds32/predicates.md: Implement new predicates for DSP 23643 extension. 23644 236452018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com> 23646 23647 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32): 23648 Reformat alternatives and attributes so it is easier to identify 23649 which constraints/attributes go with which instruction. 23650 (mov<mode>_hardfloat32, FMOVE64): Likewise. 23651 (mov<mode>_softfloat32, FMOVE64): Likewise. 23652 (mov<mode>_hardfloat64, FMOVE64): Likewise. 23653 (mov<mode>_softfloat64, FMOVE64): Likewise. 23654 236552018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org> 23656 23657 * doc/extend.texi (PowerPC Built-in Functions): Rename this 23658 subsection. 23659 (Basic PowerPC Built-in Functions): The new name of the 23660 subsection previously known as "PowerPC Built-in Functions". 23661 (Basic PowerPC Built-in Functions Available on all Configurations): 23662 New subsubsection. 23663 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise. 23664 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise. 23665 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise. 23666 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise. 23667 236682018-05-11 Martin Jambor <mjambor@suse.cz> 23669 23670 PR ipa/85655 23671 * ipa-cp.c (intersect_with_plats): Check that the lattice contains 23672 single const. 23673 236742018-05-11 Richard Earnshaw <rearnsha@arm.com> 23675 23676 PR target/85733 23677 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature. 23678 236792018-05-11 Sebastian Peryt <sebastian.peryt@intel.com> 23680 23681 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET, 23682 OPTION_MASK_ISA_WAITPKG_UNSET): New defines. 23683 (ix86_handle_option): Handle -mwaitpkg. 23684 * config.gcc: New header. 23685 * config/i386/cpuid.h (bit_WAITPKG): New bit. 23686 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg. 23687 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New 23688 function type. 23689 * config/i386/i386-c.c (ix86_target_macros_internal): Handle 23690 OPTION_MASK_ISA_WAITPKG. 23691 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg. 23692 (ix86_option_override_internal): Add PTA_WAITPKG. 23693 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg. 23694 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT, 23695 IX86_BUILTIN_TPAUSE. 23696 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor, 23697 __builtin_ia32_umwait and __builtin_ia32_tpause. 23698 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR, 23699 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE. 23700 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New. 23701 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR, 23702 UNSPECV_TPAUSE): New. 23703 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New. 23704 * config/i386/i386.opt: Add -mwaitpkg. 23705 * config/i386/waitpkgintrin.h: New file. 23706 * config/i386/x86intrin.h: New header. 23707 * doc/invoke.texi: Add -mwaitpkg. 23708 237092018-05-11 Richard Earnshaw <rearnsha@arm.com> 23710 23711 PR target/85606 23712 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now 23713 equivalent. 23714 (cortex-m0): Use armv6s-m isa. 23715 (cortex-m0plus): Likewise. 23716 (cortex-m1): Likewise. 23717 (cortex-m0.small-multiply): Likewise. 23718 (cortex-m0plus.small-multiply): Likewise. 23719 (cortex-m1.small-multiply): Likewise. 23720 237212018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io> 23722 Jakub Jelinek <jakub@redhat.com> 23723 23724 PR tree-optimization/85692 23725 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two 23726 source permute as well. 23727 237282018-05-11 Martin Liska <mliska@suse.cz> 23729 23730 PR sanitizer/85556 23731 * doc/extend.texi: Document LLVM style format for no_sanitize 23732 attribute. 23733 237342018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com> 23735 23736 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename 23737 mode_supports_vsx_dform_quad to mode_supports_dq_form. 23738 (mode_supports_vsx_dform_quad): Likewise. 23739 (mode_supports_vmx_dform): Move these functions to be next to the 23740 other mode_supports functions. 23741 (mode_supports_dq_form): Likewise. 23742 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to 23743 mode_supports_dq_form. 23744 (reg_offset_addressing_ok_p): Likewise. 23745 (offsettable_ok_by_alignment): Likewise. 23746 (rs6000_legitimate_offset_address_p): Likewise. 23747 (legitimate_lo_sum_address_p): Likewise. 23748 (rs6000_legitimize_address): Likewise. 23749 (rs6000_legitimize_reload_address): Likewise. 23750 (rs6000_secondary_reload_inner): Likewise. 23751 (rs6000_preferred_reload_class): Likewise. 23752 (rs6000_output_move_128bit): Likewise. 23753 237542018-05-10 Uros Bizjak <ubizjak@gmail.com> 23755 23756 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>: 23757 Generate SImode target register for null target. 23758 <case IX86_BUILTIN_XGETBV>: Ditto. 23759 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation. 23760 * config/i386/xsaveintrin.h (_xgetbv): Add missing return. 23761 237622018-05-10 Carl Love <cel@us.ibm.com> 23763 23764 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions 23765 dcbtt and dcbtstt if operands[2] is 0. 23766 237672018-05-10 Uros Bizjak <ubizjak@gmail.com> 23768 23769 PR target/85693 23770 * config/i386/sse.md (usadv64qi): New expander. 23771 237722018-05-10 Segher Boessenkool <segher@kernel.crashing.org> 23773 23774 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh, 23775 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove 23776 -maltivec=be support. 23777 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi, 23778 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi, 23779 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si, 23780 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi, 23781 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi, 23782 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx, 23783 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us, 23784 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws, 23785 altivec_vsumsws): Adjust. 23786 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth, 23787 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal, 23788 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be 23789 support. 23790 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns, 23791 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx, 23792 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust. 23793 (altivec_lve<VI_char>x): Delete expand. 23794 (*altivec_lve<VI_char>x_internal): Rename to... 23795 (altivec_lve<VI_char>x): ... this. 23796 (altivec_lvxl_<mode>): Delete expand. 23797 (*altivec_lvxl_<mode>_internal): Rename to ... 23798 (altivec_lvxl_<mode>): ... this. 23799 (altivec_stvxl_<mode>): Delete expand. 23800 (*altivec_stvxl_<mode>_internal): Rename to ... 23801 (altivec_stvxl_<mode>): ... this. 23802 (altivec_stve<VI_char>x): Delete expand. 23803 (*altivec_stve<VI_char>x_internal): Rename to ... 23804 (altivec_stve<VI_char>x): ... this. 23805 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2, 23806 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2, 23807 reduc_plus_scal_<mode>): Adjust. 23808 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust 23809 comment. 23810 (rs6000_cpu_cpp_builtins): Adjust. 23811 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support. 23812 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be, 23813 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete. 23814 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove 23815 -maltivec=be support. 23816 (rs6000_split_vec_extract_var): Adjust. 23817 (rs6000_split_v4si_init): Adjust. 23818 (swap_selector_for_mode): Delete. 23819 (altivec_expand_lvx_be, altivec_expand_stvx_be, 23820 altivec_expand_stvex_be): Delete. 23821 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove 23822 -maltivec=be support. 23823 (rs6000_gimple_fold_builtin): Ditto. 23824 (rs6000_generate_float2_double_code, rs6000_generate_float2_code): 23825 Adjust. 23826 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete. 23827 (TARGET_DIRECT_MOVE_64BIT): Adjust. 23828 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust. 23829 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete. 23830 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>, 23831 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df, 23832 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si, 23833 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df, 23834 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9, 23835 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an 23836 anonymous split): Adjust. 23837 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support. 23838 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust. 23839 238402018-05-10 Eric Botcazou <ebotcazou@adacore.com> 23841 23842 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only 23843 when --with-gxx-include-dir is also specified. 23844 * configure: Regenerate. 23845 238462018-05-09 Jim Wilson <jimw@sifive.com> 23847 23848 PR target/84797 23849 * config.gcc (riscv*-*-*): Handle --with-multilib-list. 23850 * config/riscv/t-withmultilib: New. 23851 * config/riscv/withmultilib.h: New. 23852 * doc/install.texi: Document RISC-V --with-multilib-list support. 23853 238542018-05-09 Richard Biener <rguenther@suse.de> 23855 23856 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost 23857 vector. 23858 (vect_bb_vectorization_profitable_p): Adjust. Compute 23859 actual scalar cost using the cost vector and the add_stmt_cost 23860 machinery. 23861 238622018-05-09 Segher Boessenkool <segher@kernel.crashing.org> 23863 23864 PR rtl-optimization/85645 23865 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET 23866 in the REG_CFA_REGISTER note for LR, don't leave it empty. 23867 238682018-05-09 Segher Boessenkool <segher@kernel.crashing.org> 23869 23870 PR rtl-optimization/85645 23871 * shrink-wrap.c (spread_components): Return a boolean saying if 23872 anything was changed. 23873 (try_shrink_wrapping_separate): Iterate spread_components until 23874 nothing changes anymore. 23875 238762018-05-09 Segher Boessenkool <segher@kernel.crashing.org> 23877 23878 PR rtl-optimization/85645 23879 * regrename.c (build_def_use): Also kill the chains that include the 23880 destination of a REG_CFA_REGISTER note. 23881 238822018-05-09 Segher Boessenkool <segher@kernel.crashing.org> 23883 23884 PR rtl-optimization/85645 23885 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an 23886 insn that has a REG_CFA_REGISTER note. 23887 238882018-05-09 Richard Sandiford <richard.sandiford@linaro.org> 23889 23890 * cfgexpand.c (expand_clobber): New function. 23891 (expand_gimple_stmt_1): Use it. 23892 * tree-vect-stmts.c (vect_clobber_variable): New function, 23893 split out from... 23894 (vectorizable_simd_clone_call): ...here. 23895 (vectorizable_store): Emit a clobber either side of an 23896 IFN_STORE_LANES sequence. 23897 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence. 23898 238992018-05-09 Tom de Vries <tom@codesourcery.com> 23900 23901 PR target/85626 23902 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true") 23903 (define_insn "trap_if_false"): Add exit after trap. 23904 239052018-05-09 Eric Botcazou <ebotcazou@adacore.com> 23906 23907 PR rtl-optimization/85638 23908 * bb-reorder.c: Include common/common-target.h. 23909 (create_forwarder_block): New function extracted from... 23910 (fix_up_crossing_landing_pad): ...here. Rename into... 23911 (dw2_fix_up_crossing_landing_pad): ...this. 23912 (sjlj_fix_up_crossing_landing_pad): New function. 23913 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode, 23914 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges 23915 from both partitions and exit the loop after one iteration. 23916 239172018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org> 23918 23919 Revert: 23920 * doc/extend.texi (PowerPC Built-in Functions): Rename this 23921 subsection. 23922 (Basic PowerPC Built-in Functions): The new name of the 23923 subsection previously known as "PowerPC Built-in Functions". 23924 (Basic PowerPC Built-in Functions Available on all Configurations): 23925 New subsubsection. 23926 (Basic PowerPC Built-in Functions Available on ISA 2.05): New 23927 subsubsection. 23928 (Basic PowerPC Built-in Functions Available on ISA 2.06): New 23929 subsubsection. 23930 (Basic PowerPC Built-in Functions Available on ISA 2.07): New 23931 subsubsection. 23932 (Basic PowerPC Built-in Functions Available on ISA 3.0): New 23933 subsubsection. 23934 239352018-05-08 Jim Wilson <jimw@sifive.com> 23936 23937 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash. 23938 (LD_EMUL_SUFFIX): New. 23939 (LINK_SPEC): Use it. 23940 239412018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org> 23942 23943 * doc/extend.texi (PowerPC Built-in Functions): Rename this 23944 subsection. 23945 (Basic PowerPC Built-in Functions): The new name of the 23946 subsection previously known as "PowerPC Built-in Functions". 23947 (Basic PowerPC Built-in Functions Available on all Configurations): 23948 New subsubsection. 23949 (Basic PowerPC Built-in Functions Available on ISA 2.05): New 23950 subsubsection. 23951 (Basic PowerPC Built-in Functions Available on ISA 2.06): New 23952 subsubsection. 23953 (Basic PowerPC Built-in Functions Available on ISA 2.07): New 23954 subsubsection. 23955 (Basic PowerPC Built-in Functions Available on ISA 3.0): New 23956 subsubsection. 23957 239582018-05-08 Jakub Jelinek <jakub@redhat.com> 23959 23960 PR target/85683 23961 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0 23962 after cmpelim optimization. 23963 239642018-05-08 Olga Makhotina <olga.makhotina@intel.com> 23965 23966 * config.gcc: Support "goldmont". 23967 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont". 23968 * config/i386/i386-c.c (ix86_target_macros_internal): Handle 23969 PROCESSOR_GOLDMONT. 23970 * config/i386/i386.c (m_GOLDMONT): Define. 23971 (processor_target_table): Add "goldmont". 23972 (PTA_GOLDMONT): Define. 23973 (ix86_lea_outperforms): Add TARGET_GOLDMONT. 23974 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT. 23975 (fold_builtin_cpu): Add M_INTEL_GOLDMONT. 23976 (fold_builtin_cpu): Add "goldmont". 23977 (ix86_add_stmt_cost): Add TARGET_GOLDMONT. 23978 (ix86_option_override_internal): Add "goldmont". 23979 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT. 23980 (processor_type): Add PROCESSOR_GOLDMONT. 23981 * config/i386/i386.md: Add CPU "glm". 23982 * config/i386/glm.md: New file. 23983 * config/i386/x86-tune.def: Add m_GOLDMONT. 23984 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type. 23985 239862018-05-08 Jakub Jelinek <jakub@redhat.com> 23987 23988 PR target/85572 23989 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and 23990 E_V4DImode. 23991 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of 23992 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not 23993 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes. 23994 23995 PR target/85317 23996 * config/i386/i386.c (ix86_fold_builtin): Handle 23997 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}. 23998 23999 PR target/85480 24000 * config/i386/sse.md (ssequaterinsnmode): New mode attribute. 24001 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern. 24002 240032018-05-08 Richard Earnshaw <rearnsha@arm.com> 24004 24005 PR target/85658 24006 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence. 24007 (check_arch): Likewise. 24008 (check_fpu): Return the result rather than printing it. 24009 (end arch): Fix operator precedence. 24010 (end cpu): Likewise. 24011 (END): Print the result from check_fpu. 24012 240132018-05-08 Richard Sandiford <richard.sandiford@linaro.org> 24014 Alan Hayward <alan.hayward@arm.com> 24015 David Sherwood <david.sherwood@arm.com> 24016 24017 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine) 24018 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine) 24019 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and) 24020 (*fcmuo<mode>_and): New patterns. 24021 240222018-05-08 Richard Sandiford <richard.sandiford@linaro.org> 24023 24024 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS) 24025 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete. 24026 (SVE_INT_CMP, SVE_FP_CMP): New code iterators. 24027 (cmp_op, sve_imm_con): New code attributes. 24028 (SVE_COND_INT_CMP, imm_con): Delete. 24029 (cmp_op): Remove above unspecs from int attribute. 24030 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename 24031 to... 24032 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of 24033 comparison-specific unspecs. 24034 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to... 24035 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise. 24036 (*vec_cmp<cmp_op>_<mode>_cc): Rename to... 24037 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise. 24038 (*vec_fcm<cmp_op><mode>): Rename to... 24039 (*fcm<cmp_op><mode>): ...this and adjust likewise. 24040 (*vec_fcmuo<mode>): Rename to... 24041 (*fcmuo<mode>): ...this and adjust likewise. 24042 (*pred_fcm<cmp_op><mode>): New pattern. 24043 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop) 24044 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New 24045 functions. 24046 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU 24047 and UNORDERED. 24048 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete. 24049 (aarch64_emit_sve_predicated_cond): New function. 24050 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc. 24051 (aarch64_emit_unspec_cond_or): Replace with... 24052 (aarch64_emit_sve_or_conds): ...this new function. Use 24053 aarch64_emit_sve_ptrue_op for the individual comparisons and 24054 aarch64_emit_binop to OR them together. 24055 (aarch64_emit_inverted_unspec_cond): Replace with... 24056 (aarch64_emit_sve_inverted_cond): ...this new function. Use 24057 aarch64_emit_sve_ptrue_op for the comparison and 24058 aarch64_emit_unop to invert the result. 24059 (aarch64_expand_sve_vec_cmp_float): Update after the above 24060 changes. Use aarch64_emit_sve_ptrue_op for native comparisons. 24061 240622018-05-07 Nathan Sidwell <nathan@acm.org> 24063 24064 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope. 24065 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope 24066 (Backwards Compatibility): Likewise. 24067 240682018-05-07 Luis Machado <luis.machado@linaro.org> 24069 24070 PR bootstrap/85681 24071 Revert: 24072 2018-05-07 Luis Machado <luis.machado@linaro.org> 24073 24074 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune) 24075 <prefetch_dynamic_strides>: New const bool field. 24076 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include 24077 prefetch_dynamic_strides. 24078 (exynosm1_prefetch_tune): Likewise. 24079 (thunderxt88_prefetch_tune): Likewise. 24080 (thunderx_prefetch_tune): Likewise. 24081 (thunderx2t99_prefetch_tune): Likewise. 24082 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides 24083 to false. 24084 (aarch64_override_options_internal): Update to set 24085 PARAM_PREFETCH_DYNAMIC_STRIDES. 24086 * doc/invoke.texi (prefetch-dynamic-strides): Document new option. 24087 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New. 24088 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define. 24089 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for 24090 prefetch-dynamic-strides setting. 24091 24092 2018-05-07 Luis Machado <luis.machado@linaro.org> 24093 24094 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune) 24095 <minimum_stride>: New const int field. 24096 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include 24097 minimum_stride field. 24098 (exynosm1_prefetch_tune): Likewise. 24099 (thunderxt88_prefetch_tune): Likewise. 24100 (thunderx_prefetch_tune): Likewise. 24101 (thunderx2t99_prefetch_tune): Likewise. 24102 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048. 24103 (aarch64_override_options_internal): Update to set 24104 PARAM_PREFETCH_MINIMUM_STRIDE. 24105 * doc/invoke.texi (prefetch-minimum-stride): Document new option. 24106 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New. 24107 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define. 24108 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if 24109 stride is constant and is below the minimum stride threshold. 24110 241112018-05-07 Luis Machado <luis.machado@linaro.org> 24112 24113 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set 24114 to 512. 24115 241162018-05-07 Luis Machado <luis.machado@linaro.org> 24117 24118 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune) 24119 <prefetch_dynamic_strides>: New const bool field. 24120 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include 24121 prefetch_dynamic_strides. 24122 (exynosm1_prefetch_tune): Likewise. 24123 (thunderxt88_prefetch_tune): Likewise. 24124 (thunderx_prefetch_tune): Likewise. 24125 (thunderx2t99_prefetch_tune): Likewise. 24126 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides 24127 to false. 24128 (aarch64_override_options_internal): Update to set 24129 PARAM_PREFETCH_DYNAMIC_STRIDES. 24130 * doc/invoke.texi (prefetch-dynamic-strides): Document new option. 24131 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New. 24132 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define. 24133 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for 24134 prefetch-dynamic-strides setting. 24135 241362018-05-07 Luis Machado <luis.machado@linaro.org> 24137 24138 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune) 24139 <minimum_stride>: New const int field. 24140 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include 24141 minimum_stride field. 24142 (exynosm1_prefetch_tune): Likewise. 24143 (thunderxt88_prefetch_tune): Likewise. 24144 (thunderx_prefetch_tune): Likewise. 24145 (thunderx2t99_prefetch_tune): Likewise. 24146 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048. 24147 (aarch64_override_options_internal): Update to set 24148 PARAM_PREFETCH_MINIMUM_STRIDE. 24149 * doc/invoke.texi (prefetch-minimum-stride): Document new option. 24150 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New. 24151 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define. 24152 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if 24153 stride is constant and is below the minimum stride threshold. 24154 241552018-05-06 Jakub Jelinek <jakub@redhat.com> 24156 24157 PR c++/85659 24158 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if 24159 the type is addressable. Don't force op into register if it has 24160 BLKmode. 24161 241622018-05-05 Roland McGrath <mcgrathr@google.com> 24163 24164 PR other/77609 24165 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for 24166 any section for which we don't know a specific type it should have, 24167 regardless of name. Previously this was done only for the exact 24168 names ".init_array", ".fini_array", and ".preinit_array". 24169 (default_elf_asm_named_section): Add comment about 24170 relationship with default_section_type_flags and SECTION_NOTYPE. 24171 (get_section): Don't consider it a type conflict if one side has 24172 SECTION_NOTYPE and the other doesn't, as long as neither has the 24173 SECTION_BSS et al used in the default_section_type_flags logic. 24174 241752018-05-05 Tom de Vries <tom@codesourcery.com> 24176 24177 PR target/85653 24178 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define. 24179 (workaround_barsyncs): New function. 24180 (nvptx_reorg): Use workaround_barsyncs. 24181 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR. 24182 (define_expand "nvptx_membar_cta"): New define_expand. 24183 (define_insn "*nvptx_membar_cta"): New insn. 24184 241852018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com> 24186 24187 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs. 24188 To improve optimization opportunities. 24189 * builtin-types.def: The new needed builtin types for the above. 24190 241912018-05-04 Richard Biener <rguenther@suse.de> 24192 24193 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check. 24194 * gimple-ssa-store-merging.c 24195 (imm_store_chain_info::output_merged_store): Remove redundant create, 24196 release split_store vector contents on failure. 24197 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking 24198 scalar stmt vector on cache hit. 24199 242002018-05-04 Segher Boessenkool <segher@kernel.crashing.org> 24201 24202 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove 24203 Xilinx FP support. 24204 * config.gcc (powerpc-xilinx-eabi*): Remove. 24205 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP 24206 support. 24207 (fusion_addis_mem_combo_load): Ditto. 24208 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx 24209 FP support. 24210 (rs6000_cpu_cpp_builtins): Ditto. 24211 * config/rs6000/rs6000-linux.c 24212 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto. 24213 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete. 24214 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP 24215 support. 24216 (rs6000_setup_reg_addr_masks): Ditto. 24217 (rs6000_init_hard_regno_mode_ok): Ditto. 24218 (rs6000_option_override_internal): Ditto. 24219 (legitimate_lo_sum_address_p): Ditto. 24220 (rs6000_legitimize_address): Ditto. 24221 (rs6000_legitimize_reload_address): Ditto. 24222 (rs6000_legitimate_address_p): Ditto. 24223 (abi_v4_pass_in_fpr): Ditto. 24224 (setup_incoming_varargs): Ditto. 24225 (rs6000_gimplify_va_arg): Ditto. 24226 (rs6000_split_multireg_move): Ditto. 24227 (rs6000_savres_strategy): Ditto. 24228 (rs6000_emit_prologue_components): Ditto. 24229 (rs6000_emit_epilogue_components): Ditto. 24230 (rs6000_emit_prologue): Ditto. 24231 (rs6000_emit_epilogue): Ditto. 24232 (rs6000_elf_file_end): Ditto. 24233 (rs6000_function_value): Ditto. 24234 (rs6000_libcall_value): Ditto. 24235 * config/rs6000/rs6000.h: Ditto. 24236 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ... 24237 (TARGET_MINMAX): ... this. New. 24238 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete. 24239 * config/rs6000/rs6000.md: Remove Xilinx FP support. 24240 (*movsi_internal1_single): Delete. 24241 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu, 24242 mfpu=, mxilinx-fpu): Delete. 24243 * config/rs6000/singlefp.h: Delete. 24244 * config/rs6000/sysv4.h: Remove Xilinx FP support. 24245 * config/rs6000/t-rs6000: Ditto. 24246 * config/rs6000/t-xilinx: Delete. 24247 * config/rs6000/titan.md: Adjust for fp_type removal. 24248 * config/rs6000/vsx.md: Remove Xilinx FP support. 24249 (VStype_simple): Delete. 24250 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete. 24251 * config/rs6000/xfpu.h: Delete. 24252 * config/rs6000/xfpu.md: Delete. 24253 * config/rs6000/xilinx.h: Delete. 24254 * config/rs6000/xilinx.opt: Delete. 24255 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove 24256 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu. 24257 242582018-05-04 Tom de Vries <tom@codesourcery.com> 24259 24260 PR libgomp/85639 24261 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target 24262 if ignore == 0. 24263 242642018-05-04 Richard Biener <rguenther@suse.de> 24265 24266 PR middle-end/85627 24267 * tree-complex.c (update_complex_assignment): We are always in SSA form. 24268 (expand_complex_div_wide): Likewise. 24269 (expand_complex_operations_1): Likewise. 24270 (expand_complex_libcall): Preserve EH info of the original stmt. 24271 (tree_lower_complex): Handle removed blocks. 24272 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW 24273 on complex multiplication and division libcall builtins. 24274 242752018-05-04 Richard Biener <rguenther@suse.de> 24276 24277 PR middle-end/85574 24278 * fold-const.c (negate_expr_p): Restrict negation of operand 24279 zero of a division to when we know that can happen without 24280 overflow. 24281 (fold_negate_expr_1): Likewise. 24282 242832018-05-04 Jakub Jelinek <jakub@redhat.com> 24284 24285 PR libstdc++/85466 24286 * real.h (real_nextafter): Declare. 24287 * real.c (real_nextafter): New function. 24288 * fold-const-call.c (fold_const_nextafter): New function. 24289 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and 24290 CASE_CFN_NEXTTOWARD. 24291 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss 24292 even when arg1_mode is different from arg0_mode. 24293 242942018-05-03 Nathan Sidwell <nathan@acm.org> 24295 24296 * doc/extend.texi (Deprecated Features): Remove 24297 -ffriend-injection. 24298 (Backwards Compatibility): Likewise. 24299 * doc/invoke.texi (C++ Language Options): Likewise. 24300 (C++ Dialect Options): Likewise. 24301 243022018-05-03 Jakub Jelinek <jakub@redhat.com> 24303 24304 PR target/85530 24305 * config/i386/avx512fintrin.h (_mm512_mullox_epi64, 24306 _mm512_mask_mullox_epi64): New intrinsics. 24307 243082018-05-03 Tom de Vries <tom@codesourcery.com> 24309 24310 PR testsuite/85106 24311 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization 24312 dump files): Add offload-tree. 24313 243142018-05-03 Richard Biener <rguenther@suse.de> 24315 24316 PR tree-optimization/85615 24317 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits 24318 to loops not nested in BBs loop father to avoid creating multi-entry 24319 loops. 24320 243212018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 24322 24323 PR tree-optimization/70291 24324 * tree-complex.c (expand_complex_libcall): Add type, inplace_p 24325 arguments. Change return type to tree. Emit libcall as a new 24326 statement rather than replacing existing one when inplace_p is true. 24327 (expand_complex_multiplication_components): New function. 24328 (expand_complex_multiplication): Expand floating-point complex 24329 multiplication using the above. 24330 (expand_complex_division): Rename inner_type parameter to type. 24331 Update expand_complex_libcall call-site. 24332 (expand_complex_operations_1): Update expand_complex_multiplication 24333 and expand_complex_division call-sites. 24334 243352018-05-02 Jakub Jelinek <jakub@redhat.com> 24336 24337 PR target/85582 24338 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask, 24339 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask, 24340 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that 24341 the highest significant bit of the shift count mask is clear. In 24342 check whether and[sq]i3 is needed verify that all significant bits 24343 of the shift count other than the highest are set. 24344 243452018-05-02 Tom de Vries <tom@codesourcery.com> 24346 24347 PR libgomp/82428 24348 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define. 24349 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID) 24350 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin. 24351 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function. 24352 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size. 24353 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and 24354 __builtin_goacc_parlevel_size. 24355 243562018-05-02 Richard Biener <rguenther@suse.de> 24357 24358 PR tree-optimization/85597 24359 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP 24360 do not use split vect_get_vec_defs call but call vect_get_slp_defs 24361 directly. 24362 243632018-05-02 Tom de Vries <tom@codesourcery.com> 24364 24365 PR testsuite/85106 24366 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization 24367 dump files): Add ltrans-tree. 24368 243692018-05-02 Tom de Vries <tom@codesourcery.com> 24370 24371 PR testsuite/85106 24372 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization 24373 dump files): Add wpa-ipa. 24374 243752018-05-02 Segher Boessenkool <segher@kernel.crashing.org> 24376 24377 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the 24378 powerpc*-*-linux*paired* target. 24379 * config/rs6000/750cl.h: Delete. 24380 * config/rs6000/paired.h: Delete. 24381 * config/rs6000/paired.md: Delete. 24382 * config/rs6000/predicates.md (easy_vector_constant): Remove paired 24383 float support. 24384 * config/rs6000/rs6000-builtin.def: Remove paired float support. 24385 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update 24386 comment. Remove paired float support. 24387 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI. 24388 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete 24389 VECTOR_PAIRED. 24390 * config/rs6000/rs6000-protos.h (paired_expand_vector_init, 24391 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete 24392 declarations. 24393 * config/rs6000/rs6000.c: Remove paired float support. 24394 (paired_expand_vector_init, paired_expand_vector_move, 24395 paired_emit_vector_compare, paired_emit_vector_cond_expr, 24396 (paired_expand_lv_builtin, paired_expand_stv_builtin, 24397 paired_expand_builtin, paired_expand_predicate_builtin, 24398 paired_init_builtins): Delete. 24399 * config/rs6000/rs6000.h: Remove paired float support. 24400 * config/rs6000/rs6000.md: Remove paired float support. 24401 (move_from_CR_ov_bit): Delete. 24402 * config/rs6000/rs6000.opt (mpaired): Delete. 24403 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES. 24404 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired. 24405 244062018-05-02 Richard Biener <rguenther@suse.de> 24407 24408 PR middle-end/85567 24409 * gimplify.c (gimplify_save_expr): When in SSA form allow 24410 SAVE_EXPRs to compute to SSA vars. 24411 244122018-05-02 Jakub Jelinek <jakub@redhat.com> 24413 24414 PR target/85582 24415 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask, 24416 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask, 24417 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't 24418 clobber operands[2], instead use a new pseudo. Formatting fixes. 24419 244202018-05-02 Richard Sandiford <richard.sandiford@linaro.org> 24421 24422 PR tree-optimization/85586 24423 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only 24424 exit early for statements in the same group if the accesses are 24425 not strided. 24426 244272018-05-02 Tom de Vries <tom@codesourcery.com> 24428 24429 PR lto/85451 24430 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload" 24431 error message. 24432 244332018-05-01 Marc Glisse <marc.glisse@inria.fr> 24434 24435 PR tree-optimization/85143 24436 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR. 24437 244382018-05-01 Tom de Vries <tom@codesourcery.com> 24439 24440 PR lto/85451 24441 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler 24442 not found" error message. 24443 244442018-05-01 Tom de Vries <tom@codesourcery.com> 24445 24446 PR other/83786 24447 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define. 24448 * vec.c (test_ordered_remove_if): New function. 24449 (vec_c_tests): Call test_ordered_remove_if. 24450 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO. 24451 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF. 24452 * tree-vect-patterns.c (vect_pattern_recog_1): Use 24453 VEC_ORDERED_REMOVE_IF. 24454 244552018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org> 24456 24457 PR tree-optimization/82665 24458 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle 24459 pointer subtraction where arguments come from a memchr call. 24460 244612018-05-01 Jakub Jelinek <jakub@redhat.com> 24462 24463 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use 24464 --push-state --as-needed and --pop-state instead of --as-needed and 24465 --no-as-needed if ld supports it. 24466 * configure: Regenerated. 24467 24468 PR web/85578 24469 * doc/install.texi2html: Replace _002d with - and _002a with * in 24470 generated html files using sed. 24471 244722018-04-30 David Malcolm <dmalcolm@redhat.com> 24473 24474 PR c++/85523 24475 * gcc-rich-location.c (blank_line_before_p): New function. 24476 (use_new_line): New function. 24477 (gcc_rich_location::add_fixit_insert_formatted): New function. 24478 * gcc-rich-location.h 24479 (gcc_rich_location::add_fixit_insert_formatted): New function. 24480 244812018-04-30 David Malcolm <dmalcolm@redhat.com> 24482 24483 * selftest.c (assert_streq): Rename "expected" and "actual" to 24484 "val1" and "val2". Extend NULL-handling to cover both inputs 24485 symmetrically, while still requiring both to be non-NULL for a pass. 24486 * selftest.h (assert_streq): Rename "expected" and "actual" to 24487 "val1" and "val2". 24488 (ASSERT_EQ): Likewise. 24489 (ASSERT_EQ_AT): Likewise. 24490 (ASSERT_KNOWN_EQ): Likewise. 24491 (ASSERT_KNOWN_EQ_AT): Likewise. 24492 (ASSERT_NE): Likewise. 24493 (ASSERT_MAYBE_NE): Likewise. 24494 (ASSERT_MAYBE_NE_AT): Likewise. 24495 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for 24496 the assertion to pass. 24497 (ASSERT_STREQ_AT): Likewise. 24498 244992018-04-30 Jonathan Wakely <jwakely@redhat.com> 24500 24501 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify 24502 interaction with -pie. 24503 245042018-04-30 David Malcolm <dmalcolm@redhat.com> 24505 24506 * selftest.h: Fix alphabetization of per-source-file selftest 24507 declarations. 24508 245092018-04-30 Jason Merrill <jason@redhat.com> 24510 24511 PR c++/61982 - dead stores to destroyed objects. 24512 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS 24513 of clobber. 24514 245152018-04-30 Jason Merrill <jason@redhat.com> 24516 24517 * tree.c (build_clobber): New. 24518 * tree.h: Declare it. 24519 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it. 24520 245212018-04-30 David Malcolm <dmalcolm@redhat.com> 24522 24523 * diagnostic-show-locus.c (layout::layout): Update for 24524 location_get_source_line returning a char_span. 24525 (struct char_span): Move to input.h. 24526 (struct correction): Update for fields in char_span becoming 24527 private. 24528 (struct source_line): Update for location_get_source_line 24529 returning a char_span. 24530 (layout::print_line): Likewise. 24531 * edit-context.c (edited_file::print_content): Likewise. 24532 (edited_file::print_diff_hunk): Likewise. 24533 (edited_file::print_run_of_changed_lines): Likewise. 24534 (edited_file::get_num_lines): Likewise. 24535 (edited_line::edited_line): Likewise. 24536 * final.c (asm_show_source): Likewise. 24537 * input.c (location_get_source_line): Convert return type 24538 from const char * to char_span, losing the final "line_len" 24539 param. 24540 (dump_location_info): Update for the above. 24541 (get_substring_ranges_for_loc): Likewise. Use a char_span 24542 when handling the literal within the line. 24543 (test_reading_source_line): Update for location_get_source_line 24544 returning a char_span. 24545 * input.h (class char_span): Move here from 24546 diagnostic-show-locus.c, converting from a struct to a class. 24547 Make data members private. 24548 (char_span::operator bool): New. 24549 (char_span::length): New. 24550 (char_span::get_buffer): New. 24551 (char_span::operator[]): New. 24552 (char_span::subspan): Make const. 24553 (char_span::xstrdup): New. 24554 (location_get_source_line): Convert return type from const char * 24555 to char_span, losing the final "line_size" param. 24556 245572018-04-30 Jan Hubicka <jh@suse.cz> 24558 24559 * lto-wrapper.c (ltrans_priorities): New static var. 24560 (cmp_priority): New. 24561 (run_gcc): Read priorities and if doing parallel build order 24562 the Makefile by them. 24563 245642018-04-30 David Malcolm <dmalcolm@redhat.com> 24565 24566 * input.h (builtins_location_check): Convert to a STATIC_ASSERT. 24567 245682018-04-30 Richard Biener <rguenther@suse.de> 24569 24570 * tree-cfg.c (verify_address): Remove base argument, add 24571 flag whether to check TREE_ADDRESSABLE and do that. 24572 (verify_expr): Remove. 24573 (verify_types_in_gimple_reference): Add pieces from verify_expr. 24574 (verify_gimple_assign_single): Likewise. 24575 (verify_gimple_switch): Likewise. 24576 (verify_expr_location_1): Dereference tp once. Add (disabled) 24577 piece from verify_expr. 24578 (verify_gimple_in_cfg): Do not call verify_expr on all ops. 24579 245802018-04-30 Claudiu Zissulescu <claziss@synopsys.com> 24581 24582 * config/arc/linux.h (CLEAR_INSN_CACHE): Define. 24583 245842018-04-30 Claudiu Zissulescu <claziss@synopsys.com> 24585 24586 * config/arc/arc-protos.h (prepare_extend_operands): Remove. 24587 (small_data_pattern): Likewise. 24588 (arc_rewrite_small_data): Likewise. 24589 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove. 24590 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise. 24591 (get_symbol_alignment): New function. 24592 (legitimate_small_data_address_p): Likewise. 24593 (legitimate_scaled_address): Update, call 24594 legitimate_small_data_address_p. 24595 (output_sdata): New static variable. 24596 (arc_print_operand): Update how we handle small data operands. 24597 (arc_print_operand_address): Likewise. 24598 (arc_legitimate_address_p): Update, use 24599 legitimate_small_data_address_p. 24600 (arc_rewrite_small_data_p): Remove. 24601 (arc_rewrite_small_data_1): Likewise. 24602 (arc_rewrite_small_data): Likewise. 24603 (small_data_pattern): Likewise. 24604 (compact_sda_memory_operand): Update to use 24605 legitimate_small_data_address_p and get_symbol_alignment. 24606 (prepare_move_operands): Don't rewite sdata pattern. 24607 (prepare_extend_operands): Remove. 24608 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata 24609 pattern. 24610 (zero_extendqisi2): Likewise. 24611 (zero_extendhisi2): Likewise. 24612 (extendqihi2): Likewise. 24613 (extendqisi2): Likewise. 24614 (extendhisi2): Likewise. 24615 (addsi3): Likewise. 24616 (subsi3): Likewise. 24617 (andsi3): Likewise. 24618 * config/arc/constraints.md (Usd): Change it to memory constraint. 24619 246202018-04-30 Claudiu Zissulescu <claziss@synopsys.com> 24621 24622 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants 24623 as source of std instructions. 24624 * config/arc/arc.md (movsi_insn): Update pattern predicate to 24625 allow 6-bit constants as source for store instructions. 24626 (movdi_insn): Update instruction pattern to allow 6-bit constants 24627 as source for store instructions. 24628 246292018-04-30 Jonathan Wakely <jwakely@redhat.com> 24630 24631 * doc/invoke.texi (-fdebug-types-section): Fix grammar. 24632 246332018-04-30 Nathan Sidwell <nathan@acm.org> 24634 Sandra Loosemore <sandra@codesourcery.com> 24635 24636 * dumpfile.c (dump_open): Allow '-' for stdout. 24637 * doc/invoke.texi (Developer Options): Document dump filename 24638 determination early. Document stdin/stdout selection. 24639 246402018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com> 24641 24642 Microblaze Target: PIC data text relative 24643 24644 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel. 24645 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p): 24646 Add declaration. 24647 * config/microblaze/microblaze.h (microblaze_constant_address_p): 24648 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p. 24649 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL): 24650 New addressing mode for data-text relative position indepenedent code. 24651 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case -> 24652 'ADDRESS_SYMBOLIC_TXT_REL'. 24653 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT. 24654 (microblaze_legitimate_pic_operand): Exclude function calls from 24655 pic operands in case of TARGET_PIC_DATA_TEXT_REL option. 24656 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible 24657 addresses cases. 24658 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case. 24659 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case. 24660 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling 24661 for 'address + offset'. 24662 (microblaze_expand_prologue): Add new function prologue call for 24663 'r20' assignation. 24664 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook 24665 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector 24666 table in case of TARGET_PIC_DATA_TEXT_REL. 24667 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'. 24668 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL): 24669 Add new macros 'UNSPEC_TEXT', 24670 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue 24671 + exclude function calls from 'UNSPEC_PLT' in case of data text 24672 relative mode. 24673 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add 24674 new target hook for generating address diff vector tables in case of 24675 flag_pic. 24676 * doc/tm.texi : Regenerate. 24677 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition 24678 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case 24679 of addr diff vector generation. 24680 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add 24681 target hook definition. 24682 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): 24683 Add default function for generate_pic_addr_diff_vec -> flag_pic. 24684 * doc/invoke.texi (Add new pic option): Add new microblaze pic 24685 option for data text relative. 24686 246872018-04-30 Richard Biener <rguenther@suse.de> 24688 24689 * tree-chrec.h (evolution_function_is_constant_p): Remove 24690 redundant check. 24691 * tree-cfg.c (tree_node_can_be_shared): Re-order checks. 24692 246932018-04-30 Richard Biener <rguenther@suse.de> 24694 24695 PR bootstrap/85571 24696 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=]. 24697 246982018-04-30 Richard Biener <rguenther@suse.de> 24699 24700 PR tree-optimization/28364 24701 PR tree-optimization/85275 24702 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after 24703 copying first exit test. 24704 247052018-04-28 Mark Wielaard <mark@klomp.org> 24706 24707 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for 24708 dwarf_version >= 5. 24709 (dwarf_AT): Handle DW_AT_addr_base. 24710 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base. 24711 247122018-04-28 Uros Bizjak <ubizjak@gmail.com> 24713 24714 PR target/84431 24715 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern. 24716 (*ashl<dwi>3_doubleword_mask_1): Ditto. 24717 (*<shift_insn><dwi>3_doubleword_mask): Ditto. 24718 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto. 24719 247202018-04-28 Richard Biener <rguenther@suse.de> 24721 24722 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument. 24723 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts 24724 to reflect use. Only add interesting stmts. 24725 247262018-04-27 Martin Jambor <mjambor@suse.cz> 24727 24728 PR ipa/85549 24729 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure 24730 the jump function allows for passing through aggregate values. 24731 247322018-04-27 David Malcolm <dmalcolm@redhat.com> 24733 24734 * input.h (in_system_header_at): Convert from macro to inline 24735 function. 24736 (from_macro_expansion_at): Likewise. 24737 (from_macro_definition_at): Likewise. 24738 247392018-04-27 Jeff Law <law@redhat.com> 24740 24741 * config.gcc: Mark tile* targets as deprecated/obsolete. 24742 247432018-04-27 Richard Biener <rguenther@suse.de> 24744 24745 * config/aarch64/aarch64.c: Simplify ap.__stack advance and 24746 fix for ILP32. 24747 247482018-04-27 Richard Biener <rguenther@suse.de> 24749 24750 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable. 24751 247522018-04-27 Uros Bizjak <ubizjak@gmail.com> 24753 24754 * config/i386/i386.md (*movti_internal): Substitute Ye constraint 24755 with Yd constraint. Set "preferred_for_speed" attribute from 24756 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives 24757 with Yd constraint. 24758 (*movdi_internal): Ditto. 24759 (movti_interunit splitters): Remove 24760 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition. 24761 (movdi_interunit splitters): Ditto. 24762 * config/i386/constraints.md (Ye): Remove. 24763 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC. 24764 247652018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 24766 24767 PR target/85512 24768 * config/aarch64/constraints.md (Usg): Limit to 31. 24769 (Usj): Limit to 63. 24770 247712018-04-27 Jakub Jelinek <jakub@redhat.com> 24772 24773 PR tree-optimization/85529 24774 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB 24775 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE, 24776 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious 24777 zero extension or masking of the MSB bit. 24778 (optimize_range_tests): Add FIRST_BB argument, pass it through 24779 to optimize_range_tests_var_bound. 24780 (maybe_optimize_range_tests, reassociate_bb): Adjust 24781 optimize_range_tests callers. 24782 247832018-04-26 Richard Biener <rguenther@suse.de> 24784 Jakub Jelinek <jakub@redhat.com> 24785 24786 * cgraph.h (symbol_table): Just declare debug method here. 24787 * symtab.c (symbol_table::debug): Define. 24788 247892018-04-26 Eric Botcazou <ebotcazou@adacore.com> 24790 24791 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer. 24792 247932018-04-26 Uros Bizjak <ubizjak@gmail.com> 24794 24795 * config/i386/i386.md ("isa" attribute): Add x64_sse2. 24796 ("enabled" attribute): Handle x64_sse2 "isa" attribute. 24797 (*movdi_internal): Substitute Yi and Yj constraint with x 24798 and Ym and Yn constraint with y constraint. Update "isa" 24799 attribute and set "preferred_for_speed" attribute from 24800 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives. 24801 (*movsi_internal): Ditto. 24802 (*movdf_internal): Ditto. 24803 (*movsf_internal): Ditto. 24804 (*zero_extendsidi2): Ditto. 24805 * config/i386/sse.md (vec_set<mode>_0): Ditto. 24806 (sse2_loadld): Ditto. 24807 (*vec_extract<ssevecmodelower>_0): Ditto. 24808 (*vec_extractv4si_0_zext_sse4): Ditto. 24809 (vec_concatv2di): Ditto. 24810 (*vec_dup<mode>): Ditto. 24811 * config/i386/mmx.md (*mov<mode>_internal): Ditto. 24812 * config/i386/constraints.md (Yi): Remove. 24813 (Yj): Remove. 24814 (Ym): Remove. 24815 (Yn): Remove. 24816 248172018-04-26 Nathan Sidwell <nathan@acm.org> 24818 24819 * dumpfile.c (dump_open): New. 24820 (dump_open_alternate_stream, dump_start, dump_begin): Call it. 24821 (dump_finish): Detect stdio/stderr by value not name. 24822 248232018-04-26 Jonathan Wakely <jwakely@redhat.com> 24824 24825 * doc/invoke.texi (-Wreturn-type): Document default status for C++. 24826 248272018-04-26 Tom de Vries <tom@codesourcery.com> 24828 24829 PR target/84952 24830 * config/nvptx/nvptx.c (verify_neutering_jumps) 24831 (verify_neutering_labels): New function 24832 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels. 24833 248342018-04-26 Tom de Vries <tom@codesourcery.com> 24835 24836 PR target/84025 24837 * config/nvptx/nvptx.c (needs_neutering_p): New function. 24838 (nvptx_single): Use needs_neutering_p to skip over insns that do not 24839 need neutering. 24840 248412018-04-26 Richard Biener <rguenther@suse.de> 24842 Tom de Vries <tom@codesourcery.com> 24843 24844 PR lto/85422 24845 * lto-streamer-out.c (output_function): Fixup loops if required to match 24846 discovery done in the reader. 24847 248482018-04-26 Richard Biener <rguenther@suse.de> 24849 24850 PR tree-optimization/85116 24851 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should 24852 have a loop exit from the single latch predecessor. Remove 24853 case of header with just condition. 24854 (ch_base::copy_headers): Exclude infinite loops from any 24855 processing. 24856 (pass_ch::execute): Record exits. 24857 248582018-04-26 Richard Biener <rguenther@suse.de> 24859 24860 * tree-vect-data-refs.c (vect_get_data_access_cost): Get 24861 prologue cost vector and pass it to vect_get_load_cost. 24862 (vect_get_peeling_costs_all_drs): Likewise. 24863 (vect_peeling_hash_get_lowest_cost): Likewise. 24864 (vect_enhance_data_refs_alignment): Likewise. 24865 248662018-04-26 Richard Biener <rguenther@suse.de> 24867 24868 PR middle-end/85450 24869 * tree-cfg.c (verify_gimple_assign_unary): Restore proper 24870 checking of integer<->pointer conversions. 24871 * omp-expand.c (expand_omp_for_static_nochunk): Avoid 24872 sign-/zero-extending pointer types. 24873 (expand_omp_for_static_chunk): Likewise. 24874 248752018-03-22 Hans-Peter Nilsson <hp@axis.com> 24876 Jean Lee <xiaoyur347@gmail.com> 24877 24878 * config/mips/mips.c (mips_asan_shadow_offset): New function. 24879 (TARGET_ASAN_SHADOW_OFFSET): Define. 24880 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be 24881 true for -fsanitize=address. 24882 248832018-04-25 Mark Wielaard <mark@klomp.org> 24884 24885 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before 24886 shorter ones. 24887 248882018-04-25 Jakub Jelinek <jakub@redhat.com> 24889 24890 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather 24891 than "alu", remove explicit "memory" and "imm_disp" attributes. 24892 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise. 24893 24894 PR middle-end/85414 24895 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND, 24896 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to 24897 gen_lowpart_no_emit. 24898 248992018-04-25 Sebastian Peryt <sebastian.peryt@intel.com> 24900 24901 PR target/85473 24902 * config/i386/i386.c (ix86_expand_builtin): Change memory 24903 operand to XI, extend p0 to Pmode. 24904 * config/i386/i386.md: Change unspec volatile and operand 24905 1 mode to XI, change operand 0 mode to P. 24906 249072018-04-25 Chung-Ju Wu <jasonwucj@gmail.com> 24908 24909 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with 24910 GET_MODE_MASK before any checking. 24911 (nds32_can_use_bset_p): Likewise. 24912 (nds32_can_use_btgl_p): Likewise. 24913 249142018-04-25 Chung-Ju Wu <jasonwucj@gmail.com> 24915 24916 * config/nds32/nds32-doubleword.md: New define_split pattern for 24917 illegal register number. 24918 249192018-04-25 Chung-Ju Wu <jasonwucj@gmail.com> 24920 24921 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier. 24922 249232018-04-25 Chung-Ju Wu <jasonwucj@gmail.com> 24924 24925 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character. 24926 249272018-04-25 Richard Biener <rguenther@suse.de> 24928 24929 * lto-streamer.h (LTO_major_version): Bump to 8. 24930 249312018-04-25 Jakub Jelinek <jakub@redhat.com> 24932 24933 * BASE-VER: Set to 9.0.0. 24934 249352018-04-24 Segher Boessenkool <segher@kernel.crashing.org> 24936 24937 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes 24938 in __abskf2 and __powikf2. 24939 249402018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 24941 24942 PR target/85512 24943 * config/aarch64/constraints.md (Usg, Usj): New constraints. 24944 * config/aarch64/iterators.md (cmode_simd): New mode attribute. 24945 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3): 24946 Use the above on operand 2. Reindent. 24947 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise. 24948 249492018-04-24 H.J. Lu <hongjiu.lu@intel.com> 24950 24951 PR target/85485 24952 * common/config/i386/i386-common.c (ix86_handle_option): Don't 24953 handle OPT_mcet. 24954 * config/i386/i386.opt (mcet): Removed. 24955 * doc/install.texi: Remove -mcet documentation. 24956 * doc/invoke.texi: Likewise. 24957 249582018-04-24 H.J. Lu <hongjiu.lu@intel.com> 24959 24960 PR target/85485 24961 * doc/install.texi: Remove -mcet from bootstrap-cet. 24962 249632018-04-24 Jakub Jelinek <jakub@redhat.com> 24964 24965 PR target/85511 24966 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define 24967 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32 24968 if TARGET_64BIT. 24969 24970 PR target/85503 24971 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if 24972 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool 24973 containing a CONST_VECTOR. 24974 249752018-04-24 Cesar Philippidis <cesar@codesourcery.com> 24976 24977 * doc/install.texi: Update newlib dependency for nvptx. 24978 249792018-04-24 Jakub Jelinek <jakub@redhat.com> 24980 24981 PR target/85508 24982 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL 24983 instead of INTVAL when shifting x left. 24984 249852018-04-24 Andreas Krebbel <krebbel@linux.ibm.com> 24986 24987 PR tree-optimization/85478 24988 * tree-vect-loop.c (vect_analyze_loop_2): Do not call 24989 vect_grouped_store_supported for single element vectors. 24990 249912018-04-24 Richard Biener <rguenther@suse.de> 24992 24993 PR target/85491 24994 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided 24995 load cost increase to the case of non-constant step. 24996 249972018-04-24 Jakub Jelinek <jakub@redhat.com> 24998 24999 PR target/84828 25000 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead 25001 destination if any_malformed_asm. 25002 250032018-04-23 Eric Botcazou <ebotcazou@adacore.com> 25004 25005 PR middle-end/85496 25006 * expr.c (store_field): In the bitfield case, if the value comes from 25007 a function call and is returned in registers by means of a PARALLEL, 25008 do not change the mode of the temporary unless BLKmode and VOIDmode. 25009 250102018-04-23 Andrey Belevantsev <abel@ispras.ru> 25011 25012 PR rtl-optimization/85423 25013 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard 25014 dependencies to debug insns when the previous insn is non-debug. 25015 250162018-04-23 Claudiu Zissulescu <claziss@synopsys.com> 25017 25018 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec 25019 enums into a single definition. 25020 (fls): Fix predicates and printing. 25021 (seti): Likewise. 25022 250232018-04-23 Claudiu Zissulescu <claziss@synopsys.com> 25024 25025 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove. 25026 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers 25027 and short u6 immediate. 25028 (check_if_valid_sleep_operand): Remove. 25029 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates. 25030 250312018-04-22 Chung-Ju Wu <jasonwucj@gmail.com> 25032 25033 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider 25034 flag_always_save_lp condition. 25035 * config/nds32/nds32.opt (malways-save-lp): New option. 25036 250372018-04-22 Shiva Chen <shiva0217@gmail.com> 25038 25039 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare. 25040 * config/nds32/nds32.c (nds32_use_load_post_increment): New. 25041 * config/nds32/nds32.h 25042 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define. 25043 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define. 25044 250452018-04-22 Shiva Chen <shiva0217@gmail.com> 25046 25047 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove. 25048 * config/nds32/nds32.c (nds32_ls_333_p): Remove. 25049 250502018-04-22 Shiva Chen <shiva0217@gmail.com> 25051 Chung-Ju Wu <jasonwucj@gmail.com> 25052 25053 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode): 25054 Declare. 25055 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function. 25056 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify. 25057 250582018-04-22 Chung-Ju Wu <jasonwucj@gmail.com> 25059 25060 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value. 25061 250622018-04-22 Chung-Ju Wu <jasonwucj@gmail.com> 25063 25064 * config/nds32/nds32-protos.h (nds32_data_alignment, 25065 nds32_local_alignment): Declare. 25066 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment, 25067 nds32_local_alignment): New functions. 25068 (TARGET_CONSTANT_ALIGNMENT): Define. 25069 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define. 25070 250712018-04-22 Chung-Ju Wu <jasonwucj@gmail.com> 25072 25073 * config/nds32/nds32.c 25074 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file. 25075 (TARGET_MODES_TIEABLE_P): Likewise. 25076 250772018-04-22 Chung-Ju Wu <jasonwucj@gmail.com> 25078 25079 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization 25080 level Ofast and Og. 25081 250822018-04-22 Monk Chiang <sh.chiang04@gmail.com> 25083 Chung-Ju Wu <jasonwucj@gmail.com> 25084 25085 * config/nds32/constants.md (unspec_volatile_element): Add enum values 25086 for unaligned access. 25087 * config/nds32/nds32-intrinsic.c: Implementation of expanding 25088 unaligned access. 25089 * config/nds32/nds32-intrinsic.md: Likewise. 25090 * config/nds32/nds32_intrinsic.h: Likewise. 25091 * config/nds32/nds32.h (nds32_builtins): Likewise. 25092 * config/nds32/nds32.opt (munaligned-access): New option. 25093 * config/nds32/nds32.c (nds32_asm_file_start): Display 25094 flag_unaligned_access status. 25095 250962018-04-20 Kito Cheng <kito.cheng@gmail.com> 25097 25098 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if 25099 -mno-relax is present. 25100 * config/riscv/linux.h (LINK_SPEC): Ditto. 25101 251022018-04-20 Martin Sebor <msebor@redhat.com> 25103 25104 PR c/85365 25105 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict 25106 for null pointers. 25107 (gimple_fold_builtin_stxcpy_chk): Same. 25108 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same. 25109 251102018-04-20 Michael Meissner <meissner@linux.ibm.com> 25111 25112 PR target/85456 25113 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call 25114 __powikf2 when long double is IEEE 128-bit. 25115 251162018-04-20 Kito Cheng <kito.cheng@gmail.com> 25117 25118 * config/riscv/riscv.c (riscv_first_stack_step): Round up min 25119 step to make sure stack always aligned. 25120 251212018-04-20 Carl Love <cel@us.ibm.com> 25122 25123 PR target/83402 25124 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add 25125 size check for arg0. 25126 251272018-04-20 Nathan Sidwell <nathan@codesourcery.com> 25128 Tom de Vries <tom@codesourcery.com> 25129 25130 PR target/85445 25131 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining): 25132 Emit insns for calls too. 25133 (nvptx_find_par): Always look for worker-level predecessor insn. 25134 (nvptx_propagate): Add is_call parm, return bool. Copy frame for 25135 calls. 25136 (nvptx_vpropagate, nvptx_wpropagate): Adjust. 25137 (nvptx_process_pars): Propagate frames for calls. 25138 251392018-04-20 H.J. Lu <hongjiu.lu@intel.com> 25140 25141 PR target/85469 25142 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET): 25143 Removed. 25144 (OPTION_MASK_ISA_IBT_UNSET): Likewise. 25145 (ix86_handle_option): Don't handle OPT_mibt. 25146 * config/i386/cet.h: Check __CET__ instead of __IBT__ and 25147 __SHSTK__. 25148 * config/i386/driver-i386.c (host_detect_local_cpu): Remove 25149 has_ibt and ibt. 25150 * config/i386/i386-c.c (ix86_target_macros_internal): Don't 25151 check OPTION_MASK_ISA_IBT nor flag_cf_protection. 25152 (ix86_target_macros): Define __CET__ with flag_cf_protection 25153 for -fcf-protection. 25154 * config/i386/i386.c (isa2_opts): Remove -mibt. 25155 * config/i386/i386.h (TARGET_IBT): Removed. 25156 (TARGET_IBT_P): Likewise. 25157 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt. 25158 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT. 25159 * config/i386/i386.opt (mcet): Update help message. 25160 (mshstk): Likewise. 25161 (mibt): Removed. 25162 * doc/invoke.texi: Remove -mibt. Document __CET__. Document 25163 -mcet as an alias for -mshstk. 25164 251652018-04-20 Richard Biener <rguenther@suse.de> 25166 25167 PR middle-end/85475 25168 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential 25169 complexity by forcing a single use of the multiply operand. 25170 251712018-04-20 Martin Jambor <mjambor@suse.cz> 25172 25173 ipa/85449 25174 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding 25175 recursion dependency to only apply to non-clones. 25176 251772018-04-20 Martin Jambor <mjambor@suse.cz> 25178 25179 ipa/85447 25180 * ipa-cp.c (create_specialized_node): Check that clones of 25181 self-recursive edges exist during IPA-CP. 25182 251832018-04-19 Toon Moene <toon@moene.org> 25184 25185 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled 25186 by -O3. 25187 251882018-04-19 Jakub Jelinek <jakub@redhat.com> 25189 25190 PR tree-optimization/85467 25191 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use 25192 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the 25193 VECTOR_CST element to type. 25194 251952018-04-19 H.J. Lu <hongjiu.lu@intel.com> 25196 25197 PR target/85397 25198 * config/i386/i386.h (STACK_SAVEAREA_MODE): New. 25199 * config/i386/i386.md (builtin_setjmp_setup): Removed. 25200 (builtin_longjmp): Likewise. 25201 (save_stack_nonlocal): New pattern. 25202 (restore_stack_nonlocal): Likewise. 25203 252042018-04-19 H.J. Lu <hongjiu.lu@intel.com> 25205 25206 PR target/85404 25207 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet): 25208 Replace ASM_OUTPUT_LABEL with fprintf. 25209 252102018-04-19 H.J. Lu <hongjiu.lu@intel.com> 25211 25212 PR target/85417 25213 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet): 25214 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK. 25215 * config/i386/i386-c.c (ix86_target_macros_internal): Also 25216 define __IBT__ and __SHSTK__ for -fcf-protection. 25217 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check 25218 TARGET_IBT. 25219 (ix86_trampoline_init): Likewise. 25220 (x86_output_mi_thunk): Likewise. 25221 (ix86_notrack_prefixed_insn_p): Likewise. 25222 (ix86_option_override_internal): Don't disallow -fcf-protection. 25223 * config/i386/i386.md (rdssp<mode>): Also enable for 25224 -fcf-protection. 25225 (incssp<mode>): Likewise. 25226 (nop_endbr): Likewise. 25227 * config/i386/i386.opt (mcet): Change help message to built-in 25228 functions only. 25229 (mibt): Likewise. 25230 (mshstk): Likewise. 25231 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition 25232 on -fcf-protection. Change -mcet, -mibt and -mshstk to only 25233 enable CET built-in functions. 25234 252352018-04-19 Sebastian Peryt <sebastian.peryt@intel.com> 25236 25237 * common/config/i386/i386-common.c 25238 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET, 25239 OPTION_MASK_ISA_MOVDIRI_UNSET, 25240 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines. 25241 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b. 25242 * config.gcc (movdirintrin.h): New header. 25243 * config/i386/cpuid.h (bit_MOVDIRI, 25244 bit_MOVDIR64B): New bits. 25245 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri 25246 and -mmvodir64b. 25247 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED), 25248 (VOID, PVOID, PCVOID)): New function types. 25249 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32, 25250 __builtin_ia32_directstoreu_u64, 25251 __builtin_ia32_movdir64b): New builtins. 25252 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New. 25253 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b 25254 and -mmovdiri. 25255 (ix86_valid_target_attribute_inner_p): Ditto. 25256 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED 25257 and VOID_FTYPE_PUNSIGNED_UNSIGNED. 25258 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B. 25259 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P, 25260 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New. 25261 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New. 25262 (movdiri<mode>, movdir64b_<mode>): New. 25263 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b. 25264 * config/i386/immintrin.h: Include movdirintrin.h. 25265 * config/i386/movdirintrin.h: New file. 25266 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b. 25267 252682018-04-19 Richard Biener <rguenther@suse.de> 25269 25270 PR middle-end/85455 25271 * cfg.c (clear_bb_flags): When loop state says we have 25272 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP. 25273 252742018-04-19 Richard Biener <rguenther@suse.de> 25275 25276 PR tree-optimization/84737 25277 * tree-vect-data-refs.c (vect_copy_ref_info): New function 25278 copying restrict info. 25279 (vect_setup_realignment): Use it. 25280 * tree-vectorizer.h (vect_copy_ref_info): Declare. 25281 * tree-vect-stmts.c (vectorizable_store): Copy ref info from 25282 the first DR to all generated stores. 25283 (vectorizable_load): Likewise for loads. 25284 252852018-04-19 Jakub Jelinek <jakub@redhat.com> 25286 25287 PR tree-optimization/85446 25288 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require 25289 the integral and pointer types to have the same precision. 25290 25291 * doc/install.texi: Document --disable-cet being the default and 25292 --enable-cet=auto. 25293 252942018-04-18 Martin Liska <mliska@suse.cz> 25295 25296 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding 25297 style. 25298 252992018-04-18 Martin Liska <mliska@suse.cz> 25300 25301 Revert 25302 2018-03-02 Eric Botcazou <ebotcazou@adacore.com> 25303 25304 PR ipa/83983 25305 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both 25306 arguments if they are comparable. 25307 253082018-04-18 Martin Liska <mliska@suse.cz> 25309 25310 Revert 25311 2018-03-13 Eric Botcazou <ebotcazou@adacore.com> 25312 25313 PR lto/84805 25314 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of 25315 incomplete types. 25316 253172018-04-18 H.J. Lu <hongjiu.lu@intel.com> 25318 25319 PR target/85388 25320 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert 25321 ENDBR after calling __morestack. 25322 253232018-04-18 David Malcolm <dmalcolm@redhat.com> 25324 25325 PR jit/85384 25326 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version 25327 by using gcc_base_ver to generate a gcc_driver_version, and use 25328 it when generating GCC_DRIVER_NAME. 25329 * configure: Regenerate. 25330 253312018-04-18 Jakub Jelinek <jakub@redhat.com> 25332 25333 PR target/81084 25334 * config.gcc: Obsolete powerpc*-*-*spe*. 25335 253362018-04-17 Jakub Jelinek <jakub@redhat.com> 25337 25338 PR debug/84637 25339 * dbxout.c (dbxout_int): Perform negation in unsigned int type. 25340 (stabstr_D): Change type of unum from unsigned int to 25341 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT 25342 type. 25343 253442018-04-17 Jim Wilson <jimw@sifive.com> 25345 25346 PR 84856 25347 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to 25348 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size. 25349 Set arg_pointer_offset after using pretend_args_size. 25350 253512018-04-17 Jakub Jelinek <jakub@redhat.com> 25352 25353 PR rtl-optimization/85431 25354 * dse.c (record_store): Ignore zero width stores. 25355 25356 PR sanitizer/85230 25357 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit 25358 __asan_allocas_unpoison call and last_alloca_addr = new_sp before 25359 __builtin_stack_restore rather than after it. 25360 * builtins.c (expand_asan_emit_allocas_unpoison): Pass 25361 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second 25362 argument instead of virtual_dynamic_stack_rtx. 25363 253642018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org> 25365 25366 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p): 25367 New prototype. 25368 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): 25369 Add note to error message to explain internal mapping of overloaded 25370 built-in function name to non-overloaded built-in function name. 25371 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New 25372 function. 25373 253742018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com> 25375 25376 PR target/85424 25377 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack 25378 where the inputs overlap with the output. 25379 253802018-04-17 Jakub Jelinek <jakub@redhat.com> 25381 25382 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add 25383 (=v, v) alternative and explicit "memory" attribute. 25384 (vec_extract_lo_<mode><mask_name>): Likewise. Also add 25385 "type", "prefix", "prefix_extra", "length_immediate" and "mode" 25386 attributes. 25387 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use 25388 "sselog1" type instead of "sselog". 25389 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of 25390 "sselog". Remove explicit "memory" attribute. 25391 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory", 25392 "type", "prefix", "prefix_extra", "length_immediate" and "mode" 25393 attributes. 25394 (vec_extract_hi_v32hi): Merge all alternatives into one, use 25395 "sselog1" type instead of "sselog". Remove explicit "memory" 25396 attribute. 25397 (vec_extract_hi_v16hi): Merge each pair of alternatives into one, 25398 use "sselog1" type instead of "sselog". Remove explicit "memory" 25399 attribute. 25400 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory", 25401 "type", "prefix", "prefix_extra", "length_immediate" and "mode" 25402 attributes. 25403 (vec_extract_hi_v64qi): Merge all alternatives into one, use 25404 "sselog1" type instead of "sselog". Remove explicit "memory" 25405 attribute. 25406 (vec_extract_hi_v32qi): Merge each pair of alternatives into one, 25407 use "sselog1" type instead of "sselog". Remove explicit "memory" 25408 attribute. 25409 25410 PR target/85430 25411 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu. 25412 25413 PR middle-end/85414 25414 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG 25415 on a SUBREG. 25416 254172018-04-17 Martin Jambor <mjambor@suse.cz> 25418 25419 PR ipa/85421 25420 * ipa-cp.c (create_specialized_node): Call 25421 expand_all_artificial_thunks if necessary. 25422 254232018-04-17 Martin Liska <mliska@suse.cz> 25424 25425 PR lto/85405 25426 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing 25427 in message, remote space in between '_G' and '('. 25428 254292018-04-17 Jakub Jelinek <jakub@redhat.com> 25430 25431 PR target/85281 25432 * config/i386/sse.md (reduces<mode><mask_scalar_name>, 25433 avx512f_vmcmp<mode>3<round_saeonly_name>, 25434 avx512f_vmcmp<mode>3_mask<round_saeonly_name>, 25435 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>, 25436 avx512f_rndscale<mode><round_saeonly_name>, 25437 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>, 25438 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>): 25439 Use %<iptr>2 instead of %2 for -masm=intel. 25440 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>, 25441 avx512f_vcvttss2usi<round_saeonly_name>, 25442 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for 25443 -masm=intel. 25444 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>, 25445 avx512f_vcvttsd2usi<round_saeonly_name>, 25446 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>): 25447 Use %q1 instead of %1 for -masm=intel. 25448 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>, 25449 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead 25450 of %3 for -masm=intel. 25451 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for 25452 -masm=intel. 25453 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for 25454 -masm=intel. 25455 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for 25456 -masm=intel. 25457 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with 25458 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and 25459 %g1. 25460 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for 25461 -masm=intel. 25462 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with 25463 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and 25464 %g1 and one with %0 and %1. 25465 (avx512er_vmrcp28<mode><round_saeonly_name>, 25466 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of 25467 %1 for -masm=intel. 25468 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask, 25469 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask, 25470 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order 25471 of %0 and %{%4%} for -masm=intel. 25472 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz, 25473 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz, 25474 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap 25475 order of %0 and %{%5%}%{z%} for -masm=intel. 25476 254772018-04-17 Jan Hubicka <jh@suse.cz> 25478 25479 PR lto/85405 25480 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields. 25481 254822018-04-17 Martin Liska <mliska@suse.cz> 25483 25484 PR ipa/85329 25485 * multiple_target.c (create_dispatcher_calls): Set apostrophes 25486 for target_clone error message. Make default implementation 25487 clone to be a local declaration. 25488 (separate_attrs): Add new argument and check for an empty 25489 string. 25490 (expand_target_clones): Handle it. 25491 (ipa_target_clone): Make redirection just for target_clones 25492 functions. 25493 254942018-04-16 Cesar Philippidis <cesar@codesourcery.com> 25495 Tom de Vries <tom@codesourcery.com> 25496 25497 PR middle-end/84955 25498 * omp-expand.c (expand_oacc_for): Add dummy false branch for 25499 tiled basic blocks without omp continue statements. 25500 255012018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com> 25502 25503 PR target/83660 25504 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark 25505 vec_extract expression as having side effects to make sure it gets 25506 a cleanup point. 25507 255082018-04-16 H.J. Lu <hongjiu.lu@intel.com> 25509 25510 PR target/85403 25511 * config/i386/i386.c (get_builtin_code_for_version): Check 25512 error_mark_node. 25513 255142018-04-16 Olga Makhotina <olga.makhotina@intel.com> 25515 25516 PR target/84331 25517 * config.gcc: Support "skylake". 25518 * config/i386/i386-c.c (ix86_target_macros_internal): Handle 25519 PROCESSOR_SKYLAKE. 25520 * config/i386/i386.c (m_SKYLAKE): Define. 25521 (processor_target_table): Add "skylake". 25522 (ix86_option_override_internal): Add "skylake". 25523 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE, 25524 PROCESSOR_CANNONLAKE. 25525 (get_builtin_code_for_version): Fix priority for 25526 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER, 25527 PROCESSOR_SKYLAKE-AVX512. 25528 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE. 25529 (processor_type): Add PROCESSOR_SKYLAKE. 25530 255312018-04-16 Paolo Carlini <paolo.carlini@oracle.com> 25532 Jason Merrill <jason@redhat.com> 25533 25534 PR c++/85112 25535 * convert.c (convert_to_integer_1): Use direct recursion for 25536 enumeral types and types with a precision less than the number 25537 of bits in their mode. 25538 255392018-04-16 Julia Koval <julia.koval@intel.com> 25540 25541 PR target/84413 25542 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL, 25543 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512 25544 255452018-04-14 Segher Boessenkool <segher@kernel.crashing.org> 25546 25547 PR target/85293 25548 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated. 25549 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move 25550 and -mno-direct-move. 25551 255522018-04-13 Paul A. Clarke <pc@us.ibm.com> 25553 25554 PR target/83402 25555 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}): 25556 Ensure that vec_splat_s32 is only called with 0 <= shift < 16. 25557 Ensure negative shifts result in {0}. 25558 255592018-04-13 Vladimir Makarov <vmakarov@redhat.com> 25560 25561 PR rtl-optimization/79916 25562 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard 25563 regs (if any) to define how to gnerate SD moves when LRA is in 25564 progress. 25565 255662018-04-13 Jakub Jelinek <jakub@redhat.com> 25567 25568 PR rtl-optimization/85393 25569 * except.h (expand_dw2_landing_pad_for_region): Remove declaration. 25570 * except.c (expand_dw2_landing_pad_for_region): Make static. 25571 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just 25572 a label and unconditional jump to old_bb, rather than 25573 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ 25574 basic block. 25575 25576 PR rtl-optimization/85376 25577 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and 25578 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX 25579 instead of a specific value. 25580 255812018-04-13 Jan Hubicka <hubicka@ucw.cz> 25582 Bin Cheng <bin.cheng@arm.com> 25583 25584 PR tree-optimization/82965 25585 PR tree-optimization/83991 25586 * cfgloopanal.c (expected_loop_iterations_unbounded): Add 25587 by_profile_only parameter. 25588 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile 25589 information if the loop was predicted to iterate too many times. 25590 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype 25591 255922018-04-13 Jan Hubicka <hubicka@ucw.cz> 25593 25594 PR lto/71991 25595 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for 25596 always inline. 25597 255982018-04-13 Martin Liska <mliska@suse.cz> 25599 Jakub Jelinek <jakub@redhat.com> 25600 25601 PR middle-end/81657 25602 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET. 25603 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET. 25604 * builtins.c (expand_builtin_memory_copy_args): Use 25605 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target, 25606 handle dest_addr == pc_rtx. 25607 256082018-04-12 Segher Boessenkool <segher@kernel.crashing.org> 25609 25610 PR target/85291 25611 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if 25612 asked to not generate direct moves. 25613 (fix_trunc<mode>si2_stfiwx): Similar. 25614 (fix_trunc<mode>si2_internal): Similar. 25615 256162018-04-12 Jakub Jelinek <jakub@redhat.com> 25617 25618 PR debug/83157 25619 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST. 25620 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest, 25621 lookup if dest in some wider mode is known to be const0_rtx and 25622 if so, record permanent equivalence for it to be ZERO_EXTEND of 25623 the narrower mode destination. 25624 256252018-04-12 Cesar Philippidis <cesar@codesourcery.com> 25626 25627 * lto-streamer-out.c (output_function): Revert 259346. 25628 * omp-expand.c (expand_oacc_for): Likewise. 25629 256302018-04-12 Alexander Monakov <amonakov@ispras.ru> 25631 25632 PR rtl-optimization/85354 25633 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call... 25634 * sel-sched.c (sel_global_init): ... here. 25635 256362018-04-12 Eric Botcazou <ebotcazou@adacore.com> 25637 25638 PR target/85238 25639 * lto-wrapper.c (debug_objcopy): Open the files in binary mode. 25640 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO 25641 mode for PE-COFF targets. 25642 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare. 25643 (i386_pe_asm_lto_end): Likewise. 25644 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define. 25645 (TARGET_ASM_LTO_END): Likewise. 25646 * config/i386/winnt.c (saved_debug_info_level): New static variable. 25647 (i386_pe_asm_lto_start): New function. 25648 (i386_pe_asm_lto_end): Likewise. 25649 256502018-04-12 Cesar Philippidis <cesar@codesourcery.com> 25651 Richard Biener <rguenther@suse.de> 25652 25653 PR middle-end/84955 25654 * lto-streamer-out.c (output_function): Fix CFG loop state before 25655 streaming out. 25656 * omp-expand.c (expand_oacc_for): Handle calls to internal 25657 functions like regular functions. 25658 256592018-04-12 Richard Biener <rguenther@suse.de> 25660 25661 PR lto/85371 25662 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label] 25663 for the early LTO debug to properly generate references to it 25664 during DIE emission. Do not re-use that for the skeleton for 25665 split-dwarf. 25666 (dwarf2out_early_finish): Likewise. 25667 256682018-04-12 Jakub Jelinek <jakub@redhat.com> 25669 25670 PR target/85328 25671 * config/i386/sse.md 25672 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split, 25673 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split, 25674 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi, 25675 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg 25676 and output is a reg, avoid creating invalid lowpart subreg, but 25677 instead split into a 512-bit move. Don't split if not AVX512VL, 25678 input is xmm16+ reg and output is a mem. 25679 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi, 25680 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is 25681 xmm16+ reg and output is a mem. 25682 256832018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com> 25684 25685 * config/s390/s390.c (s390_output_indirect_thunk_function): Check 25686 also for flag_dwarf2_cfi_asm. 25687 256882018-04-12 Jakub Jelinek <jakub@redhat.com> 25689 25690 PR rtl-optimization/85342 25691 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use 25692 a bool scalar var inside of the loop instead. Don't try to update 25693 recog_data.operand after failed apply_change_group. 25694 256952018-04-12 Tom de Vries <tom@codesourcery.com> 25696 25697 PR target/85296 25698 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function. 25699 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined 25700 array with flexible array member as array without given dimension. 25701 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call 25702 argument for undefined param to true. 25703 257042018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com> 25705 25706 PR target/85321 25707 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options 25708 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe 25709 from PowerPC section. 25710 * config/rs6000/sysv4.opt (mcall-): Improve help text. 25711 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim 25712 help text that is too long. 25713 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim 25714 help text that is too long. 25715 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim 25716 help text that is too long. 25717 257182018-04-11 Uros Bizjak <ubizjak@gmail.com> 25719 25720 * config/alpha/alpha.md (stack_probe_internal): Rename 25721 from "probe_stack". Update all callers. 25722 257232018-04-11 Alexander Monakov <amonakov@ispras.ru> 25724 25725 PR rtl-optimization/84566 25726 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking 25727 sched_macro_fuse_insns. 25728 257292018-04-11 Alexander Monakov <amonakov@ispras.ru> 25730 25731 PR target/84301 25732 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here... 25733 (compute_block_dependences): ... from here. 25734 257352018-04-11 Jakub Jelinek <jakub@redhat.com> 25736 25737 PR tree-optimization/85331 25738 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type 25739 from int to HOST_WIDE_INT. 25740 257412018-04-11 Martin Jambor <mjambor@suse.cz> 25742 25743 PR ipa/84149 25744 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment. 25745 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is 25746 not the same as the source val. 25747 (cgraph_edge_brings_value_p): New parameter. 25748 (gather_edges_for_value): Pass destination value to 25749 cgraph_edge_brings_value_p. 25750 (perhaps_add_new_callers): Likewise. 25751 (get_info_about_necessary_edges): Likewise and exclude values brought 25752 only by self-recursive edges. 25753 (create_specialized_node): Redirect only clones of self-calling edges. 25754 (+self_recursive_pass_through_p): New function. 25755 (find_more_scalar_values_for_callers_subset): Use it. 25756 (find_aggregate_values_for_callers_subset): Likewise. 25757 (known_aggs_to_agg_replacement_list): Removed. 25758 (decide_whether_version_node): Re-calculate known constants for all 25759 remaining context clones. 25760 257612018-04-11 Richard Biener <rguenther@suse.de> 25762 25763 PR lto/85339 25764 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute 25765 from early DWARF output. 25766 (dwarf2out_early_finish): Output line info unconditionally into 25767 early DWARF and add reference to it. 25768 257692018-04-11 Jakub Jelinek <jakub@redhat.com> 25770 25771 PR target/85281 25772 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases. 25773 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes 25774 other than V2DFmode using iptr mode attribute. 25775 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute. 25776 257772018-04-11 Alexander Monakov <amonakov@ispras.ru> 25778 25779 PR rtl-optimization/84659 25780 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg. 25781 257822018-04-11 Jakub Jelinek <jakub@redhat.com> 25783 25784 PR debug/85302 25785 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if 25786 SIZEP is NULL. 25787 (output_loc_list): Pass address of a dummy size variable even in the 25788 locview handling loop. 25789 (index_location_lists): Add comment on why skip_loc_list_entry can't 25790 call size_of_locs. 25791 257922018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com> 25793 25794 PR target/85261 25795 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand 25796 into register. 25797 257982018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com> 25799 25800 PR target/85321 25801 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options 25802 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit, 25803 and -mstring-compare-inline-limit. 25804 258052018-04-10 Segher Boessenkool <segher@kernel.crashing.org> 25806 25807 PR target/85287 25808 * config/rs6000/rs6000.md (allocate_stack): Put the residual size 25809 for stack clash protection in a register whenever we need it to be in 25810 a register. 25811 258122018-04-10 Segher Boessenkool <segher@kernel.crashing.org> 25813 25814 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct): 25815 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF. 25816 258172018-04-10 Segher Boessenkool <segher@kernel.crashing.org> 25818 25819 PR target/85321 25820 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in 25821 the help text. 25822 (mlong-double-): Ditto. 25823 * config/rs6000/sysv4.opt (msdata=): Ditto. 25824 (mtls-size=): Ditto. 25825 258262018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org> 25827 25828 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove 25829 erroneous entries for 25830 "vector int vec_ldl (int, long int *)", and 25831 "vector unsigned int vec_ldl (int, unsigned long int *)". 25832 Add comments and entries for 25833 "vector bool char vec_ldl (int, bool char *)", 25834 "vector bool short vec_ldl (int, bool short *)", 25835 "vector bool int vec_ldl (int, bool int *)", 25836 "vector bool long long vec_ldl (int, bool long long *)", 25837 "vector pixel vec_ldl (int, pixel *)", 25838 "vector long long vec_ldl (int, long long *)", 25839 "vector unsigned long long vec_ldl (int, unsigned long long *)". 25840 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new 25841 type tree bool_long_long_type_node and correct definition of 25842 bool_V2DI_type_node to make reference to this new type tree. 25843 (rs6000_mangle_type): Replace erroneous reference to 25844 bool_long_type_node with bool_long_long_type_node. 25845 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add 25846 comments to emphasize sign distinctions for char and int types and 25847 replace RS6000_BTI_bool_long constant with 25848 RS6000_BTI_bool_long_long constant. Also add comment to restrict 25849 use of RS6000_BTI_pixel. 25850 (bool_long_type_node): Remove this macro definition. 25851 (bool_long_long_type_node): New macro definition 25852 258532018-04-10 Jakub Jelinek <jakub@redhat.com> 25854 25855 PR rtl-optimization/85300 25856 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also 25857 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if 25858 simplify_unary_operation fails. 25859 258602018-04-10 Martin Liska <mliska@suse.cz> 25861 25862 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node, 25863 cgraph_edge and ipa_ref. 25864 258652018-04-10 Jakub Jelinek <jakub@redhat.com> 25866 25867 PR target/85177 25868 PR target/85255 25869 * config/i386/sse.md 25870 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix 25871 computation of the VEC_MERGE selector from mask. 25872 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>): 25873 Fix decoding of the VEC_MERGE selector into mask. 25874 258752018-04-10 Richard Sandiford <richard.sandiford@linaro.org> 25876 25877 PR tree-optimization/85286 25878 * tree-vect-data-refs.c (vect_get_smallest_scalar_type): 25879 258802018-04-10 Richard Sandiford <richard.sandiford@linaro.org> 25881 25882 * final.c (final_1): Set insn_last_address as well as 25883 insn_current_address. 25884 258852018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 25886 25887 PR target/85173 25888 * explow.c (emit_stack_probe): Call validize_mem on memory location 25889 before passing it to gen_probe_stack. Create address operand and 25890 legitimize it for the probe_stack_address case. 25891 258922018-04-09 Jan Hubicka <jh@suse.cz> 25893 25894 PR lto/85078 25895 * ipa-devirt.c (rebuild_type_inheritance-hash): New. 25896 * ipa-utils.h (rebuild_type_inheritance-hash): Declare. 25897 * tree.c (free_lang_data_in_type): Fix handling of binfos; 25898 walk basetypes. 25899 (free_lang_data): Rebuild type inheritance graph. 25900 259012018-04-09 Martin Sebor <msebor@redhat.com> 25902 25903 * invoke.texi (-finline-small-functions): Mention other optimization 25904 options. 25905 (-findirect-inlining, -fpartial-inlining): Same. 25906 (-finline-functions-called-once): Same. 25907 (-freorder-blocks-and-partition): Same. 25908 259092018-04-09 Jan Hubicka <jh@suse.cz> 25910 25911 PR rtl/84058 25912 * cfgcleanup.c (try_forward_edges): Do not give up on crossing 25913 jumps; choose last target that matches the criteria (i.e. 25914 no partition changes for non-crossing jumps). 25915 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic 25916 support for redirecting crossing jumps to non-crossing. 25917 259182018-04-09 Alexey Brodkin <abrodkin@synopsys.com> 25919 25920 * config/arc/arc.c (arc_expand_prologue): Set stack usage info 25921 also for naked functions. 25922 259232018-04-09 Claudiu Zissulescu <claziss@synopsys.com> 25924 25925 * config/arc/arc.md (add_shift): New pattern. 25926 (add_shift2): Likewise. 25927 (sub_shift): Likewise. 25928 (sub_shift_cmp0_noout): Likewise. 25929 (compare_si_ashiftsi): Likewise. 25930 (xbfu_cmp0_noout): New combine pattern. 25931 (xbfu_cmp0"): Likewise. 25932 (movsi_set_cc_insn): Place the predicable variant first. 25933 (commutative_binary_cmp0_noout): Remove clobber. 25934 (commutative_binary_cmp0): New pattern. 25935 (noncommutative_binary_cmp0): Likewise. 25936 (noncommutative_binary_cmp0_noout): Likewise. 25937 (noncommutative_binary_comparison_result_used): Removed. 25938 (rsub_cmp0): New pattern. 25939 (rsub_cmp0_noout): Likewise. 25940 (extzvsi): Changed, keep only meaningful variants. 25941 (SQH, SEZ): New iterators. 25942 (SQH_postfix): New mode attribute. 25943 (SEZ_prefix): New code attribute. 25944 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern. 25945 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise. 25946 * config/arc/predicates.md (cc_set_register): Use CC_REG instead 25947 of numerical value. 25948 (noncommutative_operator): Check the availability of barrel 25949 shifter option. 25950 259512018-04-09 Richard Biener <rguenther@suse.de> 25952 25953 PR tree-optimization/85284 25954 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions): 25955 Only use the niter constraining form of simple_iv when the exit 25956 is always executed. 25957 259582018-04-09 Tom de Vries <tom@codesourcery.com> 25959 25960 PR target/84041 25961 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR. 25962 (define_expand "*memory_barrier"): New define_expand. 25963 (define_insn "memory_barrier"): New insn. 25964 259652018-04-09 Andrey Belevantsev <abel@ispras.ru> 25966 25967 PR rtl-optimization/80463 25968 PR rtl-optimization/83972 25969 PR rtl-optimization/83480 25970 25971 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the 25972 correct producer for the insn. 25973 (tidy_control_flow): Fixup seqnos in case of debug insns. 25974 259752018-04-09 Andrey Belevantsev <abel@ispras.ru> 25976 25977 PR rtl-optimization/83913 25978 25979 * sel-sched-ir.c (merge_expr_data): Choose the middle between two 25980 different sched-times when merging exprs. 25981 259822018-04-09 Andrey Belevantsev <abel@ispras.ru> 25983 25984 PR rtl-optimization/83962 25985 25986 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call 25987 tidy_fallthru_edge and tidy_control_flow. 25988 259892018-04-09 Andrey Belevantsev <abel@ispras.ru> 25990 25991 PR rtl-optimization/83530 25992 25993 * sel-sched.c (force_next_insn): New global variable. 25994 (remove_insn_for_debug): When force_next_insn is true, also leave only 25995 next insn in the ready list. 25996 (sel_sched_region): When the region wasn't scheduled, make another pass 25997 over it with force_next_insn set to 1. 25998 259992018-04-08 Monk Chiang <sh.chiang04@gmail.com> 26000 26001 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h 26002 into tm_file. 26003 * config/nds32/constants.md (unspec_volatile_element): Add enum values 26004 for interrupt control. 26005 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic 26006 functions for interrupt control. 26007 * config/nds32/nds32-intrinsic.md: Likewise. 26008 * config/nds32/nds32_intrinsic.h: Likewise. 26009 * config/nds32/nds32.h (nds32_builtins): Likewise. 26010 260112018-04-08 Chung-Ju Wu <jasonwucj@gmail.com> 26012 26013 * config/nds32/nds32.c (nds32_init_machine_status, 26014 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider 26015 strict_aligned_p field. 26016 (nds32_expand_to_rtl_hook): New function. 26017 (TARGET_EXPAND_TO_RTL_HOOK): Define. 26018 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field. 26019 260202018-04-08 Kito Cheng <kito.cheng@gmail.com> 26021 Chung-Ju Wu <jasonwucj@gmail.com> 26022 26023 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu. 26024 * config/nds32/nds32-n7.md: New file. 26025 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7. 26026 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7 26027 pipeline. 26028 * config/nds32/nds32-protos.h: More declarations for n7 pipeline. 26029 * config/nds32/nds32.md (pipeline_model): Add n7. 26030 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus. 26031 * config/nds32/pipelines.md: Include n7 settings. 26032 260332018-04-08 Kito Cheng <kito.cheng@gmail.com> 26034 Chung-Ju Wu <jasonwucj@gmail.com> 26035 26036 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu. 26037 * config/nds32/nds32-e8.md: New file. 26038 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8. 26039 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8 26040 pipeline. 26041 * config/nds32/nds32-protos.h: More declarations for e8 pipeline. 26042 * config/nds32/nds32.md (pipeline_model): Add e8. 26043 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus. 26044 * config/nds32/pipelines.md: Include e8 settings. 26045 260462018-04-08 Kito Cheng <kito.cheng@gmail.com> 26047 Chung-Ju Wu <jasonwucj@gmail.com> 26048 26049 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu. 26050 * config/nds32/nds32-n8.md: New file. 26051 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8. 26052 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8 26053 pipeline. 26054 * config/nds32/nds32-protos.h: More declarations for n8 pipeline. 26055 * config/nds32/nds32-utils.c: More implementations for n8 pipeline. 26056 * config/nds32/nds32.md (pipeline_model): Add n8. 26057 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus. 26058 * config/nds32/pipelines.md: Include n8 settings. 26059 260602018-04-08 Kito Cheng <kito.cheng@gmail.com> 26061 Chung-Ju Wu <jasonwucj@gmail.com> 26062 26063 * config.gcc (nds32*): Add nds32-utils.o into extra_objs. 26064 * config/nds32/nds32-n9-2r1w.md: New file. 26065 * config/nds32/nds32-n9-3r2w.md: New file. 26066 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type, 26067 nds32_register_ports): New or modify for cpu n9. 26068 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9 26069 pipeline. 26070 * config/nds32/nds32-protos.h: More declarations for n9 pipeline. 26071 * config/nds32/nds32-utils.c: New file. 26072 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE, 26073 TARGET_MUL_SLOW): Define. 26074 * config/nds32/nds32.md (pipeline_model): New attribute. 26075 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports): 26076 New options that support cpu n9. 26077 * config/nds32/pipelines.md: Include n9 settings. 26078 * config/nds32/t-nds32 (nds32-utils.o): Add dependency. 26079 260802018-04-08 Chung-Ju Wu <jasonwucj@gmail.com> 26081 26082 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align 26083 information if necessary. 26084 (output_cond_branch_compare_zero): Likewise. 26085 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case. 26086 (nds32_target_alignment): Refine for alignment. 26087 * config/nds32/nds32.h (NDS32_ALIGN_P): Define. 26088 (FUNCTION_BOUNDARY): Modify. 26089 * config/nds32/nds32.md (call_internal, call_value_internal): Consider 26090 align case. 26091 * config/nds32/nds32.opt (malways-align, malign-functions): New. 26092 260932018-04-08 Monk Chiang <sh.chiang04@gmail.com> 26094 26095 * config/nds32/constants.md (unspec_volatile_element): Add values for 26096 TLB operation and data prefetch. 26097 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic 26098 functions for TLB operation and data prefetch. 26099 * config/nds32/nds32-intrinsic.md: Likewise. 26100 * config/nds32/nds32_intrinsic.h: Likewise. 26101 * config/nds32/nds32.c (nds32_dpref_names): Likewise. 26102 (nds32_print_operand): Likewise. 26103 * config/nds32/nds32.h (nds32_builtins): Likewise. 26104 261052018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org> 26106 Andrew Pinski <pinsika@gcc.gnu.org> 26107 26108 PR middle-end/82976 26109 * match.pd: Use constant_boolean_node of correct type instead of 26110 boolean_true_node or boolean_false_node for simplifying 26111 pointer comparisons to zero. 26112 261132018-04-07 Jakub Jelinek <jakub@redhat.com> 26114 26115 PR tree-optimization/80021 26116 * tree.c (verify_type_variant): Make error call in verify_variant_match 26117 translatable and remove final full stop. 26118 261192018-04-07 Chung-Ju Wu <jasonwucj@gmail.com> 26120 26121 * config/nds32/constants.md (unspec_volatile_element): Add 26122 UNSPEC_VOLATILE_EH_RETURN. 26123 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push, 26124 nds32_output_stack_pop): Support dwarf exception handling process. 26125 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare. 26126 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf 26127 exception handling process. 26128 (nds32_compute_stack_frame): Likewise. 26129 (nds32_return_addr_rtx): Likewise. 26130 (nds32_initial_elimination_offset): Likewise. 26131 (nds32_expand_prologue): Likewise. 26132 (nds32_expand_epilogue): Likewise. 26133 (nds32_dynamic_chain_address): New function. 26134 * config/nds32/nds32.h (machine_function): Add fields for dwarf 26135 exception handling. 26136 (DYNAMIC_CHAIN_ADDRESS): Define. 26137 (EH_RETURN_DATA_REGNO): Define. 26138 (EH_RETURN_STACKADJ_RTX): Define. 26139 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement 26140 patterns for dwarf exception handling. 26141 261422018-04-07 Chung-Ju Wu <jasonwucj@gmail.com> 26143 26144 * config/nds32/nds32.h: Clean up obsolete macros. 26145 261462018-04-07 Monk Chiang <sh.chiang04@gmail.com> 26147 26148 * config/nds32/constants.md (unspec_element, unspec_volatile_element): 26149 Add enum values for particular instructions. 26150 * config/nds32/nds32-intrinsic.c: Implementation of expanding 26151 particular intrinsic functions. 26152 * config/nds32/nds32-intrinsic.md: Likewise. 26153 * config/nds32/nds32_intrinsic.h: Likewise. 26154 * config/nds32/nds32.h (nds32_builtins): Likewise. 26155 * config/nds32/nds32.md (type): Add pbsad and pbsada. 26156 (btst, ave): New patterns for particular instructions. 26157 261582018-04-07 Monk Chiang <sh.chiang04@gmail.com> 26159 26160 * config/nds32/constants.md (unspec_element, unspec_volatile_element): 26161 Add enum values for atomic load/store and memory sync. 26162 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store 26163 and memory sync. 26164 * config/nds32/nds32-intrinsic.md: Likewise. 26165 * config/nds32/nds32_intrinsic.h: Likewise. 26166 * config/nds32/nds32.h (nds32_builtins): Likewise. 26167 261682018-04-07 Jakub Jelinek <jakub@redhat.com> 26169 26170 PR tree-optimization/85257 26171 * fold-const.c (native_encode_vector): If not all elts could fit 26172 and off is -1, return 0 rather than offset. 26173 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass 26174 (offseti - offset2) / BITS_PER_UNIT as 4th argument to 26175 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't 26176 adjust buffer in native_interpret_expr call. 26177 261782018-04-07 Monk Chiang <sh.chiang04@gmail.com> 26179 26180 * config/nds32/constants.md (unspec_volatile_element): Add cache 26181 control enum values. 26182 * config/nds32/nds32-intrinsic.c: Add cache control expand functions. 26183 * config/nds32/nds32-intrinsic.md: Add cache control patterns. 26184 * config/nds32/nds32.c (nds32_cctl_names): New. 26185 (nds32_print_operand): Handle cache control register names. 26186 * config/nds32/nds32.h (nds32_builtins): New enum values. 26187 * config/nds32/nds32_intrinsic.h: Add cache control enum types and 26188 macros. 26189 * config/nds32/nds32.md (type): Add mmu. 26190 * config/nds32/pipelines.md (simple_insn): Add mmu. 26191 261922018-04-07 Chung-Ju Wu <jasonwucj@gmail.com> 26193 26194 * config/nds32/nds32.md (type): Remove call. 26195 * config/nds32/pipelines.md (simple_insn): Likewise. 26196 261972018-04-07 Monk Chiang <sh.chiang04@gmail.com> 26198 26199 * config/nds32/constants.md (unspec_volatile_element): Add 26200 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and 26201 UNSPEC_VOLATILE_FMFCFG. 26202 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin 26203 description for fmfcfg and fmfcsr. 26204 (bdesc_1arg): Add fmtcsr. 26205 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd. 26206 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions. 26207 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd, 26208 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr, 26209 unspec_fmfcfg): New patterns. 26210 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG, 26211 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS, 26212 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD. 26213 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss, 26214 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr, 26215 __nds32__fmfcfg): Define. 26216 262172018-04-07 Monk Chiang <sh.chiang04@gmail.com> 26218 26219 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more 26220 intrinsic register names. 26221 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more 26222 intrinsic register enum values and macros. 26223 262242018-04-07 Chung-Ju Wu <jasonwucj@gmail.com> 26225 26226 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition 26227 for load/store addressing form. 26228 (nds32_print_operand_address): Likewise. 26229 262302018-04-06 Eric Botcazou <ebotcazou@adacore.com> 26231 26232 PR target/85196 26233 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands 26234 based on LABEL_REF. Remove useless assertion. 26235 (pic_address_needs_scratch): Fix formatting. 26236 (sparc_legitimize_pic_address): Minor tweaks. 26237 (sparc_delegitimize_address): Adjust assertion accordingly. 26238 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand 26239 into symbolic_operand. 26240 (movsi_high_pic_label_ref): Likewise. 26241 (movsi_lo_sum_pic_label_ref): Likewise. 26242 (movdi_pic_label_ref): Likewise. 26243 (movdi_high_pic_label_ref): Likewise. 26244 (movdi_lo_sum_pic_label_ref): Likewise. 26245 262462018-04-06 Amaan Cheval <amaan.cheval@gmail.com> 26247 26248 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for 26249 custom LIB_SPEC setup. 26250 262512018-04-06 Ruslan Bukin <br@bsdpad.com> 26252 Kito Cheng <kito.cheng@gmail.com> 26253 26254 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support. 26255 * config/riscv/freebsd.h: New. 26256 262572018-04-06 Chung-Ju Wu <jasonwucj@gmail.com> 26258 26259 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine. 26260 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in 26261 file. 26262 262632018-04-06 Chung-Ju Wu <jasonwucj@gmail.com> 26264 Kito Cheng <kito.cheng@gmail.com> 26265 26266 * config/nds32/nds32-md-auxiliary.c (nds32_output_return, 26267 nds32_output_call, nds32_symbol_binds_local_p): New functions. 26268 * config/nds32/nds32-protos.h (nds32_output_call, 26269 nds32_output_return): Declare. 26270 * config/nds32/nds32.md: Refine all the call and return patterns. 26271 262722018-04-06 Jakub Jelinek <jakub@redhat.com> 26273 26274 PR debug/85252 26275 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only 26276 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST. 26277 26278 PR rtl-optimization/84872 26279 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if 26280 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is 26281 EDGE_CROSSING edge. 26282 262832018-04-06 Tamar Christina <tamar.christina@arm.com> 26284 26285 * expr.c (copy_blkmode_to_reg): Revert 254862. 26286 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise. 26287 262882018-04-06 Richard Biener <rguenther@suse.de> 26289 26290 PR middle-end/85244 26291 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref 26292 after seeing a component reference with an adjacent field. Treat 26293 refs to arrays at struct end of external decls similar to 26294 refs to unconstrained commons. 26295 262962018-04-06 Jakub Jelinek <jakub@redhat.com> 26297 26298 PR sanitizer/85213 26299 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't 26300 look through SAVE_EXPRs with non-side-effects argument. Adjust 26301 recursive calls. 26302 (fold_comparison): Adjust twoval_comparison_p caller, don't handle 26303 save_p here. 26304 263052018-04-06 Richard Biener <rguenther@suse.de> 26306 26307 PR middle-end/85180 26308 * alias.c (find_base_term): New wrapper around find_base_term 26309 unwinding CSELIB_VAL_PTR changes. 26310 (find_base_term): Do not restore CSELIB_VAL_PTR during the 26311 recursion. 26312 263132018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com> 26314 26315 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP 26316 instructions. 26317 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New 26318 constant definitions. 26319 ("nop"): lr 0,0 -> nopr r0 26320 ("nop_lr0", "nop_lr1"): New insn definitions. 26321 263222018-04-06 Chung-Ju Wu <jasonwucj@gmail.com> 26323 26324 * config/nds32/nds32.md (*stack_push, *stack_pop): Use 26325 NDS32_V3PUSH_AVAILABLE_P macro. 26326 263272018-04-06 Monk Chiang <sh.chiang04@gmail.com> 26328 Chung-Ju Wu <jasonwucj@gmail.com> 26329 26330 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking. 26331 (nds32*-*-*): Add float and fpu_config into supported_defaults. 26332 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): 26333 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA. 26334 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN, 26335 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS. 26336 * config/nds32/constraints.md: New constraints and checking for hard 26337 float configuration. 26338 * config/nds32/iterators.md: New mode iterator and attribute for hard 26339 float configuration. 26340 * config/nds32/nds32-doubleword.md: Use hard float alternatives and 26341 patterns. 26342 * config/nds32/nds32-fpu.md: New file. 26343 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to 26344 deal with hard float code generation. 26345 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and 26346 ARCH_V3S. 26347 (abi_type, float_reg_number): New enum type. 26348 * config/nds32/nds32-predicates.c: New predicates for hard float. 26349 * config/nds32/nds32-protos.h: Declare functions for hard float. 26350 * config/nds32/nds32.c: Implementation for hard float configuration. 26351 * config/nds32/nds32.h: Definitions for hard float configuration. 26352 * config/nds32/nds32.md: Include hard float machine description and 26353 modify patterns for hard float configuration. 26354 * config/nds32/nds32.opt: New options for hard float configuration. 26355 * config/nds32/predicates.md: New predicates for hard float 26356 configuration. 26357 263582018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com> 26359 26360 * common/config/nds32/nds32-common.c 26361 (nds32_option_optimization_table): Enable -mreleax-hint by default. 26362 263632018-04-05 Jakub Jelinek <jakub@redhat.com> 26364 26365 PR middle-end/85195 26366 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around 26367 CONSTRUCTOR_ELT (ctor, ...)->value. 26368 263692018-04-05 Uros Bizjak <ubizjak@gmail.com> 26370 26371 PR target/85193 26372 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type. 26373 263742018-04-05 Tom de Vries <tom@codesourcery.com> 26375 26376 PR target/85204 26377 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only 26378 cond jump. 26379 263802018-04-05 Shiva Chen <shiva0217@gmail.com> 26381 Kito Cheng <kito.cheng@gmail.com> 26382 26383 * config/nds32/constraints.md (U33): Fine-tune checking condition. 26384 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto. 26385 * config/nds32/nds32.h (nds32_16bit_address_type): Add 26386 ADDRESS_POST_MODIFY_LO_REG_IMM3U. 26387 263882018-04-05 Shiva Chen <shiva0217@gmail.com> 26389 Kito Cheng <kito.cheng@gmail.com> 26390 26391 * config/nds32/constraints.md (Ufe): New memory constraint. 26392 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format, 26393 nds32_output_16bit_load): Consider r8 register for lwi45.fe format. 26394 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe 26395 operands. 26396 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U. 26397 * config/nds32/nds32.md (*mov<mode>): Adjust pattern. 26398 263992018-04-05 Chung-Ju Wu <jasonwucj@gmail.com> 26400 26401 * config/nds32/nds32.md: Use optimize_size in the condition for 26402 alu-shift instructions. 26403 264042018-04-05 Chung-Ju Wu <jasonwucj@gmail.com> 26405 26406 * config/nds32/nds32.md (divsi4, udivsi4): New patterns. 26407 264082018-04-05 Chung-Ju Wu <jasonwucj@gmail.com> 26409 26410 * config/nds32/nds32.md (negsi2): Refine pattern. 26411 264122018-04-05 Kito Cheng <kito.cheng@gmail.com> 26413 Chung-Ju Wu <jasonwucj@gmail.com> 26414 26415 * config/nds32/iterators.md (shift_rotate): New code iterator. 26416 (shift): New code attribute. 26417 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New. 26418 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare. 26419 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases. 26420 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for 26421 bit-wise operations. 26422 (andsi3, *andsi3): Ditto. 26423 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto. 26424 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto. 26425 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto. 26426 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand, 26427 nds32_ior_operand, nds32_xor_operand): New predicates. 26428 264292018-04-05 Chung-Ju Wu <jasonwucj@gmail.com> 26430 26431 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ... 26432 (addsi3, subsi3): ... this. 26433 264342018-04-05 Chung-Ju Wu <jasonwucj@gmail.com> 26435 26436 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator. 26437 264382018-04-05 Chung-Ju Wu <jasonwucj@gmail.com> 26439 26440 * config/nds32/nds32.md: Adjust indention. 26441 264422018-04-05 Kito Cheng <kito.cheng@gmail.com> 26443 26444 * config/nds32/nds32.md (feature): New attribute. 26445 264462018-04-05 Chung-Ju Wu <jasonwucj@gmail.com> 26447 26448 * config/nds32/nds32.md (subtype): New attribute. 26449 264502018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com> 26451 26452 PR target/85203 26453 * config/arm/arm-builtins.c (arm_expand_builtin): Change 26454 expansion to perform a bitwise AND of the argument followed by a 26455 boolean negation of the result. 26456 264572018-04-04 Peter Bergner <bergner@vnet.ibm.com> 26458 26459 PR rtl-optimization/84878 26460 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine 26461 the basic block. Assert the use reference is not artificial and that 26462 it has an associated insn. 26463 264642018-04-04 Michael Matz <matz@suse.de> 26465 26466 * builtins.c (compute_objsize): Pass correct operand 26467 to array_at_struct_end_p. 26468 264692018-04-04 Richard Biener <rguenther@suse.de> 26470 26471 PR lto/85176 26472 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces 26473 from contexts for DINFO_LEVEL_TERSE and below. 26474 264752018-04-04 Kito Cheng <kito.cheng@gmail.com> 26476 26477 * config/nds32/nds32-doubleword.md (move_<mode>): Require 26478 resiter_operand condition. 26479 * config/nds32/nds32.md (*move<mode>): Ditto. 26480 264812018-04-04 Kito Cheng <kito.cheng@gmail.com> 26482 Monk Chiang <sh.chiang04@gmail.com> 26483 26484 * config/nds32/nds32.md (movmisalign<mode>): New pattern. 26485 264862018-04-04 Chung-Ju Wu <jasonwucj@gmail.com> 26487 26488 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>. 26489 264902018-04-04 Chung-Ju Wu <jasonwucj@gmail.com> 26491 Kito Cheng <kito.cheng@gmail.com> 26492 26493 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code, 26494 nds32_cond_code_str, output_cond_branch, 26495 output_cond_branch_compare_zero, nds32_expand_cbranch, 26496 nds32_expand_cstore, nds32_expand_movcc, 26497 nds32_output_cbranchsi4_equality_zero, 26498 nds32_output_cbranchsi4_equality_reg, 26499 nds32_output_cbranchsi4_equality_reg_or_const_int, 26500 nds32_output_cbranchsi4_greater_less_zero: New functions. 26501 * config/nds32/nds32-protos.h (nds32_expand_cbranch, 26502 nds32_expand_cstore, nds32_expand_movcc, 26503 nds32_output_cbranchsi4_equality_zero, 26504 nds32_output_cbranchsi4_equality_reg, 26505 nds32_output_cbranchsi4_equality_reg_or_const_int, 26506 nds32_output_cbranchsi4_greater_less_zero): Declare. 26507 * config/nds32/predicates.md (nds32_movecc_comparison_operator, 26508 nds32_rimm11s_operand): New predicates. 26509 * config/nds32/nds32.h (nds32_expand_result_type): New enum type. 26510 * config/nds32/nds32.md: Rewrite all the branch and conditional move 26511 patterns. 26512 265132018-04-04 Kito Cheng <kito.cheng@gmail.com> 26514 26515 * config/nds32/nds32-doubleword.md: Refine all the instruction type. 26516 * config/nds32/nds32.md: Ditto. 26517 * config/nds32/pipelines.md: Ditto. 26518 265192018-04-04 Richard Biener <rguenther@suse.de> 26520 26521 PR tree-optimization/85168 26522 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid 26523 propagating abnormals. 26524 265252018-04-04 Chung-Ju Wu <jasonwucj@gmail.com> 26526 26527 * config/nds32/nds32.md (enabled): Use yes/no for this attribute. 26528 265292018-04-04 Chung-Ju Wu <jasonwucj@gmail.com> 26530 Kito Cheng <kito.cheng@gmail.com> 26531 26532 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function. 26533 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare. 26534 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function. 26535 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define. 26536 * config/nds32/nds32.md (sibcall_internal): New. 26537 (sibcall_register): Remove. 26538 (sibcall_immediate): Remove. 26539 (sibcall_value_internal): New. 26540 (sibcall_value_register): Remove. 26541 (sibcall_value_immediate): Remove. 26542 * config/nds32/predicates.md (nds32_general_register_operand): New. 26543 (nds32_call_address_operand): New. 26544 265452018-04-03 Jakub Jelinek <jakub@redhat.com> 26546 26547 PR rtl-optimization/85167 26548 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and 26549 bb_defs if *split_p, instead preinitialize it to NULL. 26550 26551 PR tree-optimization/85156 26552 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid 26553 evaluating the argument multiple times. 26554 265552018-04-03 Bill Schmidt <wschmidt@linux.ibm.com> 26556 26557 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather 26558 than vector. 26559 (_mm_cvtpd_ps): Likewise. 26560 (_mm_cvttpd_epi32): Likewise. 26561 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise. 26562 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine 26563 vector, pixel, and bool following altivec.h include. 26564 265652018-04-03 Martin Sebor <msebor@redhat.com> 26566 26567 * doc/extend.texi (Common Function Attributes): Clarify. 26568 (const attribute): Likewise. 26569 (pure attribute): Likewise. 26570 265712018-04-03 Jakub Jelinek <jakub@redhat.com> 26572 26573 PR target/85169 26574 * config/i386/i386.c (ix86_expand_vector_set): Use 26575 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix. 26576 265772018-04-03 Uros Bizjak <ubizjak@gmail.com> 26578 26579 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic 26580 instructions when changing rounding bits to preserve precision bits 26581 in the x87 control word. 26582 265832018-04-03 Martin Liska <mliska@suse.cz> 26584 26585 PR tree-optimization/82491 26586 * rtl.h (strip_offset_and_add): Replace += suboffset with 26587 poly_uint64 () + suboffset. 26588 265892018-03-29 Martin Liska <mliska@suse.cz> 26590 Martin Jambor <mjambor@suse.cz> 26591 26592 PR ipa/84947 26593 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if 26594 param_type is not an integral or pointer type. 26595 265962018-04-03 Richard Biener <rguenther@suse.de> 26597 26598 * sese.h (recompute_all_dominators): Remove. 26599 266002018-04-02 Martin Sebor <msebor@redhat.com> 26601 26602 * doc/invoke.texi (-Wrestrict): Fix typos. 26603 266042018-04-02 Jim Wilson <jimw@sifive.com> 26605 26606 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero. 26607 * config/riscv/riscv.md (<optab>si3): Use QImode shift count. 26608 (<optab>di3, <optab>si3_extend): Likewise. 26609 (<optab>si3_mask, <optab>si3_mask_1): New. 26610 (<optab>di3_mask, <optab>di3_mask_1): New. 26611 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New. 26612 (lshrsi3_zero_extend_1): Use VOIDmode shift count. 26613 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count. 26614 266152018-04-02 Gerald Pfeifer <gerald@pfeifer.com> 26616 26617 * doc/cpp.texi (Variadic Macros): Fix line continuation in an 26618 example. 26619 266202018-04-02 Chung-Ju Wu <jasonwucj@gmail.com> 26621 26622 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define. 26623 (nds32_canonicalize_comparison): New function. 26624 266252018-04-01 Chung-Ju Wu <jasonwucj@gmail.com> 26626 Kito Cheng <kito.cheng@gmail.com> 26627 Kuan-Lin Chen <kuanlinchentw@gmail.com> 26628 26629 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs. 26630 * config/nds32/constants.md (unspec_volatile_element): Add 26631 UNSPEC_VOLATILE_RELAX_GROUP. 26632 * config/nds32/nds32-relax-opt.c: New file. 26633 * config/nds32/nds32-predicates.c 26634 (nds32_symbol_load_store_p): New function. 26635 * config/nds32/nds32-protos.h 26636 (nds32_symbol_load_store_p): Declare function. 26637 (make_pass_nds32_relax_opt): Declare new rtl pass function. 26638 * config/nds32/nds32.c 26639 (nds32_register_pass): New function to register pass. 26640 (nds32_register_passes): New function to register passes. 26641 * config/nds32/nds32.md (relax_group): New pattern. 26642 * config/nds32/nds32.opt (mrelax-hint): New option. 26643 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency. 26644 266452018-04-01 Kito Cheng <kito.cheng@gmail.com> 26646 26647 * config/nds32/t-nds32: Modify files dependency. 26648 266492018-04-01 Chung-Ju Wu <jasonwucj@gmail.com> 26650 26651 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character. 26652 (PROFILE_HOOK): Define its implementation. 26653 266542018-04-01 Chung-Ju Wu <jasonwucj@gmail.com> 26655 26656 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int 26657 type and 32-bit size. 26658 266592018-04-01 Jakub Jelinek <jakub@redhat.com> 26660 26661 PR middle-end/85090 26662 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F. 26663 (V_128_256): New mode iterator. 26664 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split. 26665 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise. 26666 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead 26667 of V. 26668 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and 26669 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F. 26670 266712018-03-31 Segher Boessenkool <segher@kernel.crashing.org> 26672 26673 PR target/83315 26674 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet) 26675 NaN inputs correctly. 26676 266772018-03-30 Peter Bergner <bergner@vnet.ibm.com> 26678 26679 PR target/80546 26680 * config/rs6000/vsx.md (??r): New mode attribute. 26681 (*vsx_mov<mode>_64bit): Use it. 26682 (*vsx_mov<mode>_32bit): Likewise. 26683 266842018-03-30 Martin Sebor <msebor@redhat.com> 26685 26686 PR tree-optimization/84818 26687 * builtins.c (check_access): Use warning_n. 26688 266892018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com> 26690 26691 PR target/83822 26692 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant 26693 condition. 26694 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant 26695 condition. 26696 266972018-03-30 Julia Koval <julia.koval@intel.com> 26698 26699 PR target/84413 26700 * x86-tune.def (movx, partial_reg_dependency): Enable for 26701 m_SKYLAKE_AVX512. 26702 267032018-03-29 Vladimir Makarov <vmakarov@redhat.com> 26704 26705 PR inline-asm/84985 26706 * lra-constraints.c (process_alt_operands): Move setting 26707 this_alternative_matches below. 26708 267092018-03-29 Martin Liska <mliska@suse.cz> 26710 26711 PR lto/84995. 26712 * doc/invoke.texi: Document how LTO works with debug info. 26713 Describe auto-load support of binutils. Mention 'x86-64' 26714 as valid option value of -march option. 26715 267162018-03-29 Jakub Jelinek <jakub@redhat.com> 26717 26718 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>. 26719 26720 PR c/85094 26721 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT. 26722 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe 26723 OEP_NO_HASH_CHECK for recursive call, to avoid exponential 26724 checking. 26725 267262018-03-28 Peter Bergner <bergner@vnet.ibm.com> 26727 26728 PR target/84912 26729 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define. 26730 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64. 26731 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support 26732 for RS6000_BTM_POWERPC64. 26733 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64 26734 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64. 26735 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro 26736 definition. 26737 (DIVDE): Use it. 26738 (DIVDEU): Likewise. 26739 267402018-03-28 Carl Love <cel@us.ibm.com> 26741 26742 Revert 26743 2017-09-27 Carl Love <cel@us.ibm.com> 26744 26745 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro. 26746 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins. 26747 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the 26748 fctiw instruction. 26749 267502018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com> 26751 26752 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool 26753 instead of __vector bool. 26754 (_mm_max_pu8): Likewise. 26755 (_mm_min_pi16): Likewise. 26756 267572018-03-28 Peter Bergner <bergner@vnet.ibm.com> 26758 26759 PR target/84912 26760 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion. 26761 (DIVWEUO): Likewise. 26762 (DIVDEO): Likewise. 26763 (DIVDEUO): Likewise. 26764 * config/rs6000/rs6000.c (builtin_function_type): Remove support for 26765 DIVWEUO and DIVDEUO. 26766 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs. 26767 (UNSPEC_DIV_EXTEND): Remove deleted unspecs. 26768 (div_extend): Likewise. 26769 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted 26770 builtin function. 26771 (__builtin_divweuo): Likewise. 26772 (__builtin_divdeo): Likewise. 26773 (__builtin_divdeuo): Likewise. 26774 267752018-03-28 Jakub Jelinek <jakub@redhat.com> 26776 26777 PR target/85095 26778 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0, 26779 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns. 26780 26781 PR tree-optimization/82004 26782 * gimple-match-head.c (optimize_pow_to_exp): New function. 26783 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE. 26784 Don't fold to exp if optimize_pow_to_exp is false. 26785 267862018-03-28 Martin Liska <mliska@suse.cz> 26787 26788 PR other/84819 26789 * calls.c (initialize_argument_information): Fix trailing space. 26790 * common.opt: Fix typo and provide better explanation for 26791 -fsanitize-coverage option. 26792 * config/i386/i386.opt: Fix typo. 26793 267942018-03-28 Jakub Jelinek <jakub@redhat.com> 26795 Martin Liska <mliska@suse.cz> 26796 26797 PR sanitizer/85081 26798 * gimplify.c (asan_poison_variable): Don't do the check for 26799 gimplify_omp_ctxp here. 26800 (gimplify_decl_expr): Do it here. 26801 (gimplify_target_expr): Likewise. 26802 268032018-03-28 Martin Liska <mliska@suse.cz> 26804 26805 PR target/84988 26806 * config/i386/i386.c (ix86_function_arg_advance): Do not call 26807 chkp_type_bounds_count if MPX is not enabled. 26808 268092018-03-27 Chung-Ju Wu <jasonwucj@gmail.com> 26810 26811 * config/nds32/nds32.h (BRANCH_COST): Adjust cost. 26812 268132018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com> 26814 26815 PR target/84914 26816 * config/rs6000/rs6000.c (create_complex_muldiv): New helper 26817 function to create the function decl for complex long double 26818 multiply and divide for -mabi=ieeelongdouble. 26819 (init_float128_ieee): Call it. 26820 268212018-03-27 H.J. Lu <hongjiu.lu@intel.com> 26822 26823 PR target/85044 26824 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for 26825 -fcf-protection=branch -mibt. 26826 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes. 26827 268282018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com> 26829 26830 PR target/81863 26831 * config/arm/arm.c (arm_valid_symbolic_address): Handle 26832 arm_word_relocations. 26833 268342018-03-27 Cesar Philippidis <cesar@codesourcery.com> 26835 26836 PR target/85056 26837 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to 26838 extern array declarations. 26839 268402018-03-27 Richard Biener <rguenther@suse.de> 26841 26842 PR middle-end/84067 26843 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with 26844 explicit single_use checks. 26845 268462018-03-27 Richard Biener <rguenther@suse.de> 26847 26848 PR tree-optimization/85082 26849 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces): 26850 Valueize the VUSE. 26851 268522018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com> 26853 26854 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES 26855 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]): 26856 Turn on fasynchronous-unwind-tables and funwind-tables. 26857 268582018-03-26 Uros Bizjak <ubizjak@gmail.com> 26859 26860 PR target/85073 26861 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern. 26862 (*bmi_blsr_<mode>_ccz): Ditto. 26863 268642018-03-26 Tom de Vries <tom@codesourcery.com> 26865 26866 PR tree-optimization/85063 26867 * omp-general.c (offloading_function_p): New function. Factor out 26868 of ... 26869 * omp-offload.c (pass_omp_target_link::gate): ... here. 26870 * omp-general.h (offloading_function_p): Declare. 26871 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable 26872 with attribute omp declare target for offloading functions. 26873 268742018-03-24 Richard Sandiford <richard.sandiford@linaro.org> 26875 26876 PR tree-optimization/84005 26877 * tree-data-ref.h (get_base_for_alignment): Declare. 26878 * tree-data-ref.c (get_base_for_alignment_1): New function. 26879 (get_base_for_alignment): Likewise. 26880 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use 26881 get_base_for_alignment to find a suitable base object, instead 26882 of always using drb->base_address. 26883 268842018-03-23 Jakub Jelinek <jakub@redhat.com> 26885 26886 PR inline-asm/85022 26887 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have 26888 known size by default. 26889 268902018-03-23 Vladimir Makarov <vmakarov@redhat.com> 26891 26892 PR inline-asm/85030 26893 * lra-constraints.c (process_alt_operands): Don't match BLKmode 26894 and non BLKmode operands. 26895 268962018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 26897 26898 PR target/85026 26899 * config/arm/arm.md (unaligned_loadhis): Remove first alternative. 26900 Clean up attributes. 26901 269022018-03-23 Richard Biener <rguenther@suse.de> 26903 26904 PR debug/85020 26905 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when 26906 we are going to emit early debug for LTO. 26907 269082018-03-23 Jakub Jelinek <jakub@redhat.com> 26909 26910 PR inline-asm/85034 26911 * function.c (match_asm_constraints_1): Don't optimize if input 26912 doesn't satisfy general_operand predicate for output's mode. 26913 26914 PR inline-asm/85022 26915 * alias.c (write_dependence_p): Don't require for x_canonicalized 26916 non-VOIDmode if x has VOIDmode. 26917 26918 PR sanitizer/85029 26919 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base), 26920 just don't try to optimize it rather than assert it never happens. 26921 269222018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org> 26923 26924 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X 26925 macro expansions for definition of ST_INTERNAL_<mode> and 26926 LD_INTERNAL_<mode> builtins. 26927 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec): 26928 Remove prototype. 26929 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this 26930 function. 26931 (altivec_expand_st_builtin): Likewise. 26932 (altivec_expand_builtin): Remove calls to deleted functions. 26933 (rs6000_address_for_altivec): Delete this function. 26934 * config/rs6000/vector.md: Remove expands for 26935 vector_altivec_load_<mode> and vector_altivec_store_<mode>. 26936 269372018-03-22 Sudakshina Das <sudi.das@arm.com> 26938 26939 PR target/84826 26940 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes. 26941 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid 26942 re-computing once computed. 26943 (arm_expand_prologue): Compute machine->static_chain_stack_bytes. 26944 (arm_init_machine_status): Initialize 26945 machine->static_chain_stack_bytes. 26946 269472018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org> 26948 26949 PR target/84760 26950 * doc/extend.texi: Add four new prototypes for vec_ld. 26951 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol 26952 definitions for more logical presentation. 26953 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add 26954 entries for V1TI variants of __builtin_altivec_ld builtin. 26955 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and 26956 handling of V1TI variant of LVX icode pattern. 26957 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI. 26958 (rs6000_gimple_fold_builtin): Likewise. 26959 (altivec_init_builtins): Add code to define 26960 __builtin_altivec_lvx_v1ti function. 26961 269622018-03-22 Jakub Jelinek <jakub@redhat.com> 26963 26964 PR inline-asm/84941 26965 * function.c (match_asm_constraints_1): Don't do the optimization 26966 if input isn't a REG, SUBREG, MEM or constant. 26967 269682018-03-22 Tom de Vries <tom@codesourcery.com> 26969 26970 PR tree-optimization/84956 26971 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with 26972 bb_has_abnormal_pred. 26973 269742018-03-22 Jakub Jelinek <jakub@redhat.com> 26975 26976 PR sanitizer/85018 26977 * dwarf2asm.c (dw2_output_indirect_constant_1): Set 26978 DECL_INITIAL (decl) to decl at the end. 26979 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change, 26980 adjust the comment. 26981 269822018-03-21 Joseph Myers <joseph@codesourcery.com> 26983 26984 * doc/extend.texi (__builtin_tgmath): Document when complex 26985 integer types are treated as _Complex _Float64. 26986 269872018-03-21 Tom de Vries <tom@codesourcery.com> 26988 26989 * doc/extend.texi (__builtin_extend_pointer): Remove pasto. 26990 269912018-03-21 Jakub Jelinek <jakub@redhat.com> 26992 26993 PR tree-optimization/84960 26994 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb 26995 if it is ENTRY block, move them into single succ of ENTRY in that case. 26996 269972018-03-21 Richard Sandiford <richard.sandiford@linaro.org> 26998 26999 PR tree-optimization/84811 27000 * poly-int.h (poly_span_traits): Remove the T3 parameter and 27001 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT. 27002 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p): 27003 (known_subrange_p): Update accordingly. Cast each value involved 27004 in the size comparison, rather than casting the result of the 27005 subtraction. 27006 270072018-03-21 Jakub Jelinek <jakub@redhat.com> 27008 27009 PR tree-optimization/84982 27010 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion 27011 by flipping the least significant bit rather than all bits from 27012 bitpos to bitpos + bitsize - 1. 27013 270142018-03-21 Nathan Sidwell <nathan@acm.org> 27015 27016 * doc/extend.texi (Deprecated Features): Remove mention of 27017 long-deleted deprecations. 27018 270192018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE> 27020 27021 PR jit/84288 27022 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set. 27023 * configure: Regenerate. 27024 270252018-03-21 Tom de Vries <tom@codesourcery.com> 27026 27027 PR tree-optimization/83126 27028 * tree-parloops.c (num_phis): New function. 27029 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure. 27030 270312018-03-21 Nathan Sidwell <nathan@acm.org> 27032 27033 * doc/extend.texi (Deprecated Features): Update deprecated flags, 27034 mention anon-struct/union members and trailing attributes. 27035 270362018-03-21 Bin Cheng <bin.cheng@arm.com> 27037 27038 PR tree-optimization/84969 27039 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder 27040 builtin memset partitions if they set different rhs values. 27041 270422018-03-21 Jakub Jelinek <jakub@redhat.com> 27043 27044 PR rtl-optimization/84989 27045 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify 27046 VEC_DUPLICATE with scalar result mode. 27047 270482018-03-21 Martin Liska <mliska@suse.cz> 27049 27050 PR ipa/84963 27051 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove 27052 not intended return statement. 27053 270542018-03-21 Martin Liska <mliska@suse.cz> 27055 27056 PR target/84988 27057 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro. 27058 (chkp_find_bound_slots_1): Limit number of iterations. 27059 270602018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca> 27061 27062 PR target/84838 27063 * Minor grammar fixes for x86 options. 27064 270652018-03-20 Jakub Jelinek <jakub@redhat.com> 27066 27067 PR debug/84875 27068 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves 27069 holding REG_CFA_RESTORE notes, instead turn them into a USE. 27070 270712018-03-20 Peter Bergner <bergner@vnet.ibm.com> 27072 27073 PR target/83789 27074 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn. 27075 (altivec_lvx_<mode>_1op): Likewise. 27076 (altivec_stvx_<mode>_2op): Likewise. 27077 (altivec_stvx_<mode>_1op): Likewise. 27078 (altivec_lvx_<VM2:mode>): New define_expand. 27079 (altivec_stvx_<VM2:mode>): Likewise. 27080 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn. 27081 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise. 27082 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise. 27083 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise. 27084 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders. 27085 (rs6000_gen_lvx): Likewise. 27086 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise. 27087 (altivec_expand_stv_builtin): Likewise. 27088 (altivec_expand_builtin): Likewise. 27089 * config/rs6000/vector.md: Likewise. 27090 270912018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 27092 27093 PR target/82518 27094 * config/arm/arm.c (arm_array_mode_supported_p): Return false for 27095 BYTES_BIG_ENDIAN. 27096 270972018-03-20 Richard Biener <rguenther@suse.de> 27098 27099 PR target/84986 27100 * config/i386/i386.c (ix86_add_stmt_cost): Only cost 27101 sign-conversions as zero, fall back to standard scalar_stmt 27102 cost for the rest. 27103 271042018-03-20 Martin Liska <mliska@suse.cz> 27105 27106 PR ipa/84825 27107 * predict.c (rebuild_frequencies): Handle case when we have 27108 PROFILE_ABSENT, but flag_guess_branch_prob is false. 27109 271102018-03-20 Jakub Jelinek <jakub@redhat.com> 27111 27112 PR target/84990 27113 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off 27114 flag_section_anchors. 27115 * varasm.c (use_blocks_for_decl_p): Remove hack for 27116 dw2_force_const_mem. 27117 27118 PR target/84845 27119 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename 27120 to ... 27121 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't 27122 be created, use lowpart_subreg of operands[0] rather than operands[0] 27123 itself. 27124 (*aarch64_reg_<mode>3_minus_mask): Rename to ... 27125 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this. 27126 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate 27127 and n constraint instead of aarch64_shift_imm_di and Usd. 27128 (*aarch64_reg_<optab>_minus<mode>3): Rename to ... 27129 (*aarch64_<optab>_reg_minus<mode>3): ... this. 27130 271312018-03-20 Sudakshina Das <sudi.das@arm.com> 27132 27133 PR target/82989 27134 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints 27135 to favor GPR over NEON registers. 27136 (<shift>di3_neon): Likewise. 27137 271382018-03-20 Tom de Vries <tom@codesourcery.com> 27139 27140 PR target/84952 27141 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync. 27142 (nvptx_process_pars): Emit bar.sync asap and alap. 27143 271442018-03-20 Tom de Vries <tom@codesourcery.com> 27145 27146 PR target/84954 27147 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update 27148 seen_label if seen_label is already set. 27149 271502018-03-20 Jakub Jelinek <jakub@redhat.com> 27151 27152 PR target/84945 27153 * config/i386/i386.c (fold_builtin_cpu): For features above 31 27154 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0]. 27155 Use 1U instead of 1. Formatting fixes. 27156 27157 PR c/84953 27158 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type 27159 instead of TREE_TYPE (s1) for the return value. 27160 271612018-03-19 Jakub Jelinek <jakub@redhat.com> 27162 27163 PR tree-optimization/84946 27164 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute 27165 bitsize + bitsize in poly_uint64 rather than poly_int64. 27166 27167 PR sanitizer/78651 27168 * dwarf2asm.c: Include fold-const.c. 27169 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR 27170 of decl rather than decl itself. 27171 27172 PR rtl-optimization/84643 27173 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator. 27174 271752018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com> 27176 27177 PR sanitizer/78651 27178 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before 27179 calling assemble_variable. 27180 271812018-03-19 Sudakshina Das <sudi.das@arm.com> 27182 27183 PR target/81647 27184 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify 27185 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED. 27186 271872018-03-19 Jim Wilson <jimw@sifive.com> 27188 27189 PR bootstrap/84856 27190 * config/riscv/riscv.c (riscv_function_arg_boundary): Use 27191 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY. 27192 (riscv_first_stack_step): Likewise. 27193 (riscv_option_override): Use STACK_BOUNDARY instead of 27194 MIN_STACK_BOUNDARY. 27195 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from 27196 MIN_STACK_BOUNDARY. 27197 (BIGGEST_ALIGNMENT): Set to 128. 27198 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY. 27199 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of 27200 STACK_BOUNDARY. 27201 272022018-03-19 Richard Biener <rguenther@suse.de> 27203 27204 PR tree-optimization/84933 27205 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound 27206 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE. 27207 272082018-03-19 Richard Biener <rguenther@suse.de> 27209 27210 PR tree-optimization/84859 27211 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function. 27212 (cond_if_else_store_replacement): Perform sinking operation on 27213 single-store BBs regardless of MAX_STORES_TO_SINK setting. 27214 Generalize what a BB with a single eligible store is. 27215 272162018-03-19 Richard Biener <rguenther@suse.de> 27217 27218 PR tree-optimization/84929 27219 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard 27220 chrec_is_positive against non-chrec arg. 27221 272222018-03-19 Tamar Christina <tamar.christina@arm.com> 27223 27224 PR target/84711 27225 * config/arm/arm.c (arm_can_change_mode_class): revert r258554. 27226 272272018-03-18 Martin Liska <mliska@suse.cz> 27228 27229 PR rtl-optimization/84635 27230 * regrename.c (build_def_use): Use matches_mode only when 27231 matches >= 0. 27232 272332018-03-18 Richard Sandiford <richard.sandiford@linaro.org> 27234 27235 PR tree-optimization/84913 27236 * tree-vect-loop.c (vectorizable_reduction): Don't try to 27237 vectorize chains of COND_EXPRs. 27238 272392018-03-18 Chung-Ju Wu <jasonwucj@gmail.com> 27240 27241 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value. 27242 272432018-03-18 Chung-Ju Wu <jasonwucj@gmail.com> 27244 27245 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define. 27246 272472018-03-18 Chung-Ju Wu <jasonwucj@gmail.com> 27248 27249 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define. 27250 272512018-03-17 Chung-Ju Wu <jasonwucj@gmail.com> 27252 Kito Cheng <kito.cheng@gmail.com> 27253 27254 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare. 27255 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array. 27256 (nds32_adjust_reg_alloc_order): New function. 27257 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define. 27258 272592018-03-17 Kito Cheng <kito.cheng@gmail.com> 27260 27261 * config/nds32/nds32.c (nds32_asm_output_mi_thunk, 27262 nds32_print_operand, nds32_print_operand_address): Use 27263 HOST_WIDE_INT_PRINT_DEC instead. 27264 272652018-03-17 Chung-Ju Wu <jasonwucj@gmail.com> 27266 27267 * config/nds32/nds32.c (nds32_register_priority): Modify cost. 27268 272692018-03-17 Jakub Jelinek <jakub@redhat.com> 27270 27271 PR target/84902 27272 * config/i386/i386.c (initial_ix86_tune_features, 27273 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than 27274 unsigned long long. 27275 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int 27276 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune 27277 rather than 1u << ix86_tune. Formatting fix. 27278 (ix86_option_override_internal): Change ix86_arch_mask from 27279 unsigned int to unsigned HOST_WIDE_INT, initialize to 27280 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch. 27281 (ix86_function_specific_restore): Likewise. 27282 272832018-03-16 Jakub Jelinek <jakub@redhat.com> 27284 27285 PR target/84899 27286 * postreload.c (reload_combine_recognize_pattern): Perform 27287 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and 27288 truncate_int_for_mode the result for the destination's mode. 27289 27290 PR c/84909 27291 * hsa-gen.c (mem_type_for_type): Fix comment typo. 27292 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks): 27293 Likewise. 27294 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset): 27295 Likewise. 27296 272972018-03-16 Vladimir Makarov <vmakarov@redhat.com> 27298 27299 PR target/84876 27300 * lra-assigns.c (lra_split_hard_reg_for): Don't use 27301 regno_allocno_class_array and sorted_pseudos. 27302 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in 27303 insns where regno is used. 27304 273052018-03-16 Martin Liska <mliska@suse.cz> 27306 27307 PR ipa/84833 27308 * multiple_target.c (create_dispatcher_calls): Redirect 27309 reference in the symbol table. 27310 273112018-03-16 Martin Liska <mliska@suse.cz> 27312 27313 PR ipa/84722 27314 * multiple_target.c (create_dispatcher_calls): Redirect also 27315 an alias. 27316 273172018-03-16 Jakub Jelinek <jakub@redhat.com> 27318 27319 PR c++/79937 27320 PR c++/82410 27321 * tree.h (TARGET_EXPR_NO_ELIDE): Define. 27322 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with 27323 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR. 27324 273252018-03-16 Julia Koval <julia.koval@intel.com> 27326 27327 * doc/invoke.texi (Skylake Server): Add CLWB. 27328 Cannonlake): Remove CLWB. 27329 273302018-03-16 Jakub Jelinek <jakub@redhat.com> 27331 27332 PR tree-optimization/84841 27333 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from 27334 1 << 3. 27335 (FLOAT_ONE_CONST_TYPE): Define. 27336 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0. 27337 (sort_by_operand_rank): Put entries with higher constant_type last 27338 rather than first to match comments. 27339 273402018-03-15 Sandra Loosemore <sandra@codesourcery.com> 27341 27342 * config/nios2/nios2.md (movsi_internal): Fix thinko in 27343 split predicate. 27344 273452018-03-15 Jakub Jelinek <jakub@redhat.com> 27346 27347 PR c++/79085 27348 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment 27349 check and use address of target always. 27350 273512018-03-15 H.J. Lu <hongjiu.lu@intel.com> 27352 27353 PR target/84574 27354 * config/i386/i386.c (indirect_thunk_needed): Update comments. 27355 (indirect_thunk_bnd_needed): Likewise. 27356 (indirect_thunks_used): Likewise. 27357 (indirect_thunks_bnd_used): Likewise. 27358 (indirect_return_needed): New. 27359 (indirect_return_bnd_needed): Likewise. 27360 (output_indirect_thunk_function): Add a bool argument for 27361 function return. 27362 (output_indirect_thunk_function): Don't generate alias for 27363 function return thunk. 27364 (ix86_code_end): Call output_indirect_thunk_function to generate 27365 function return thunks. 27366 (ix86_output_function_return): Set indirect_return_bnd_needed 27367 and indirect_return_needed instead of indirect_thunk_bnd_needed 27368 and indirect_thunk_needed. 27369 273702018-03-15 Olga Makhotina <olga.makhotina@intel.com> 27371 27372 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic. 27373 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions. 27374 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves. 27375 273762018-03-15 David Malcolm <dmalcolm@redhat.com> 27377 Paul Hua <paul.hua.gm@gmail.com> 27378 27379 PR c/84852 27380 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp. 27381 273822018-03-15 Segher Boessenkool <segher@kernel.crashing.org> 27383 27384 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the 27385 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode 27386 resp. SFmode cases. 27387 273882018-03-15 Tamar Christina <tamar.christina@arm.com> 27389 27390 PR target/84711 27391 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE 27392 instead of GET_MODE_SIZE when comparing Units. 27393 273942018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com> 27395 27396 PR target/68256 27397 * varasm.c (hash_section): Return an unchangeble hash value 27398 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p): 27399 Return !aarch64_can_use_per_function_literal_pools_p (). 27400 274012018-03-15 Jakub Jelinek <jakub@redhat.com> 27402 27403 PR target/84860 27404 * optabs.c (emit_conditional_move): Pass address of cmode's copy 27405 rather than address of cmode as last argument to prepare_cmp_insn. 27406 274072018-03-15 Julia Koval <julia.koval@intel.com> 27408 27409 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ, 27410 F_AVX512VNNI, F_AVX512BITALG): New. 27411 274122018-03-14 John David Anglin <danglin@gcc.gnu.org> 27413 27414 PR target/83451 27415 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload 27416 insn for floating-point loads and stores. 27417 274182018-03-14 Carl Love <cel@us.ibm.com> 27419 27420 * config/rs6000/rs6000-c.c: Add macro definitions for 27421 ALTIVEC_BUILTIN_VEC_PERMXOR. 27422 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin. 27423 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR. 27424 * config/rs6000/altivec.md (altivec_vpermxor): New define expand. 27425 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case 27426 UNSPEC_VPERMXOR. 27427 * config/doc/extend.texi: Add prototypes for vec_permxor. 27428 274292018-03-14 David Malcolm <dmalcolm@redhat.com> 27430 27431 PR c/84852 27432 * diagnostic-show-locus.c (class layout_point): Convert m_line 27433 from int to linenum_type. 27434 (line_span::comparator): Use linenum "compare" function when 27435 comparing line numbers. 27436 (test_line_span): New function. 27437 (layout_range::contains_point): Convert param "row" from int to 27438 linenum_type. 27439 (layout_range::intersects_line_p): Likewise. 27440 (layout::will_show_line_p): Likewise. 27441 (layout::print_source_line): Likewise. 27442 (layout::should_print_annotation_line_p): Likewise. 27443 (layout::print_annotation_line): Likewise. 27444 (layout::print_leading_fixits): Likewise. 27445 (layout::annotation_line_showed_range_p): Likewise. 27446 (struct line_corrections): Likewise for field m_row. 27447 (line_corrections::line_corrections): Likewise for param "row". 27448 (layout::print_trailing_fixits): Likewise. 27449 (layout::get_state_at_point): Likewise. 27450 (layout::get_x_bound_for_row): Likewise. 27451 (layout::print_line): Likewise. 27452 (diagnostic_show_locus): Likewise for locals "last_line" and "row". 27453 (selftest::diagnostic_show_locus_c_tests): Call test_line_span. 27454 * input.c (selftest::test_linenum_comparisons): New function. 27455 (selftest::input_c_tests): Call it. 27456 * selftest.c (selftest::test_assertions): Test ASSERT_GT, 27457 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT. 27458 * selftest.h (ASSERT_GT): New macro. 27459 (ASSERT_GT_AT): New macro. 27460 (ASSERT_LT): New macro. 27461 (ASSERT_LT_AT): New macro. 27462 274632018-03-14 Segher Boessenkool <segher@kernel.crashing.org> 27464 27465 PR rtl-optimization/84780 27466 * combine.c (distribute_links): Don't make a link based on pc_rtx. 27467 274682018-03-14 Martin Liska <mliska@suse.cz> 27469 27470 * tree.c (record_node_allocation_statistics): Use 27471 get_stats_node_kind. 27472 (get_stats_node_kind): New function extracted from 27473 record_node_allocation_statistics. 27474 (free_node): Use get_stats_node_kind. 27475 274762018-03-14 Richard Biener <rguenther@suse.de> 27477 27478 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts 27479 that the value-set of ANTIC_IN doesn't grow. 27480 27481 Revert 27482 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs 27483 member. 27484 (BB_VISITED_WITH_VISITED_SUCCS): New define. 27485 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS. 27486 274872018-03-14 Julia Koval <julia.koval@intel.com> 27488 27489 * config.gcc (icelake-client, icelake-server): New. 27490 (icelake): Remove. 27491 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit. 27492 (initial_ix86_arch_features): Ditto. 27493 (PTA_SKYLAKE): Add SGX. 27494 (PTA_ICELAKE): Remove. 27495 (PTA_ICELAKE_CLIENT): New. 27496 (PTA_ICELAKE_SERVER): New. 27497 (ix86_option_override_internal): Split up icelake on icelake client and 27498 icelake server. 27499 (get_builtin_code_for_version): Ditto. 27500 (fold_builtin_cpu): Ditto. 27501 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto. 27502 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto 27503 * config/i386/i386.h (processor_type): Ditto. 27504 * doc/invoke.texi: Ditto. 27505 275062018-03-14 Jakub Jelinek <jakub@redhat.com> 27507 27508 PR sanitizer/83392 27509 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also 27510 INTEGER_CST offset, add it together with bitpos / 8 and 27511 sign extend based on POINTER_SIZE. 27512 27513 PR target/84844 27514 Revert 27515 2017-04-20 Uros Bizjak <ubizjak@gmail.com> 27516 27517 PR target/78090 27518 * config/i386/constraints.md (Yc): New register constraint. 27519 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed): 27520 Use Yc constraint for alternative 2 of operand 0. Remove 27521 preferred_for_speed attribute. 27522 275232018-03-14 Richard Biener <rguenther@suse.de> 27524 27525 PR tree-optimization/84830 27526 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN 27527 with the old one to avoid oscillations. 27528 275292018-03-13 Vladimir Makarov <vmakarov@redhat.com> 27530 27531 PR target/83712 27532 * lra-assigns.c (find_all_spills_for): Ignore uninteresting 27533 pseudos. 27534 (assign_by_spills): Return a flag of reload assignment failure. 27535 Do not process the reload assignment failures. Do not spill other 27536 reload pseudos if they has the same reg class. Update n if 27537 necessary. 27538 (lra_assign): Add a return arg. Set up from the result of 27539 assign_by_spills call. 27540 (find_reload_regno_insns, lra_split_hard_reg_for): New functions. 27541 * lra-constraints.c (split_reg): Add a new arg. Use it instead of 27542 usage_insns if it is not NULL. 27543 (spill_hard_reg_in_range): New function. 27544 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg. 27545 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New 27546 function prototypes. 27547 (lra_assign): Change prototype. 27548 * lra.c (lra): Add code to deal with fails by splitting hard reg 27549 live ranges. 27550 275512018-03-01 Palmer Dabbelt <palmer@sifive.com> 27552 27553 * config/riscv/riscv.opt (mrelax): New option. 27554 * config/riscv/riscv.c (riscv_file_start): Emit ".option 27555 "norelax" when riscv_mrelax is disabled. 27556 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax". 27557 275582018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com> 27559 27560 PR target/84743 27561 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel 27562 reassociation for int modes. 27563 275642018-03-13 Richard Sandiford <richard.sandiford@linaro.org> 27565 27566 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks): 27567 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR 27568 for big-endian. 27569 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute. 27570 * config/aarch64/aarch64-sve.md 27571 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to... 27572 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this. 27573 (*extend<mode><Vwide>2): Rename to... 27574 (aarch64_sve_extend<mode><Vwide>2): ...this. 27575 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand, 27576 renaming the old pattern to... 27577 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define 27578 unsigned packs. 27579 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a 27580 define_expand, renaming the old pattern to... 27581 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this. 27582 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete. 27583 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into 27584 account when deciding which SVE instruction the optab should use. 27585 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise. 27586 275872018-03-13 Richard Sandiford <richard.sandiford@linaro.org> 27588 27589 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM) 27590 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM) 27591 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM) 27592 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants. 27593 (tlsdesc_small_<mode>): Turn a define_expand and use 27594 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to... 27595 (tlsdesc_small_advsimd_<mode>): ...this. 27596 (tlsdesc_small_sve_<mode>): New pattern. 27597 275982018-03-13 Richard Sandiford <richard.sandiford@linaro.org> 27599 27600 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART) 27601 (UNSPEC_UMUL_HIGHPART): New constants. 27602 (MUL_HIGHPART): New int iteraor. 27603 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART. 27604 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New 27605 define_expand. 27606 (*<su>mul<mode>3_highpart): New define_insn. 27607 276082018-03-13 Eric Botcazou <ebotcazou@adacore.com> 27609 27610 PR lto/84805 27611 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of 27612 incomplete types. 27613 276142018-03-13 Martin Liska <mliska@suse.cz> 27615 27616 PR ipa/84658. 27617 * (sem_item_optimizer::sem_item_optimizer): Initialize new 27618 vector. 27619 (sem_item_optimizer::~sem_item_optimizer): Release it. 27620 (sem_item_optimizer::merge_classes): Register variable aliases. 27621 (sem_item_optimizer::fixup_pt_set): New function. 27622 (sem_item_optimizer::fixup_points_to_sets): Likewise. 27623 * ipa-icf.h: Declare new variables and functions. 27624 276252018-03-13 Jakub Jelinek <jakub@redhat.com> 27626 27627 PR middle-end/84834 27628 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of 27629 integer_pow2p@2 and test integer_pow2p in condition. 27630 (A < 0 ? C : 0): Similarly for @1. 27631 27632 PR middle-end/84831 27633 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p) 27634 characters starting at p contain '\0' character, don't look beyond 27635 that. 27636 27637 PR target/84827 27638 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable 27639 pattern if -ftrapping-math -fno-fp-int-builtin-inexact. 27640 27641 PR target/84828 27642 * reg-stack.c (change_stack): Change update_end var from int to 27643 rtx_insn *, if non-NULL don't update just BB_END (current_block), but 27644 also call set_block_for_insn on the newly added insns and rescan. 27645 27646 PR target/84786 27647 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v 27648 on the last operand. 27649 27650 PR c++/84704 27651 * tree.c (stabilize_reference_1): Return save_expr (e) for 27652 STATEMENT_LIST even if it doesn't have side-effects. 27653 276542018-03-12 Jonathan Wakely <jwakely@redhat.com> 27655 27656 * doc/invoke.texi (-mclflushopt): Fix spelling of option. 27657 276582018-03-12 Renlin Li <renlin.li@arm.com> 27659 27660 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to 27661 aarch64_output_scalar_simd_mov_immediate. 27662 276632018-03-12 Martin Sebor <msebor@redhat.com> 27664 27665 PR tree-optimization/83456 27666 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning 27667 for perfectly overlapping calls to memcpy. 27668 (gimple_fold_builtin_memory_chk): Same. 27669 (gimple_fold_builtin_strcpy): Handle no-warning. 27670 (gimple_fold_builtin_stxcpy_chk): Same. 27671 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning. 27672 276732018-03-12 Segher Boessenkool <segher@kernel.crashing.org> 27674 27675 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named" 27676 parameter. Use it for SFmode. 27677 (rs6000_function_arg_advance_1): Adjust. 27678 (rs6000_function_arg): Adjust. 27679 (rs6000_gimplify_va_arg): Pass false for that new parameter. 27680 276812018-03-12 Segher Boessenkool <segher@kernel.crashing.org> 27682 27683 PR rtl-optimization/84169 27684 PR rtl-optimization/84780 27685 * combine.c (can_combine_p): Check for a 2-insn combination whether 27686 the destination register is used between the two insns, too. 27687 276882018-03-12 Richard Biener <rguenther@suse.de> 27689 27690 PR tree-optimization/84803 27691 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything 27692 for refs DR analysis didn't process. 27693 276942018-03-12 Richard Biener <rguenther@suse.de> 27695 27696 PR tree-optimization/84777 27697 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For 27698 force-vectorize loops ignore whether we are optimizing for size. 27699 277002018-03-12 Chung-Ju Wu <jasonwucj@gmail.com> 27701 27702 * config/nds32/nds32.c (nds32_md_asm_adjust): New function. 27703 (TARGET_MD_ASM_ADJUST): Define. 27704 277052018-03-12 Monk Chiang <sh.chiang04@gmail.com> 27706 Kito Cheng <kito.cheng@gmail.com> 27707 Chung-Ju Wu <jasonwucj@gmail.com> 27708 27709 * config/nds32/nds32.c (nds32_compute_stack_frame, 27710 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple, 27711 nds32_emit_stack_v3push, nds32_emit_stack_v3pop, 27712 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue, 27713 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine. 27714 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM, 27715 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New. 27716 * config/nds32/nds32.md (prologue, epilogue): Use macro 27717 NDS32_V3PUSH_AVAILABLE_P to do checking. 27718 277192018-03-11 Jakub Jelinek <jakub@redhat.com> 27720 27721 PR debug/58150 27722 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding 27723 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf, 27724 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating 27725 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard 27726 addition of most attributes on !orig_type_die or the attribute not 27727 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE. 27728 277292018-03-11 Kito Cheng <kito.cheng@gmail.com> 27730 Chung-Ju Wu <jasonwucj@gmail.com> 27731 27732 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define 27733 __NDS32_VH__ macro. 27734 * config/nds32/nds32.opt (mvh): New option. 27735 277362018-03-11 Kito Cheng <kito.cheng@gmail.com> 27737 Chung-Ju Wu <jasonwucj@gmail.com> 27738 27739 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare 27740 function. 27741 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function. 27742 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its 27743 definition. 27744 277452018-03-11 Kito Cheng <kito.cheng@gmail.com> 27746 Chung-Ju Wu <jasonwucj@gmail.com> 27747 27748 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New 27749 function. 27750 * config/nds32/nds32-multiple.md (strlensi): New pattern. 27751 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function. 27752 277532018-03-11 Monk Chiang <sh.chiang04@gmail.com> 27754 Kito Cheng <kito.cheng@gmail.com> 27755 Chung-Ju Wu <jasonwucj@gmail.com> 27756 27757 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB, 27758 UNSPEC_FFMISM and UNSPEC_FLMISM. 27759 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description 27760 for ffb, ffmism and flmism. 27761 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern. 27762 (unspec_ffmism): Ditto. 27763 (unspec_flmism): Ditto. 27764 (nds32_expand_builtin_impl): Check if string extension is available. 27765 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB, 27766 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM. 27767 277682018-03-10 Vladimir Makarov <vmakarov@redhat.com> 27769 27770 Reverting patch: 27771 2018-03-09 Vladimir Makarov <vmakarov@redhat.com> 27772 27773 PR target/83712 27774 * lra-assigns.c (assign_by_spills): Return a flag of reload 27775 assignment failure. Do not process the reload assignment 27776 failures. Do not spill other reload pseudos if they has the same 27777 reg class. 27778 (lra_assign): Add a return arg. Set up from the result of 27779 assign_by_spills call. 27780 (find_reload_regno_insns, lra_split_hard_reg_for): New functions. 27781 * lra-constraints.c (split_reg): Add a new arg. Use it instead of 27782 usage_insns if it is not NULL. 27783 (spill_hard_reg_in_range): New function. 27784 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg. 27785 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New 27786 function prototypes. 27787 (lra_assign): Change prototype. 27788 * lra.c (lra): Add code to deal with fails by splitting hard reg 27789 live ranges. 27790 277912018-03-10 H.J. Lu <hongjiu.lu@intel.com> 27792 27793 PR target/84807 27794 * config/i386/i386.opt: Replace Enforcment with Enforcement. 27795 277962018-03-10 Alexandre Oliva <aoliva@redhat.com> 27797 27798 PR debug/84620 27799 * dwarf2out.h (dw_val_class): Add dw_val_class_symview. 27800 (dw_val_node): Add val_symbolic_view. 27801 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset. 27802 (symview_upper_bound): New. 27803 (new_line_info_table): Initialize symviews_since_reset. 27804 (dwarf2out_source_line): Count symviews_since_reset and set 27805 symview_upper_bound. 27806 (dw_val_equal_p): Handle symview. 27807 (add_AT_symview): New. 27808 (print_dw_val): Handle symview. 27809 (attr_checksum, attr_checksum_ordered): Likewise. 27810 (same_dw_val_p, size_of_die): Likewise. 27811 (value_format, output_die): Likewise. 27812 (add_high_low_attributes): Use add_AT_symview for entry_view. 27813 (dwarf2out_finish): Reset symview_upper_bound, clear 27814 zero_view_p. 27815 278162018-03-09 Peter Bergner <bergner@vnet.ibm.com> 27817 27818 PR target/83969 27819 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype. 27820 Add strict argument and use it. 27821 (rs6000_split_multireg_move): Update for new strict argument. 27822 (mem_operand_gpr): Disallow all non-offsettable addresses. 27823 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint. 27824 278252018-03-09 Jakub Jelinek <jakub@redhat.com> 27826 27827 PR target/84772 27828 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp 27829 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY. 27830 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise. 27831 27832 PR c++/84767 27833 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped 27834 decl, use remap_type if we want to use the type. 27835 278362018-03-09 Martin Sebor <msebor@redhat.com> 27837 27838 PR tree-optimization/84526 27839 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset): 27840 Remove dead code. 27841 (builtin_access::generic_overlap): Be prepared to handle non-array 27842 base objects. 27843 278442018-03-09 Alexandre Oliva <aoliva@redhat.com> 27845 27846 PR rtl-optimization/84682 27847 * lra-constraints.c (process_address_1): Check is_address flag 27848 for address constraints. 27849 (process_alt_operands): Likewise. 27850 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to 27851 preprocess_constraints. 27852 * recog.h (preprocess_constraints): Add oploc parameter. 27853 Adjust callers. 27854 * recog.c (preprocess_constraints): Test address_operand for 27855 CT_ADDRESS constraints. 27856 278572018-03-09 Vladimir Makarov <vmakarov@redhat.com> 27858 27859 PR target/83712 27860 * lra-assigns.c (assign_by_spills): Return a flag of reload 27861 assignment failure. Do not process the reload assignment 27862 failures. Do not spill other reload pseudos if they has the same 27863 reg class. 27864 (lra_assign): Add a return arg. Set up from the result of 27865 assign_by_spills call. 27866 (find_reload_regno_insns, lra_split_hard_reg_for): New functions. 27867 * lra-constraints.c (split_reg): Add a new arg. Use it instead of 27868 usage_insns if it is not NULL. 27869 (spill_hard_reg_in_range): New function. 27870 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg. 27871 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New 27872 function prototypes. 27873 (lra_assign): Change prototype. 27874 * lra.c (lra): Add code to deal with fails by splitting hard reg 27875 live ranges. 27876 278772018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 27878 27879 PR target/83193 27880 * common/config/arm/arm-common.c (arm_parse_arch_option_name): 27881 Accept complain bool parameter. Only emit errors if it is true. 27882 (arm_parse_cpu_option_name): Likewise. 27883 (arm_target_thumb_only): Adjust callers of the above. 27884 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust 27885 prototype to take a default true bool parameter. 27886 (arm_parse_arch_option_name): Likewise. 27887 278882018-03-09 David Malcolm <dmalcolm@redhat.com> 27889 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org> 27890 27891 PR jit/64089 27892 PR jit/84288 27893 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New. 27894 * configure: Regenerate. 27895 * configure.ac ("linker --version-script option"): New. 27896 ("linker soname option"): New. 27897 278982018-03-09 Richard Biener <rguenther@suse.de> 27899 27900 PR tree-optimization/84775 27901 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink 27902 immediate uses of predicate stmts and mark them modified. 27903 27904 Revert 27905 PR tree-optimization/84178 27906 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates 27907 to caller. 27908 (version_loop_for_if_conversion): Delay update_ssa call. 27909 (tree_if_conversion): Delay update_ssa until after predicate 27910 insertion. 27911 279122018-03-09 Eric Botcazou <ebotcazou@adacore.com> 27913 27914 PR target/84763 27915 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation 27916 when the function accesses prior frames. 27917 279182018-03-08 Jakub Jelinek <jakub@redhat.com> 27919 27920 PR debug/84456 27921 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call 27922 gen_llsym, otherwise call maybe_gen_llsym. 27923 27924 PR inline-asm/84742 27925 * recog.c (asm_operand_ok): Return 0 if multi-character constraint 27926 has ',' character inside of it. 27927 279282018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 27929 27930 PR target/84748 27931 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern 27932 as clobbering CC_REGNUM. 27933 279342018-03-08 Richard Biener <rguenther@suse.de> 27935 27936 PR middle-end/84552 27937 * tree-scalar-evolution.c: Include tree-into-ssa.h. 27938 (follow_copies_to_constant): Do not follow SSA names registered 27939 for update. 27940 279412018-03-08 Richard Biener <rguenther@suse.de> 27942 27943 PR tree-optimization/84178 27944 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates 27945 to caller. 27946 (version_loop_for_if_conversion): Delay update_ssa call. 27947 (tree_if_conversion): Delay update_ssa until after predicate 27948 insertion. 27949 279502018-03-08 David Malcolm <dmalcolm@redhat.com> 27951 27952 PR tree-optimization/84178 27953 * tree-if-conv.c (release_bb_predicate): Remove the 27954 the assertion that the stmts have NULL use_ops. 27955 Discard the statements, asserting that they haven't 27956 yet been added to a BB. 27957 279582018-03-08 Richard Biener <rguenther@suse.de> 27959 27960 PR tree-optimization/84746 27961 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL. 27962 (phi_translate): Pass in destination ANTIC_OUT set. 27963 (phi_translate_1): Likewise. For a simplified result lookup 27964 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets. 27965 (phi_translate_set): Adjust. 27966 (do_pre_regular_insertion): Likewise. 27967 (do_pre_partial_partial_insertion): Likewise. 27968 279692018-03-08 Martin Liska <mliska@suse.cz> 27970 27971 PR gcov-profile/84735 27972 * doc/gcov.texi: Document usage of profile files. 27973 * gcov-io.h: Document changes in the format. 27974 279752018-03-08 Alexandre Oliva <aoliva@redhat.com> 27976 27977 PR debug/84404 27978 PR debug/84408 27979 * dwarf2out.c (struct dw_line_info_table): Update comments for 27980 view == -1. 27981 (FORCE_RESET_NEXT_VIEW): New. 27982 (FORCE_RESETTING_VIEW_P): New. 27983 (RESETTING_VIEW_P): Check for -1 too. 27984 (ZERO_VIEW_P): Likewise. 27985 (new_line_info_table): Force-reset next view. 27986 (dwarf2out_begin_function): Likewise. 27987 (dwarf2out_source_line): Simplify zero_view_p initialization. 27988 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of 27989 view directly. Omit view when omitting .loc at line 0. 27990 279912018-03-08 Jakub Jelinek <jakub@redhat.com> 27992 27993 PR tree-optimization/84740 27994 * tree-switch-conversion.c (process_switch): Call build_constructors 27995 only if info.phi_count is non-zero. 27996 27997 PR tree-optimization/84739 27998 * tree-tailcall.c (find_tail_calls): Check call arguments against 27999 DECL_ARGUMENTS (current_function_decl) rather than 28000 DECL_ARGUMENTS (func) when checking for tail recursion. 28001 280022018-03-07 Jakub Jelinek <jakub@redhat.com> 28003 28004 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm, 28005 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and 28006 Volker Reichelt's entry and add entries for people that perform 28007 GCC fuzzy testing and report numerous bugs. 28008 280092018-03-07 Segher Boessenkool <segher@kernel.crashing.org> 28010 28011 PR target/82411 28012 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put 28013 readonly data in sdata, if that is disabled. 28014 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option. 28015 * doc/invoke.texi (RS/6000 and PowerPC Options): Document 28016 -mreadonly-in-sdata option. 28017 280182018-03-07 Martin Sebor <msebor@redhat.com> 28019 28020 PR tree-optimization/84468 28021 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor 28022 basic block when looking for nul assignment. 28023 280242018-03-07 Eric Botcazou <ebotcazou@adacore.com> 28025 28026 PR target/84277 28027 * except.h (output_function_exception_table): Adjust prototype. 28028 * except.c (output_function_exception_table): Remove FNNAME parameter 28029 and add SECTION parameter. Ouput one part of the table at a time. 28030 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output 28031 the first part of the exception table and emit unwind directives. 28032 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare. 28033 (i386_pe_seh_cold_init): Likewise. 28034 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro. 28035 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise. 28036 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment. 28037 (ix86_output_call_insn): Emit a nop in one more case for SEH. 28038 * config/i386/winnt.c: Include except.h. 28039 (struct seh_frame_state): Add reg_offset, after_prologue and 28040 in_cold_section fields. 28041 (i386_pe_seh_end_prologue): Set seh->after_prologue. 28042 (i386_pe_seh_cold_init): New function. 28043 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal 28044 to seh->in_cold_section. 28045 (seh_emit_push): Record the offset of the push. 28046 (seh_emit_save): Record the offet of the save. 28047 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS. 28048 Test seh->after_prologue to disregard the epilogue. 28049 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini. 28050 (i386_pe_end_cold_function): New function. 28051 280522018-03-07 Jakub Jelinek <jakub@redhat.com> 28053 28054 PR fortran/84565 28055 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use 28056 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero. 28057 28058 PR c++/84704 28059 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag 28060 on tmp_var. 28061 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG, 28062 don't print names of DECL_NAMELESS DECL_IGNORED_P decls. 28063 28064 PR middle-end/84723 28065 * multiple_target.c: Include tree-inline.h and intl.h. 28066 (expand_target_clones): Diagnose and fail if node->definition and 28067 !tree_versionable_function_p (node->decl). 28068 280692018-03-06 John David Anglin <danglin@gcc.gnu.org> 28070 28071 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use 28072 sprint_ul. 28073 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change. 28074 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise. 28075 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above. 28076 280772018-03-06 Jakub Jelinek <jakub@redhat.com> 28078 28079 PR target/84710 28080 * combine.c (try_combine): Use reg_or_subregno instead of handling 28081 just paradoxical SUBREGs and REGs. 28082 280832018-03-06 Claudiu Zissulescu <claziss@synopsys.com> 28084 28085 * config/arc/arc.c (arc_finalize_pic): Remove function. 28086 (arc_must_save_register): We use single base PIC register, remove 28087 checks to save/restore the PIC register. 28088 (arc_expand_prologue): Likewise. 28089 * config/arc/arc-protos.h (arc_set_default_type_attributes): 28090 Remove. 28091 (arc_verify_short): Likewise. 28092 (arc_attr_type): Likewise. 28093 * config/arc/arc.c (arc_set_default_type_attributes): Remove. 28094 (walk_stores): Likewise. 28095 (arc_address_cost): Make it static. 28096 (arc_verify_short): Likewise. 28097 (branch_dest): Likewise. 28098 (arc_attr_type): Likewise. 28099 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove. 28100 (TARGET_INSN_LENGTH_PARAMETERS): Likewise. 28101 (arc_final_prescan_insn): Remove inserting the nops due to 28102 hardware hazards. It is done in reorg step. 28103 (insn_length_variant_t): Remove. 28104 (insn_length_parameters_t): Likewise. 28105 (arc_insn_length_parameters): Likewise. 28106 (arc_get_insn_variants): Likewise. 28107 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove. 28108 281092018-03-06 Jakub Jelinek <jakub@redhat.com> 28110 28111 PR inline-asm/84683 28112 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid 28113 assertion failure. 28114 28115 PR tree-optimization/84687 28116 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS 28117 on new_node->decl. 28118 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization. 28119 281202018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com> 28121 28122 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier): 28123 Rename to ppc_speculation_barrier. 28124 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to 28125 __builtin_ppc_speculation_barrier. 28126 281272018-03-05 Jakub Jelinek <jakub@redhat.com> 28128 28129 PR target/84700 28130 * combine.c (combine_simplify_rtx): Don't try to simplify if 28131 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx 28132 are equal to x. 28133 281342018-03-05 Segher Boessenkool <segher@kernel.crashing.org> 28135 28136 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops 28137 to 32 bytes when compiling for POWER9. 28138 281392018-03-05 Jakub Jelinek <jakub@redhat.com> 28140 28141 PR target/84564 28142 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for 28143 regparm >= 3 with no arg reg available also for calls with 28144 flag_force_indirect_call. Pass decl to ix86_function_regparm. 28145 28146 PR target/84524 28147 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with 28148 orig,vex. 28149 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses. 28150 281512018-03-05 Peter Bergner <bergner@vnet.ibm.com> 28152 28153 PR target/84264 28154 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands. 28155 281562018-03-05 Richard Biener <rguenther@suse.de> 28157 28158 PR tree-optimization/84486 28159 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code. 28160 When inserting a __builtin_assume_aligned call set the LHS 28161 SSA name alignment info accordingly. 28162 281632018-03-05 Wilco Dijkstra <wdijkstr@arm.com> 28164 28165 PR tree-optimization/84114 28166 * config/aarch64/aarch64.c (aarch64_reassociation_width) 28167 Avoid reassociation of FLOAT_MODE addition. 28168 281692018-03-05 Olga Makhotina <olga.makhotina@intel.com> 28170 28171 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET, 28172 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET, 28173 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions. 28174 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd. 28175 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers. 28176 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New. 28177 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig 28178 and -mwbnoinvd. 28179 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd, 28180 __builtin_ia32_wbinvd): New builtins. 28181 (SPECIAL_ARGS2): New. 28182 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New. 28183 (SPECIAL_ARGS2): New. 28184 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd. 28185 (ix86_valid_target_attribute_inner_p): Ditto. 28186 (ix86_init_mmx_sse_builtins): Add special_args2. 28187 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD, 28188 TARGET_WBNOINVD_P): New. 28189 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New. 28190 (define_insn "wbinvd", define_insn "wbnoinvd"): New. 28191 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd. 28192 * config/i386/immintrin.h (_wbinvd): New intrinsic. 28193 * config/i386/pconfigintrin.h: New file. 28194 * config/i386/wbnoinvdintrin.h: Ditto. 28195 * config/i386/x86intrin.h: Add headers pconfigintrin.h and 28196 wbnoinvdintrin.h. 28197 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New. 28198 281992018-03-05 Richard Biener <rguenther@suse.de> 28200 28201 PR tree-optimization/84670 28202 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs 28203 member. 28204 (BB_VISITED_WITH_VISITED_SUCCS): New define. 28205 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS. 28206 (compute_antic_aux): Only assert the number of values in ANTIC_IN 28207 doesn't grow if all successors (recursively) were visited at least 28208 once. 28209 282102018-03-05 Richard Biener <rguenther@suse.de> 28211 28212 PR tree-optimization/84650 28213 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache 28214 if executed in the loop pipeline. 28215 282162018-03-05 Sandra Loosemore <sandra@codesourcery.com> 28217 28218 * doc/configfiles.texi (Configuration Files): Move info about 28219 conditionalizing $target-protos.h to... 28220 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h 28221 differs from $target-protos.h. 28222 282232018-03-05 Kito Cheng <kito.cheng@gmail.com> 28224 Chung-Ju Wu <jasonwucj@gmail.com> 28225 28226 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare. 28227 * config/nds32/nds32-multiple.md (setmemsi): Define. 28228 * config/nds32/nds32-memory-manipulation.c 28229 (nds32_gen_dup_4_byte_to_word_value): New. 28230 (emit_setmem_word_loop): New. 28231 (emit_setmem_byte_loop): New. 28232 (nds32_expand_setmem_loop): New. 28233 (nds32_expand_setmem_loop_v3m): New. 28234 (nds32_expand_setmem_unroll): New. 28235 (nds32_expand_setmem): New. 28236 282372018-03-04 Kito Cheng <kito.cheng@gmail.com> 28238 Chung-Ju Wu <jasonwucj@gmail.com> 28239 28240 * config/nds32/nds32-memory-manipulation.c 28241 (nds32_emit_load_store): New. 28242 (nds32_emit_post_inc_load_store): New. 28243 (nds32_emit_mem_move): New. 28244 (nds32_emit_mem_move_block): New. 28245 (nds32_expand_movmemsi_loop_unknown_size): New. 28246 (nds32_expand_movmemsi_loop_known_size): New. 28247 (nds32_expand_movmemsi_loop): New. 28248 (nds32_expand_movmemsi_unroll): New. 28249 (nds32_expand_movmemqi): Rename ... 28250 (nds32_expand_movmemsi): ... to this. 28251 * config/nds32/nds32-multiple.md (movmemqi): Rename ... 28252 (movmemsi): ... to this. 28253 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ... 28254 (nds32_expand_movmemsi): ... to this. 28255 282562018-03-04 Kito Cheng <kito.cheng@gmail.com> 28257 Monk Chiang <sh.chiang04@gmail.com> 28258 Chung-Ju Wu <jasonwucj@gmail.com> 28259 28260 * config/nds32/nds32-protos.h 28261 (nds32_expand_load_multiple): New arguments. 28262 (nds32_expand_store_multiple): Ditto. 28263 (nds32_valid_multiple_load_store): Rename ... 28264 (nds32_valid_multiple_load_store_p): ... to this. 28265 * config/nds32/nds32-memory-manipulation.c 28266 (nds32_expand_load_multiple): Refine implementation. 28267 (nds32_expand_store_multiple): Ditto. 28268 * config/nds32/nds32-multiple.md 28269 (load_multiple): Update nds32_expand_load_multiple interface. 28270 (store_multiple): Update nds32_expand_store_multiple interface. 28271 * config/nds32/nds32-predicates.c 28272 (nds32_valid_multiple_load_store): Rename ... 28273 (nds32_valid_multiple_load_store_p): ... to this and refine 28274 implementation. 28275 * config/nds32/predicates.md 28276 (nds32_load_multiple_and_update_address_operation): New predicate. 28277 (nds32_store_multiple_and_update_address_operation): New predicate. 28278 282792018-03-04 Kito Cheng <kito.cheng@gmail.com> 28280 Chung-Ju Wu <jasonwucj@gmail.com> 28281 28282 * config/nds32/nds32.md (type): Add load_multiple and store_multiple. 28283 (combo): New attribute. 28284 * config/nds32/nds32-multiple.md: Refine patterns with new attributes. 28285 282862018-03-03 Chung-Ju Wu <jasonwucj@gmail.com> 28287 28288 * config/nds32/nds32.opt: Change -mcmodel= default value. 28289 282902018-03-03 Kito Cheng <kito.cheng@gmail.com> 28291 Monk Chiang <sh.chiang04@gmail.com> 28292 Chung-Ju Wu <jasonwucj@gmail.com> 28293 28294 * config/nds32/constants.md (unspec_element): New enum. 28295 * config/nds32/constraints.md (Umw): New constraint. 28296 * config/nds32/nds32-intrinsic.c: Add more builtin functions. 28297 * config/nds32/nds32-intrinsic.md: Likewise. 28298 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New. 28299 (nds32_valid_smw_lwm_base_p): New. 28300 (nds32_output_smw_single_word): New. 28301 (nds32_output_lmw_single_word): New. 28302 (nds32_expand_unaligned_load): New. 28303 (nds32_expand_unaligned_store): New. 28304 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare. 28305 (nds32_output_smw_single_word): Declare. 28306 (nds32_output_lmw_single_word): Declare. 28307 (nds32_expand_unaligned_load): Declare. 28308 (nds32_expand_unaligned_store): Declare. 28309 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW, 28310 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW, 28311 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W, 28312 NDS32_BUILTIN_UASTORE_DW. 28313 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New 28314 predicate. 28315 283162018-03-03 Monk Chiang <sh.chiang04@gmail.com> 28317 Kito Cheng <kito.cheng@gmail.com> 28318 Chung-Ju Wu <jasonwucj@gmail.com> 28319 28320 * config/nds32/nds32-intrinsic.c 28321 (nds32_expand_builtin_null_ftype_reg): Delete. 28322 (nds32_expand_builtin_reg_ftype_imm): Ditto. 28323 (nds32_expand_builtin_null_ftype_reg_imm): Ditto. 28324 (nds32_read_argument): New. 28325 (nds32_legitimize_target): Ditto. 28326 (nds32_legitimize_argument): Ditto. 28327 (nds32_check_constant_argument): Ditto. 28328 (nds32_expand_unop_builtin): Ditto. 28329 (nds32_expand_unopimm_builtin): Ditto. 28330 (nds32_expand_binop_builtin): Ditto. 28331 (nds32_builtin_decl_impl): Ditto. 28332 (builtin_description): Ditto. 28333 (nds32_expand_builtin_impl): Rewrite with new infrastructure. 28334 (nds32_init_builtins_impl): Ditto. 28335 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define. 28336 (nds32_builtin_decl): New. 28337 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT. 28338 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare. 28339 283402018-03-02 Jeff Law <law@redhat.com> 28341 28342 * reorg.c (stop_search_p): Handle DEBUG_INSN. 28343 (redundant_insn, fill_simple_delay_slots): Likewise. 28344 (fill_slots_from_thread): Likewise. 28345 * resource.c (mark_referenced_resources): Likewise. 28346 (mark_set_resources, find_dead_or_set_registers): Likewise. 28347 283482018-03-02 Jakub Jelinek <jakub@redhat.com> 28349 28350 * substring-locations.h (format_warning_va): Formatting fix for 28351 ATTRIBUTE_GCC_DIAG. 28352 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second 28353 argument. 28354 (format_warning_n_va, format_warning_at_substring_n): New prototypes. 28355 * substring-locations.c: Include intl.h. 28356 (format_warning_va): Turned into small wrapper around 28357 format_warning_n_va, renamed to ... 28358 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments, 28359 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID, 28360 use ngettext. 28361 (format_warning_at_substring_n): New function. 28362 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma. 28363 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of 28364 format_warning_at_substring with just a shorter name instead of 28365 const function pointer. 28366 (fmtwarn_n): New function. 28367 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where 28368 appropriate, get rid of all the fmtstr temporaries, move conditionals 28369 with G_() wrapped string literals directly into fmtwarn arguments, 28370 cast dir.len to (int), formatting fixes. 28371 283722018-03-02 Thomas Schwinge <thomas@codesourcery.com> 28373 28374 * doc/invoke.texi: Remove "Cilk Plus" references. 28375 283762018-03-02 Jakub Jelinek <jakub@redhat.com> 28377 Richard Biener <rguenther@suse.de> 28378 28379 PR ipa/84628 28380 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics 28381 for error or warning attributes if CALL_FROM_THUNK_P is set. 28382 Formatting fixes. 28383 283842018-03-02 Jakub Jelinek <jakub@redhat.com> 28385 28386 PR target/56540 28387 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine 28388 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY. 28389 28390 PR target/56540 28391 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine 28392 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros. 28393 28394 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED 28395 instead of -1U in last predictors element's probability member. 28396 283972018-03-02 Eric Botcazou <ebotcazou@adacore.com> 28398 28399 PR ipa/83983 28400 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both 28401 arguments if they are comparable. 28402 284032018-03-02 Richard Sandiford <richard.sandiford@linaro.org> 28404 28405 PR tree-optimization/84634 28406 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace 28407 masks and masked_loop_p with a single loop_masks, making sure it's 28408 null for bb vectorization. 28409 284102018-03-02 Richard Sandiford <richard.sandiford@linaro.org> 28411 28412 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence) 28413 (vect_analyze_data_ref_access): Use loop->safe_len rather than 28414 loop->force_vectorize to check whether there is no alias. 28415 284162018-03-02 Jakub Jelinek <jakub@redhat.com> 28417 28418 PR target/84614 28419 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New 28420 prototypes. 28421 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function 28422 comments. 28423 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions. 28424 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn 28425 instead of a loop around prev_real_insn. 28426 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of 28427 prev_real_insn. 28428 28429 PR inline-asm/84625 28430 * config/i386/i386.c (ix86_print_operand): Use conditional 28431 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not 28432 zero vector. 28433 284342018-03-02 Richard Biener <rguenther@suse.de> 28435 28436 PR tree-optimization/84427 28437 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove. 28438 (bitmap_set_subtract_values): Rewrite to handle multiple 28439 exprs per value. 28440 (clean): Likewise. 28441 (prune_clobbered_mems): Likewise. 28442 (phi_translate): Take edge instead of pred/phiblock. 28443 (phi_translate_1): Likewise. 28444 (phi_translate_set): Likewise. Insert all translated 28445 exprs for a value into the set, keeping possibly multiple 28446 expressions per value. 28447 (compute_antic_aux): Adjust for phi_translate changes. 28448 When intersecting union the expressions and prune those 28449 not in the final value set, keeping possibly multiple 28450 expressions per value. Do not use value-insertion 28451 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge 28452 all expressions. Add verification that the value-sets 28453 only shrink during iteration. 28454 (compute_partial_antic_aux): Adjust for the phi_translate changes. 28455 (do_pre_regular_insertion): Likewise. 28456 (do_pre_partial_partial_insertion): Likewise. 28457 284582018-03-02 Richard Biener <rguenther@suse.de> 28459 28460 PR target/82005 28461 * config/darwin.c (saved_debug_info_level): New static global. 28462 (darwin_asm_lto_start): Disable debug info generation for LTO out. 28463 (darwin_asm_lto_end): Restore debug info generation settings. 28464 284652018-03-01 Martin Liska <mliska@suse.cz> 28466 28467 PR sanitizer/82484 28468 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle 28469 volatile arguments. 28470 284712018-03-01 Richard Biener <rguenther@suse.de> 28472 28473 PR debug/84645 28474 * dwarf2out.c (gen_variable_die): Properly handle late VLA 28475 type annotation with LTO when debug was disabled at compile-time. 28476 284772018-03-01 Matthew Fortune <mfortune@gmail.com> 28478 28479 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect 28480 XINT with INTVAL. 28481 (mips_final_postscan_insn): Likewise. 28482 284832018-03-01 Richard Sandiford <richard.sandiford@linaro.org> 28484 28485 PR rtl-optimization/84528 28486 * alias.c (init_alias_target): Add commentary. 28487 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM 28488 a unique base value if the frame pointer is not eliminated 28489 to the stack pointer. 28490 284912018-03-01 Tom de Vries <tom@codesourcery.com> 28492 28493 PR rtl-optimization/83327 28494 * lra-int.h (hard_regs_spilled_into): Declare. 28495 * lra.c (hard_regs_spilled_into): Define. 28496 (init_reg_info): Init hard_regs_spilled_into. 28497 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into. 28498 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead) 28499 (process_bb_lives): Handle hard_regs_spilled_into. 28500 (lra_create_live_ranges_1): Before doing liveness propagation, clear 28501 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into. 28502 285032018-02-28 David Edelsohn <dje.gcc@gmail.com> 28504 28505 * config.gcc (powerpc-ibm-aix7.1.*): New stanza. 28506 (powerpc-ibm-aix[789]*): Default to AIX 7.2. 28507 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA. 28508 * config/rs6000/aix72.h: New file. 28509 285102018-02-28 Jakub Jelinek <jakub@redhat.com> 28511 28512 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n 28513 instead of warning_at with conditional singular and plural messages 28514 where possible. 28515 28516 PR target/52991 28517 * stor-layout.c (update_alignment_for_field): For 28518 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield 28519 && !DECL_PACKED (field), do the alignment update, just use 28520 only desired_align instead of MAX (type_align, desired_align) 28521 as the alignment. 28522 (place_field): Don't do known_align < desired_align handling 28523 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field 28524 is non-NULL, instead do it after rli->prev_field handling and 28525 only if not within a bitfield word. For DECL_PACKED (field) 28526 use type_align of BITS_PER_UNIT. 28527 285282018-02-28 Eric Botcazou <ebotcazou@adacore.com> 28529 28530 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove 28531 superfluous parentheses and trailing spaces. 28532 285332018-02-28 Richard Biener <rguenther@suse.de> 28534 28535 PR tree-optimization/84584 28536 * graphite-scop-detection.c (scop_detection::add_scop): Discard 28537 SCoPs with fake exit edge. 28538 285392018-02-28 Martin Liska <mliska@suse.cz> 28540 28541 PR testsuite/84597 28542 * timevar.c (timer::print): Fix format to properly print 100% 28543 values. 28544 285452018-02-28 Richard Biener <rguenther@suse.de> 28546 28547 PR middle-end/84607 28548 * genmatch.c (capture_info::walk_match): Do not mark 28549 captured expressions without operands as expr_p given 28550 they act more like predicates and should be subject to 28551 "lost tail" side-effect preserving. 28552 285532018-02-28 Alexandre Oliva <aoliva@redhat.com> 28554 28555 PR rtl-optimization/81611 28556 * auto-inc-dec.c (attempt_change): Move dead note from 28557 mem_insn if it's the next use of regno 28558 (find_address): Take address use of reg holding 28559 non-incremented value. Add parm to limit search to the named 28560 reg only. 28561 (merge_in_block): Attempt to use a mem insn that is the next 28562 use of the original regno. 28563 285642018-02-27 Martin Sebor <msebor@redhat.com> 28565 28566 PR c++/83871 28567 * doc/invoke.texi (-Wmissing-attributes): New option. 28568 * print-tree.c (print_node): Handle DECL_UNINLINABLE. 28569 285702018-02-27 Martin Sebor <msebor@redhat.com> 28571 28572 PR translation/84207 28573 * diagnostic-core.h (warning_n, error_n, inform_n): Change 28574 n argument to unsigned HOST_WIDE_INT. 28575 * diagnostic.c (warning_n, error_n, inform_n): Ditto. 28576 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX. 28577 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call. 28578 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n. 28579 285802018-02-27 Richard Biener <rguenther@suse.de> 28581 28582 PR tree-optimization/84512 28583 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost): 28584 Do not use the estimate returned from record_stmt_cost for 28585 the scalar iteration cost but sum properly using add_stmt_cost. 28586 285872018-02-27 Richard Biener <rguenther@suse.de> 28588 28589 PR tree-optimization/84466 28590 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p): 28591 Adjust last change to less strictly validate use operands. 28592 285932018-02-27 Martin Liska <mliska@suse.cz> 28594 28595 PR gcov-profile/84548 28596 * gcov.c (process_file): Allow partial overlap and consider it 28597 also as group functions. 28598 (output_lines): Properly calculate range of lines for a group. 28599 286002018-02-27 Martin Liska <mliska@suse.cz> 28601 28602 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and 28603 'ggc' suffixes. Change first column width. 28604 (timer::print): Fix formatting of the column. 28605 286062018-02-27 Alexandre Oliva <aoliva@redhat.com> 28607 28608 * tree-ssa-live.c (remove_unused_scope_block_p): Do not 28609 preserve inline entry blocks for the sake of debug inline 28610 entry point markers alone. 28611 (remove_unused_locals): Suggest in comments a better place to 28612 force the preservation of inline entry blocks that are 28613 otherwise unused, but do not preserve them. 28614 286152018-02-26 H.J. Lu <hongjiu.lu@intel.com> 28616 28617 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments. 28618 286192018-02-26 H.J. Lu <hongjiu.lu@intel.com> 28620 28621 PR target/84039 28622 * config/i386/constraints.md (Bs): Replace 28623 ix86_indirect_branch_register with 28624 TARGET_INDIRECT_BRANCH_REGISTER. 28625 (Bw): Likewise. 28626 * config/i386/i386.md (indirect_jump): Likewise. 28627 (tablejump): Likewise. 28628 (*sibcall_memory): Likewise. 28629 (*sibcall_value_memory): Likewise. 28630 Peepholes of indirect call and jump via memory: Likewise. 28631 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER. 28632 (*sibcall_value_GOT_32): Likewise. 28633 * config/i386/predicates.md (indirect_branch_operand): Likewise. 28634 (GOT_memory_operand): Likewise. 28635 (call_insn_operand): Likewise. 28636 (sibcall_insn_operand): Likewise. 28637 (GOT32_symbol_operand): Likewise. 28638 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New. 28639 286402018-02-26 Eric Botcazou <ebotcazou@adacore.com> 28641 28642 PR rtl-optimization/83496 28643 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from 28644 booleans to RTXes. Call fix_reg_dead_note on every non-null element. 28645 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a 28646 redundant insn, if any. 28647 (relax_delay_slots): Likewise. 28648 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN. 28649 286502018-02-26 Richard Sandiford <richard.sandiford@linaro.org> 28651 28652 PR tree-optimization/83965 28653 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume 28654 that grouped statements are part of a reduction chain. Return 28655 true if the statement is not marked as a reduction itself but 28656 is part of a group. 28657 (vect_recog_dot_prod_pattern): Don't check whether the statement 28658 is part of a group here. 28659 (vect_recog_sad_pattern): Likewise. 28660 (vect_recog_widen_sum_pattern): Likewise. 28661 286622018-02-26 Eric Botcazou <ebotcazou@adacore.com> 28663 28664 PR debug/84545 28665 * final.c (rest_of_clean_state): Also look for calls inside sequences. 28666 286672018-02-26 H.J. Lu <hongjiu.lu@intel.com> 28668 28669 PR target/84530 28670 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove 28671 the bool argument. 28672 (ix86_output_indirect_function_return): New prototype. 28673 (ix86_split_simple_return_pop_internal): Likewise. 28674 * config/i386/i386.c (indirect_return_via_cx): New. 28675 (indirect_return_via_cx_bnd): Likewise. 28676 (indirect_thunk_name): Handle return va CX_REG. 28677 (output_indirect_thunk_function): Create alias for 28678 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd. 28679 (ix86_output_indirect_jmp): Remove the bool argument. 28680 (ix86_output_indirect_function_return): New function. 28681 (ix86_split_simple_return_pop_internal): Likewise. 28682 * config/i386/i386.md (*indirect_jump): Don't pass false 28683 to ix86_output_indirect_jmp. 28684 (*tablejump_1): Likewise. 28685 (simple_return_pop_internal): Change it to define_insn_and_split. 28686 Call ix86_split_simple_return_pop_internal to split it for 28687 -mfunction-return=. 28688 (simple_return_indirect_internal): Call 28689 ix86_output_indirect_function_return instead of 28690 ix86_output_indirect_jmp. 28691 286922018-02-26 Jakub Jelinek <jakub@redhat.com> 28693 28694 PR bootstrap/84405 28695 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use 28696 memset and value initialization afterwards. 28697 286982018-02-26 Christophe Lyon <christophe.lyon@linaro.org> 28699 28700 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS. 28701 287022018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com> 28703 28704 PR target/84521 28705 * common/config/aarch64/aarch64-common.c 28706 (aarch_option_optimization_table[]): Switch 28707 off fomit-frame-pointer 28708 287092018-02-26 Kito Cheng <kito.cheng@gmail.com> 28710 Chung-Ju Wu <jasonwucj@gmail.com> 28711 28712 * config/nds32/nds32-multiple.md (load_multiple): Disallow 28713 volatile memory. 28714 (store_multiple): Ditto. 28715 287162018-02-26 Kito Cheng <kito.cheng@gmail.com> 28717 28718 * config.gcc: Add --with-cpu support for nds32 target. 28719 * config/nds32/nds32-opts.h (nds32_cpu_type): New. 28720 * config/nds32/nds32.opt: Add -mcpu= option. 28721 287222018-02-25 Segher Boessenkool <segher@kernel.crashing.org> 28723 28724 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no, 28725 isel=yes): Warn for these deprecated options. 28726 287272018-02-23 David Edelsohn <dje.gcc@gmail.com> 28728 28729 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to 28730 ISA_2_5_MASKS_EMBEDDED. 28731 287322018-02-23 Jakub Jelinek <jakub@redhat.com> 28733 28734 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and 28735 p->max as pointers rather than using iterative_hash_expr. 28736 287372018-02-23 Carl Love <cel@us.ibm.com> 28738 28739 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2 28740 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to 28741 BU_P8V_OVERLOAD_2. 28742 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to 28743 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to 28744 P8V_BUILTIN_VEC_VUNSIGNED2. 28745 287462018-02-22 Vladimir Makarov <vmakarov@redhat.com> 28747 28748 PR target/81572 28749 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros. 28750 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use 28751 LRA_UNKNOWN_ALT. 28752 * lra-constraints.c (curr_insn_transform): Set up 28753 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use 28754 LRA_UNKNOWN_ALT. 28755 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT. 28756 * lra-eliminations.c (spill_pseudos): Ditto. 28757 (process_insn_for_elimination): Ditto. 28758 * lra-lives.c (reg_early_clobber_p): Use the new macros. 28759 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and 28760 LRA_NON_CLOBBERED_ALT. 28761 287622018-02-22 Martin Sebor <msebor@redhat.com> 28763 28764 PR tree-optimization/84480 28765 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings 28766 to maybe_diag_stxncpy_trunc. Call it. 28767 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings 28768 from gimple_fold_builtin_strcpy. Print inlining stack. 28769 (handle_builtin_stxncpy): Print inlining stack. 28770 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare. 28771 287722018-02-22 H.J. Lu <hongjiu.lu@intel.com> 28773 28774 PR target/84176 28775 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an 28776 error when -mindirect-branch=thunk-extern, -fcf-protection=branch 28777 and -fcheck-pointer-bounds are used together. 28778 (indirect_thunk_prefix): New enum. 28779 (indirect_thunk_need_prefix): New function. 28780 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use 28781 "_nt" instead of "_bnd" for NOTRACK prefix. 28782 (output_indirect_thunk): Replace need_bnd_p with need_prefix. 28783 (output_indirect_thunk_function): Likewise. 28784 (): Likewise. 28785 (ix86_code_end): Update output_indirect_thunk_function calls. 28786 (ix86_output_indirect_branch_via_reg): Replace 28787 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix. 28788 (ix86_output_indirect_branch_via_push): Likewise. 28789 (ix86_output_function_return): Likewise. 28790 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is 28791 incompatible with -fcf-protection=branch and 28792 -fcheck-pointer-bounds. 28793 287942018-02-22 Steve Ellcey <sellcey@cavium.com> 28795 28796 PR target/83335 28797 * config/aarch64/aarch64.c (aarch64_print_address_internal): 28798 Change gcc_assert call to output_operand_lossage. 28799 288002018-02-22 Steve Ellcey <sellcey@cavium.com> 28801 28802 * doc/extend.texi (__builtin_extend_pointer): Document builtin. 28803 288042018-02-22 DJ Delorie <dj@redhat.com> 28805 Sebastian Perta <sebastian.perta@renesas.com> 28806 Oleg Endo <olegendo@gcc.gnu.org> 28807 28808 * config/rx/rx.c (rx_rtx_costs): New function. 28809 (TARGET_RTX_COSTS): Override to use rx_rtx_costs. 28810 288112018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com> 28812 28813 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs. 28814 288152018-02-22 Martin Liska <mliska@suse.cz> 28816 28817 PR driver/83193 28818 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option): 28819 Add "native" as a possible value. 28820 288212018-02-22 Martin Liska <mliska@suse.cz> 28822 28823 PR driver/83193 28824 * config/i386/i386.c (ix86_option_override_internal): 28825 Add "native" as a possible value for -march and -mtune. 28826 288272018-02-22 Jakub Jelinek <jakub@redhat.com> 28828 28829 PR target/84502 28830 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag 28831 to all type variants. 28832 28833 PR tree-optimization/84503 28834 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute 28835 width as info->bitpos + info->bitsize - start. 28836 (merged_store_group::merge_overlapping): Simplify width computation. 28837 (check_no_overlap): New function. 28838 (imm_store_chain_info::try_coalesce_bswap): Compute expected 28839 start + width and last_order of the group, fail if check_no_overlap 28840 fails. 28841 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info 28842 to group if check_no_overlap fails. 28843 288442018-02-21 Segher Boessenkool <segher@kernel.crashing.org> 28845 28846 * config/rs6000/altivec.md: Delete contraint arguments to 28847 define_expand, define_split, and define_peephole2, and in 28848 define_insn_and_split if always unused. 28849 * config/rs6000/darwin.md: Ditto. 28850 * config/rs6000/dfp.md: Ditto. 28851 * config/rs6000/rs6000.md: Ditto. 28852 * config/rs6000/sync.md: Ditto. 28853 * config/rs6000/vector.md: Ditto. 28854 * config/rs6000/vsx.md: Ditto. 28855 288562018-02-21 Segher Boessenkool <segher@kernel.crashing.org> 28857 28858 * config/rs6000/altivec.md: Write output control strings as braced 28859 blocks instead of double-quoted strings. 28860 * config/rs6000/darwin.md: Ditto. 28861 * config/rs6000/rs6000.md: Ditto. 28862 * config/rs6000/vector.md: Ditto. 28863 * config/rs6000/vsx.md: Ditto. 28864 288652018-02-21 Jason Merrill <jason@redhat.com> 28866 28867 PR c++/84314 - ICE with templates and fastcall attribute. 28868 * attribs.c (build_type_attribute_qual_variant): Remove assert. 28869 288702018-02-21 Jan Hubicka <hubicka@ucw.cz> 28871 28872 * ipa-cp.c (determine_versionability): Fix comment typos. 28873 288742018-02-21 Jan Hubicka <hubicka@ucw.cz> 28875 28876 PR c/84229 28877 * ipa-cp.c (determine_versionability): Do not version functions caling 28878 va_arg_pack. 28879 288802018-02-21 Martin Liska <mliska@suse.cz> 28881 28882 PR driver/83193 28883 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): 28884 Add "native" as a possible value. 28885 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define 28886 the macro when native cpu detection is available. 28887 288882018-02-21 Martin Liska <mliska@suse.cz> 28889 28890 PR driver/83193 28891 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option): 28892 Add "native" as a possible value. 28893 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro 28894 when native cpu detection is available. 28895 288962018-02-21 Jakub Jelinek <jakub@redhat.com> 28897 Martin Sebor <msebor@redhat.com> 28898 28899 PR tree-optimization/84478 28900 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to 28901 false. 28902 * gimple-fold.c (get_range_strlen): Make minlen const and assume it 28903 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR 28904 support which is conservatively correct, for 2 only stay conservative 28905 for maxlen. Formatting and comment capitalization fixes. Add STRICT 28906 argument to the 2 argument get_range_strlen, adjust 6 arg 28907 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned 28908 false. 28909 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller. 28910 (gimple_fold_builtin_strlen): Pass true as last argument to 28911 get_range_strlen. 28912 289132018-02-20 Martin Sebor <msebor@redhat.com> 28914 28915 PR middle-end/84095 28916 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New. 28917 (builtin_memref::set_base_and_offset): Same. Handle inner references. 28918 (builtin_memref::builtin_memref): Factor out parts into 28919 set_base_and_offset and call it. 28920 289212018-02-20 Richard Sandiford <richard.sandiford@linaro.org> 28922 28923 PR middle-end/84406 28924 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode 28925 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with 28926 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the 28927 search at the associated MODE_INT. 28928 289292018-02-20 Jeff Law <law@redhat.com> 28930 28931 PR middle-end/82123 28932 PR tree-optimization/81592 28933 PR middle-end/79257 28934 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer 28935 for range data rather than using global data. 28936 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for 28937 range data rather than using global data. 28938 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter 28939 pass it to children as needed. 28940 (struct directive::fmtresult): Similarly. 28941 (struct directive::set_width): Similarly. 28942 (struct directive::set_precision): Similarly. 28943 (format_integer, format_directive, parse_directive): Similarly. 28944 (format_none): Accept unnamed vr_values parameter. 28945 (format_percent, format_floating, format_character): Similarly. 28946 (format_string, format_plain): Similarly. 28947 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query 28948 the EVRP range analyzer for range data rather than using global data. 28949 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and 28950 gimple-ssa-evrp-analyze.h 28951 (class sprintf_dom_walker): Add after_dom_children member function. 28952 Add evrp_range_analyzer member. 28953 (sprintf_dom_walker::before_dom_children): Call into the EVRP 28954 range analyzer as needed. 28955 (sprintf_dom_walker::after_dom_children): New member function. 28956 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing 28957 if not optimizing. 28958 (evrp_range_analyzer::record_ranges_from_stmt): Likewise. 28959 (evrp_range_analyzer::pop_to_marker): Likewise. 28960 289612018-02-20 Richard Sandiford <richard.sandiford@linaro.org> 28962 28963 PR tree-optimization/84419 28964 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF 28965 with the required type if its current type is compatible but 28966 different. 28967 289682018-02-20 Jakub Jelinek <jakub@redhat.com> 28969 28970 PR middle-end/82004 28971 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until 28972 after vectorization. 28973 289742018-02-20 Martin Liska <mliska@suse.cz> 28975 28976 PR driver/83193 28977 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print 28978 possible values if we don't have a hint. 28979 289802018-02-20 Martin Liska <mliska@suse.cz> 28981 28982 PR c/84310 28983 PR target/79747 28984 * final.c (shorten_branches): Build align_tab array with one 28985 more element. 28986 * opts.c (finish_options): Add alignment option limit check. 28987 (MAX_CODE_ALIGN): Likewise. 28988 (MAX_CODE_ALIGN_VALUE): Likewise. 28989 * doc/invoke.texi: Document maximum allowed option value for 28990 all -falign-* options. 28991 289922018-02-19 Jakub Jelinek <jakub@redhat.com> 28993 28994 PR target/84146 28995 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note. 28996 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove. 28997 * var-tracking.c (emit_note_insn_var_location): Remove all references 28998 to NOTE_INSN_CALL_ARG_LOCATION. 28999 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on 29000 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note. 29001 Use copy_rtx_if_shared. 29002 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of 29003 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note. 29004 (dwarf2out_var_location): Remove handling of 29005 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note 29006 on call_insn. 29007 * final.c (final_scan_insn): Remove all references to 29008 NOTE_INSN_CALL_ARG_LOCATION. 29009 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes 29010 before dumping final insns. 29011 * except.c (emit_note_eh_region_end): Remove all references to 29012 NOTE_INSN_CALL_ARG_LOCATION. 29013 * config/alpha/alpha.c (alpha_pad_function_end): Likewise. 29014 * config/c6x/c6x.c (c6x_gen_bundles): Likewise. 29015 * config/arc/arc.c (hwloop_optimize): Likewise. 29016 * config/arm/arm.c (create_fix_barrier): Likewise. 29017 * config/s390/s390.c (s390_chunkify_start): Likewise. 29018 * config/sh/sh.c (find_barrier): Likewise. 29019 * config/i386/i386.c (rest_of_insert_endbranch, 29020 ix86_seh_fixup_eh_fallthru): Likewise. 29021 * config/xtensa/xtensa.c (hwloop_optimize): Likewise. 29022 * config/iq2000/iq2000.c (final_prescan_insn): Likewise. 29023 * config/frv/frv.c (frv_function_prologue): Likewise. 29024 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION 29025 reg note. 29026 (note_outside_basic_block_p): Remove all references to 29027 NOTE_INSN_CALL_ARG_LOCATION. 29028 * gengtype.c (adjust_field_rtx_def): Likewise. 29029 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn): 29030 Likewise. 29031 * jump.c (cleanup_barriers, delete_related_insns): Likewise. 29032 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise. 29033 29034 PR c++/84444 29035 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t) 29036 is ADDR_EXPR. 29037 29038 PR tree-optimization/84452 29039 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call 29040 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen 29041 is NULL. 29042 290432018-02-19 Martin Liska <mliska@suse.cz> 29044 29045 PR sanitizer/82183 29046 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0. 29047 290482018-02-19 Martin Liska <mliska@suse.cz> 29049 Richard Sandiford <richard.sandiford@linaro.org> 29050 29051 PR tree-optimization/82491 29052 * gimple-fold.c (get_base_constructor): Make earlier bail out 29053 to prevent ubsan. 29054 290552018-02-19 Carl Love <cel@us.ibm.com> 29056 29057 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from 29058 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to 29059 BU_P8V_OVERLOAD_1. 29060 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to 29061 P8V_BUILTIN_VEC_NEG. 29062 290632018-02-19 Sebastian Perta <sebastian.perta@renesas.com> 29064 29065 * config/rl78/rl78.md (movdf): New define expand. 29066 290672018-02-19 Martin Liska <mliska@suse.cz> 29068 29069 PR other/80589 29070 * doc/invoke.texi: Fix typo. 29071 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise. 29072 290732018-02-18 Segher Boessenkool <segher@kernel.crashing.org> 29074 29075 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't 29076 handle rs6000_single_float and rs6000_double_float specially for 29077 e500 family CPUs. 29078 290792018-02-16 Jeff Law <law@redhat.com> 29080 29081 * config/rx/rx.c (add_pop_cfi_notes): New function.; 29082 (pop_regs): Use it. 29083 290842018-02-16 Jakub Jelinek <jakub@redhat.com> 29085 29086 PR ipa/84425 29087 * ipa-inline.c (inline_small_functions): Fix a typo. 29088 290892018-02-16 Nathan Sidwell <nathan@acm.org> 29090 29091 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'. 29092 290932018-02-16 Carl Love <cel@us.ibm.com> 29094 29095 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition. 29096 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and 29097 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2. 29098 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2 29099 expansion to P8V_BUILTIN_VEC_FLOAT2. 29100 291012018-02-16 Vladimir Makarov <vmakarov@redhat.com> 29102 29103 PR rtl-optimization/70023 29104 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of 29105 src_regno into account. 29106 291072018-02-16 Carl Love <cel@us.ibm.com> 29108 29109 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b. 29110 * config/rs6000/rs6000-builtin.def: Remove macro expansion for 29111 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B. 29112 * config/rs6000/rs6000.c: Remove case statements for 29113 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B, 29114 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI, 29115 and P9V_BUILTIN_VEC_VINSERT4B. 29116 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for 29117 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B. 29118 * config/rs6000/vsx.md: 29119 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for 29120 vec_insert4b. 29121 291222018-02-16 Carl Love <cel@us.ibm.com> 29123 29124 * config/rs6000/altivec.h: Add builtin names vec_extract4b 29125 vec_insert4b. 29126 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B 29127 definitions. 29128 * config/rs6000/rs6000-c.c: Add the definitions for 29129 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B. 29130 * config/rs6000/rs6000.c (altivec_expand_builtin): Add 29131 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements. 29132 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand 29133 definition for insert4b and define insn *insert3b_internal. 29134 * doc/extend.texi: Add documentation for vec_extract4b. 29135 291362018-02-16 Nathan Sidwell <nathan@acm.org> 29137 29138 * doc/extend.texi (Backwards Compatibility): Mention friend 29139 injection. Note for-scope is deprecated. 29140 * doc/invoke.texi (-ffriend-injection): Deprecate. 29141 291422018-02-16 Segher Boessenkool <segher@kernel.crashing.org> 29143 29144 * combine.c (try_combine): When adjusting LOG_LINKS for the destination 29145 that moved to I2, also allow destinations that are a paradoxical 29146 subreg (instead of a normal reg). 29147 291482018-02-16 Oleg Endo <olegendo@gcc.gnu.org> 29149 29150 PR target/83831 29151 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand 29152 to QImode. 29153 291542018-02-16 Richard Biener <rguenther@suse.de> 29155 29156 PR tree-optimization/84037 29157 PR tree-optimization/84016 29158 PR target/82862 29159 * config/i386/i386.c (ix86_builtin_vectorization_cost): 29160 Adjust vec_construct for the fact we need additional higher latency 29161 128bit inserts for AVX256 and AVX512 vector builds. 29162 (ix86_add_stmt_cost): Scale vector construction cost for 29163 elementwise loads. 29164 291652018-02-16 Richard Biener <rguenther@suse.de> 29166 29167 PR tree-optimization/84417 29168 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain 29169 the MEM_REF offset when conversion to BIT_FIELD_REF is desired. 29170 (non_rewritable_lvalue_p): Likewise, use poly-ints. 29171 291722018-02-16 Martin Liska <mliska@suse.cz> 29173 29174 PR sanitizer/84307 29175 * internal-fn.def (ASAN_CHECK): Set proper flags. 29176 (ASAN_MARK): Likewise. 29177 291782018-02-16 Julia Koval <julia.koval@intel.com> 29179 29180 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB 29181 from PTA_CANNONLAKE. 29182 291832018-02-16 Jakub Jelinek <jakub@redhat.com> 29184 29185 PR target/84272 29186 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest): 29187 Use ++iter rather than iter++ for std::list iterators. 29188 (func_fma_steering::dfs): Likewise. Don't delete nodes right away, 29189 defer deleting them until all nodes in the forest are processed. Do 29190 free even leaf nodes. Change to_process into auto_vec. 29191 29192 PR bootstrap/84405 29193 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3. 29194 * vec.h (vec_default_construct): Use memset instead of placement new 29195 if BROKEN_VALUE_INITIALIZATION is defined. 29196 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use 29197 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION 29198 is defined. 29199 29200 PR rtl-optimization/83723 29201 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument. 29202 * lra.c (lra_substitute_pseudo): Likewise. If true, use 29203 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to 29204 recursive calls. 29205 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo 29206 callers. 29207 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise. 29208 292092018-02-16 Eric Botcazou <ebotcazou@adacore.com> 29210 29211 PR rtl-optimization/81443 29212 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results 29213 from inner REGs to paradoxical SUBREGs. 29214 292152018-02-16 Richard Biener <rguenther@suse.de> 29216 29217 PR tree-optimization/84399 29218 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p): 29219 For operands we can analyze at their definition make sure we can 29220 analyze them at each use as well. 29221 292222018-02-16 Richard Biener <rguenther@suse.de> 29223 29224 PR tree-optimization/84190 29225 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch 29226 volatile accesses if the decl isn't volatile. 29227 292282018-02-15 Jason Merrill <jason@redhat.com> 29229 29230 PR c++/84314 - ICE with templates and fastcall attribute. 29231 * attribs.c (build_type_attribute_qual_variant): Don't clobber 29232 TYPE_CANONICAL on an existing type. 29233 292342018-02-15 Jakub Jelinek <jakub@redhat.com> 29235 29236 PR tree-optimization/84383 29237 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at 29238 dstoff nor call operand_equal_p if dstbase is NULL. 29239 29240 PR tree-optimization/84334 29241 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is 29242 also a CONSTANT_CLASS_P, punt. 29243 292442018-02-14 Jim Wilson <jimw@sifive.com> 29245 29246 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after 29247 first SMALL_OPERAND check. New local min_second_step. Move assert 29248 to where locals are set. Add TARGET_RVC support. 29249 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New. 29250 292512018-02-14 Indu Bhagat <indu.bhagat@oracle.com> 29252 29253 * doc/invoke.texi: Correct -Wformat-overflow code sample. 29254 292552018-02-14 Martin Sebor <msebor@redhat.com> 29256 29257 PR tree-optimization/83698 29258 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For 29259 arrays constrain the offset range to their bounds. 29260 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset. 29261 (builtin_access::overlap): Avoid setting the size of overlap if it's 29262 already been set. 29263 (maybe_diag_overlap): Also consider arrays when deciding what values 29264 of offsets to include in diagnostics. 29265 292662018-02-14 Martin Sebor <msebor@redhat.com> 29267 29268 PR c/84108 29269 * attribs.c (diag_attr_exclusions): Consider the exclusion(s) 29270 that correspond to the kind of a declaration. 29271 292722018-02-14 John David Anglin <danglin@gcc.gnu.org> 29273 29274 PR target/83984 29275 * config/pa/pa.md: Load address of PIC label using the linkage table 29276 if the label is nonlocal. 29277 292782018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org> 29279 29280 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue 29281 warning message if user requests -maltivec=be. 29282 * doc/invoke.texi: Document deprecation of -maltivec=be. 29283 292842018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com> 29285 29286 PR target/84220 29287 * config/rs6000/rs6000-c.c: Update definitions for 29288 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW, 29289 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins. 29290 292912018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com> 29292 29293 PR target/84239 29294 * config/i386/cetintrin.h: Remove _rdssp[d|q] and 29295 add _get_ssp intrinsics. Remove argument from 29296 __builtin_ia32_rdssp[d|q]. 29297 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID. 29298 * config/i386/i386-builtin.def: Remove argument from 29299 __builtin_ia32_rdssp[d|q]. 29300 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use 29301 ix86_expand_special_args_builtin for _rdssp[d|q]. 29302 * config/i386/i386.md: Remove argument from rdssp[si|di] insn. 29303 Clear register before usage. 29304 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q]. 29305 Add documentation for new _get_ssp and _inc_ssp intrinsics. 29306 293072018-02-14 Richard Sandiford <richard.sandiford@linaro.org> 29308 29309 PR tree-optimization/84357 29310 * tree-data-ref.c (object_address_invariant_in_loop_p): Check 29311 operand 1 of an ARRAY_REF too. 29312 293132018-02-14 Oleg Endo <olegendo@gcc.gnu.org> 29314 29315 PR target/83831 29316 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn, 29317 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New 29318 declarations. 29319 (set_of_reg): New struct. 29320 (rx_find_set_of_reg, rx_find_use_of_reg): New functions. 29321 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn, 29322 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New 29323 functions. 29324 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split. 29325 Split into bitclr, bitset, bitinvert patterns if appropriate. 29326 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and 29327 use rx_fuse_in_memory_bitop. 29328 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert 29329 to named insn, correct maximum insn length. 29330 293312018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com> 29332 29333 PR target/79242 29334 * machmode.def: Define a complex mode for PARTIAL_INT. 29335 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for 29336 MODE_PARTIAL_INT. 29337 * doc/rtl.texi: Document CSPImode. 29338 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode 29339 handling. 29340 (msp430_hard_regno_nregs_with_padding): Likewise. 29341 293422018-02-13 Peter Bergner <bergner@vnet.ibm.com> 29343 29344 PR target/84279 29345 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses. 29346 293472018-02-13 Segher Boessenkool <segher@kernel.crashing.org> 29348 29349 PR rtl-optimization/84169 29350 * combine.c (try_combine): New variable split_i2i3. Set it to true if 29351 we generated a parallel as new i3 and we split that to new i2 and i3 29352 instructions. Handle split_i2i3 similar to swap_i2i3: scan the 29353 LOG_LINKs of i3 to see which of those need to link to i2 now. Link 29354 those to i2, not i1. Partially rewrite this scan code. 29355 293562018-02-13 Jakub Jelinek <jakub@redhat.com> 29357 29358 PR c/82210 29359 * stor-layout.c (place_field): For variable length fields, adjust 29360 offset_align afterwards not just based on the field's alignment, 29361 but also on the size. 29362 29363 PR middle-end/84309 29364 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead 29365 of exps and logs in the use_exp2 case. 29366 293672018-02-13 Jeff Law <law@redhat.com> 29368 29369 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and 29370 entry for "vector". 29371 29372 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark 29373 ARGS as unused. 29374 293752018-02-13 Alexandre Oliva <aoliva@redhat.com> 29376 29377 PR debug/84342 29378 PR debug/84319 29379 * common.opt (gas-loc-support, gas-locview-support): New. 29380 (ginline-points, ginternal-reset-location-views): New. 29381 * doc/invoke.texi: Document them. Use @itemx where intended. 29382 (gvariable-location-views): Adjust. 29383 * target.def (reset_location_view): New. 29384 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New. 29385 (TARGET_RESET_LOCATION_VIEW): New. 29386 * doc/tm.texi: Rebuilt. 29387 * dwarf2out.c (dwarf2out_default_as_loc_support): New. 29388 (dwarf2out_default_as_locview_support): New. 29389 (output_asm_line_debug_info): Use option variables. 29390 (dwarf2out_maybe_output_loclist_view_pair): Likewise. 29391 (output_loc_list): Likewise. 29392 (add_high_low_attributes): Check option variables. 29393 Don't output entry view attribute in strict mode. 29394 (gen_inlined_subroutine_die): Check option variables. 29395 (dwarf2out_inline_entry): Likewise. 29396 (init_sections_and_labels): Likewise. 29397 (dwarf2out_early_finish): Likewise. 29398 (maybe_reset_location_view): New, from... 29399 (dwarf2out_var_location): ... here. Call it. 29400 * debug.h (dwarf2out_default_as_loc_support): Declare. 29401 (dwarf2out_default_as_locview_support): Declare. 29402 * hooks.c (hook_int_rtx_insn_0): New. 29403 * hooks.h (hook_int_rtx_insn_0): Declare. 29404 * toplev.c (process_options): Take -gas-loc-support and 29405 -gas-locview-support from dwarf2out. Enable 29406 -gvariable-location-views by default only with locview 29407 assembler support. Enable -ginternal-reset-location-views by 29408 default only if the target defines the corresponding hook. 29409 Enable -ginline-points by default if location views are 29410 enabled; force it disabled if statement frontiers are 29411 disabled. 29412 * tree-inline.c (expand_call_inline): Check option variables. 29413 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise. 29414 294152018-02-13 Richard Sandiford <richard.sandiford@linaro.org> 29416 29417 PR tree-optimization/84321 29418 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE 29419 handling. Also check whether the anti-range contains any values 29420 that satisfy the mask; switch to a VR_RANGE if not. 29421 294222018-02-13 Paolo Bonzini <bonzini@gnu.org> 29423 29424 PR sanitizer/84340 29425 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec. 29426 294272018-02-13 Martin Jambor <mjambor@suse.cz> 29428 29429 PR c++/83990 29430 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location 29431 of call statements, also set location of a load to a temporary. 29432 294332018-02-13 Sebastian Perta <sebastian.perta@renesas.com> 29434 29435 * config/rl78/rl78.c (add_vector_labels): New function. 29436 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function. 29437 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels. 29438 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert 29439 which checks that no arguments are passed. 29440 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute. 29441 * doc/extend.texi: Documentation for the new attribute. 29442 294432018-02-13 Andreas Schwab <schwab@suse.de> 29444 29445 * config/riscv/linux.h (CPP_SPEC): Define. 29446 294472018-02-13 Jakub Jelinek <jakub@redhat.com> 29448 29449 PR target/84335 29450 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass 29451 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of 29452 OPTION_MASK_ISA_AES as first argument to def_builtin_const 29453 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2 29454 instead of OPTION_MASK_ISA_PCLMUL as first argument to 29455 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin. 29456 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it 29457 temporarily for AES and PCLMUL builtins. 29458 29459 PR tree-optimization/84339 29460 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling 29461 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF. 29462 Formatting fixes. 29463 29464 PR middle-end/84309 29465 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into 29466 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available. 29467 * generic-match-head.c (canonicalize_math_after_vectorization_p): New 29468 inline function. 29469 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New 29470 inline function. 29471 * omp-simd-clone.h: New file. 29472 * omp-simd-clone.c: Include omp-simd-clone.h. 29473 (expand_simd_clones): No longer static. 29474 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h, 29475 cgraph.h and omp-simd-clone.h. 29476 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x). 29477 (vect_recog_widen_shift_pattern): Formatting fix. 29478 (vect_pattern_recog_1): Don't check optab for calls. 29479 29480 PR target/84336 29481 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force 29482 operands[2] into a REG before using gen_lowpart on it. 29483 294842018-02-12 Jeff Law <law@redhat.com> 29485 29486 PR target/83760 29487 * config/sh/sh.c (find_barrier): Consider a sibling call 29488 a barrier as well. 29489 29490 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when 29491 successfully back substituting a reg. 29492 294932018-02-12 Richard Biener <rguenther@suse.de> 29494 29495 PR tree-optimization/84037 29496 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited 29497 parameter, move visited init to caller. 29498 (vect_slp_analyze_operations): Separate cost from validity 29499 check, initialize visited once for all instances. 29500 (vect_schedule_slp): Analyze map to CSE vectorized nodes once 29501 for all instances. 29502 * tree-vect-stmts.c (vect_model_simple_cost): Make early 29503 out an assert. 29504 (vect_model_promotion_demotion_cost): Likewise. 29505 (vectorizable_bswap): Guard cost modeling with !slp_node 29506 instead of !PURE_SLP_STMT to avoid double-counting on hybrid 29507 SLP stmts. 29508 (vectorizable_call): Likewise. 29509 (vectorizable_conversion): Likewise. 29510 (vectorizable_assignment): Likewise. 29511 (vectorizable_shift): Likewise. 29512 (vectorizable_operation): Likewise. 29513 (vectorizable_store): Likewise. 29514 (vectorizable_load): Likewise. 29515 (vectorizable_condition): Likewise. 29516 (vectorizable_comparison): Likewise. 29517 295182018-02-12 Paolo Bonzini <bonzini@gnu.org> 29519 29520 PR sanitizer/84307 29521 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value. 29522 (ASAN_MARK): Fix fnspec to account for return value, change pointer 29523 argument from 'R' to 'W' so that the pointed-to datum is clobbered. 29524 295252018-02-08 Jan Hubicka <hubicka@ucw.cz> 29526 29527 PR middle-end/83665 29528 * params.def (inline-min-speedup): Increase from 8 to 15. 29529 (max-inline-insns-auto): Decrease from 40 to 30. 29530 * ipa-split.c (consider_split): Add some buffer for function to 29531 be considered inlining candidate. 29532 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate 29533 default values. 29534 295352018-02-12 Richard Biener <rguenther@suse.de> 29536 29537 PR tree-optimization/84037 29538 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the 29539 matched stmts if we cannot swap the non-matched ones. 29540 295412018-02-12 Olga Makhotina <olga.makhotina@intel.com> 29542 29543 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd, 29544 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss, 29545 _mm_maskz_scalef_round_ss): New intrinsics. 29546 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix. 29547 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round, 29548 __builtin_ia32_scalefss_round): Remove. 29549 (__builtin_ia32_scalefsd_mask_round, 29550 __builtin_ia32_scalefss_mask_round): New intrinsics. 29551 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ... 29552 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this. 29553 ((match_operand:VF_128 2 "<round_nimm_predicate>" 29554 "<round_constraint>")): Changed to ... 29555 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>" 29556 "<round_scalar_constraint>")): ... this. 29557 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0| 29558 %0, %1, %2<round_op3>}"): Changed to ... 29559 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1, 29560 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1, 29561 %2<round_scalar_mask_op3>}"): ... this. 29562 * config/i386/subst.md (round_scalar_nimm_predicate): New. 29563 295642018-02-12 Olga Makhotina <olga.makhotina@intel.com> 29565 29566 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd) 29567 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss) 29568 (_mm_maskz_sqrt_round_ss): New intrinsics. 29569 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove. 29570 (__builtin_ia32_sqrtsd_mask_round) 29571 (__builtin_ia32_sqrtss_mask_round): New builtins. 29572 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round) 29573 (__builtin_ia32_sqrtss_round): Remove. 29574 (__builtin_ia32_sqrtsd_mask_round) 29575 (__builtin_ia32_sqrtss_mask_round): New builtins. 29576 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ... 29577 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this. 29578 ((match_operand:VF_128 1 "vector_operand" 29579 "xBm,<round_constraint>")): Changed to ... 29580 ((match_operand:VF_128 1 "vector_operand" 29581 "xBm,<round_scalar_constraint>")): ... this. 29582 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0| 29583 %0, %2, %<iptr>1<round_op3>}): Changed to ... 29584 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2, 29585 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2, 29586 %<iptr>1<round_scalar_mask_op3>}): ... this. 29587 ((set_attr "prefix" "<round_prefix>")): Changed to ... 29588 ((set_attr "prefix" "<round_scalar_prefix>")): ... this. 29589 295902018-02-11 Steven Munroe <munroesj@gcc.gnu.org> 29591 29592 PR target/84266 29593 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]): 29594 Cast vec_cmpeq result to correct type. 29595 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]): 29596 Cast vec_cmpgt result to correct type. 29597 295982018-02-11 Alexandre Oliva <aoliva@redhat.com> 29599 29600 * final.c (final_scan_insn_1): Renamed from... 29601 (final_scan_insn): ... this. New wrapper, to recover 29602 seen from the outermost call in recursive ones. 29603 * config/sparc/sparc.c (output_return): Drop seen from call. 29604 (output_sibcall): Likewise. 29605 * config/visium/visium.c (output_branch): Likewise. 29606 296072018-02-10 John David Anglin <danglin@gcc.gnu.org> 29608 29609 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as 29610 function label. 29611 296122018-02-10 Alan Modra <amodra@gmail.com> 29613 29614 PR target/84300 29615 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..). 29616 Specify LR as an input. 29617 296182018-02-10 Jakub Jelinek <jakub@redhat.com> 29619 29620 PR sanitizer/83987 29621 * omp-low.c (maybe_remove_omp_member_access_dummy_vars, 29622 remove_member_access_dummy_vars): New functions. 29623 (lower_omp_for, lower_omp_taskreg, lower_omp_target, 29624 lower_omp_1, execute_lower_omp): Use them. 29625 29626 PR rtl-optimization/84308 29627 * shrink-wrap.c (spread_components): Release todo vector. 29628 296292018-02-09 Vladimir Makarov <vmakarov@redhat.com> 29630 29631 PR rtl-optimization/57193 29632 * ira-color.c (struct allocno_color_data): Add member 29633 conflict_allocno_hard_prefs. 29634 (update_conflict_allocno_hard_prefs): New. 29635 (bucket_allocno_compare_func): Add a preference based on 29636 conflict_allocno_hard_prefs. 29637 (push_allocno_to_stack): Update conflict_allocno_hard_prefs. 29638 (color_allocnos): Remove a dead code. Initiate 29639 conflict_allocno_hard_prefs. Call update_costs_from_prefs. 29640 296412018-02-09 Jakub Jelinek <jakub@redhat.com> 29642 29643 PR target/84226 29644 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand 29645 constraint from =wa to wa. Avoid a subreg on the output operand, 29646 instead use a pseudo and subreg it in a move. 29647 (p9_xxbrd_<mode>): Changed to ... 29648 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator. 29649 (p9_xxbrd_v2df): New expander. 29650 (p9_xxbrw_<mode>): Changed to ... 29651 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator. 29652 (p9_xxbrw_v4sf): New expander. 29653 296542018-02-09 Sebastian Perta <sebastian.perta@renesas.com> 29655 29656 * config/rx/rx.md (movsicc): Update expander to be matched by GCC. 29657 296582018-02-09 Peter Bergner <bergner@vnet.ibm.com> 29659 29660 PR target/83926 29661 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit 29662 multiply in 32-bit mode. 29663 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode. 29664 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit 29665 mode. 29666 296672018-02-09 Sebastian Perta <sebastian.perta@renesas.com> 29668 29669 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint 29670 to allow or block "symbol_ref" depending on the value of TARGET_JSR. 29671 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF. 29672 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF. 29673 296742018-02-09 Pierre-Marie de Rodat <derodat@adacore.com> 29675 29676 PR lto/84213 29677 * dwarf2out.c (is_trivial_indirect_ref): New function. 29678 (dwarf2out_late_global_decl): Do not generate a location 29679 attribute for variables that have a non-trivial DECL_VALUE_EXPR 29680 and that are not defined in the current unit. 29681 296822018-02-09 Eric Botcazou <ebotcazou@adacore.com> 29683 29684 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison 29685 instead of a libcall for UNORDERED. 29686 296872018-02-09 Tamar Christina <tamar.christina@arm.com> 29688 29689 PR target/82641 29690 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX, 29691 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP. 29692 296932018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com> 29694 29695 PR target/PR84295 29696 * config/s390/s390.c (s390_set_current_function): Invoke 29697 s390_indirect_branch_settings also if fndecl didn't change. 29698 296992018-02-09 Alexandre Oliva <aoliva@redhat.com> 29700 29701 * config/rs6000/rs6000.md (blockage): Set length to zero. 29702 297032018-02-09 Eric Botcazou <ebotcazou@adacore.com> 29704 29705 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion. 29706 297072018-02-09 Jakub Jelinek <jakub@redhat.com> 29708 29709 PR sanitizer/84285 29710 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS, 29711 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like 29712 -static-lib*san. 29713 29714 PR debug/84252 29715 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart 29716 PARALLEL incoming that failed vt_get_decl_and_offset check. 29717 29718 PR middle-end/84237 29719 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false. 29720 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore 29721 TREE_READONLY bit. 29722 (get_variable_section): For decls in named .bss* sections pass true as 29723 second argument to bss_initializer_p. 29724 297252018-02-09 Marek Polacek <polacek@redhat.com> 29726 Jakub Jelinek <jakub@redhat.com> 29727 29728 PR c++/83659 29729 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro. 29730 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01). 29731 Sync some changes from cxx_fold_indirect_ref. 29732 297332018-02-09 Alexandre Oliva <aoliva@redhat.com> 29734 29735 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry 29736 markers. 29737 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook. 29738 (BLOCK_INLINE_ENTRY_LABEL): New. 29739 (dwarf2out_var_location): Disregard inline entry markers. 29740 (inline_entry_data): New struct. 29741 (inline_entry_data_hasher): New hashtable type. 29742 (inline_entry_data_hasher::hash): New. 29743 (inline_entry_data_hasher::equal): New. 29744 (inline_entry_data_table): New variable. 29745 (add_high_low_attributes): Add DW_AT_entry_pc and 29746 DW_AT_GNU_entry_view attributes if a pending entry is found 29747 in inline_entry_data_table. Add old entry_pc attribute only 29748 if debug nonbinding markers are disabled. 29749 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding 29750 markers are enabled. 29751 (block_within_block_p, dwarf2out_inline_entry): New. 29752 (dwarf2out_finish): Check that no entries remained in 29753 inline_entry_data_table. 29754 * final.c (reemit_insn_block_notes): Handle inline entry notes. 29755 (final_scan_insn, notice_source_line): Likewise. 29756 (rest_of_clean_state): Skip inline entry markers. 29757 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry 29758 markers. 29759 * gimple.c (gimple_build_debug_inline_entry): New. 29760 * gimple.h (enum gimple_debug_subcode): Add 29761 GIMPLE_DEBUG_INLINE_ENTRY. 29762 (gimple_build_debug_inline_entry): Declare. 29763 (gimple_debug_inline_entry_p): New. 29764 (gimple_debug_nonbind_marker_p): Adjust. 29765 * insn-notes.def (INLINE_ENTRY): New. 29766 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle 29767 inline entry marker notes. 29768 (print_insn): Likewise. 29769 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support. 29770 (INSN_DEBUG_MARKER_KIND): Likewise. 29771 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New. 29772 * tree-inline.c (expand_call_inline): Build and insert 29773 debug_inline_entry stmt. 29774 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve 29775 inline entry blocks early, if nonbind markers are enabled. 29776 (dump_scope_block): Dump fragment info. 29777 * var-tracking.c (reemit_marker_as_note): Handle inline entry note. 29778 * doc/gimple.texi (gimple_debug_inline_entry_p): New. 29779 (gimple_build_debug_inline_entry): New. 29780 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers): 29781 Enable/disable inline entry points too. 29782 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New. 29783 (DEBUG_INSN): Describe inline entry markers. 29784 29785 * common.opt (gvariable-location-views): New. 29786 (gvariable-location-views=incompat5): New. 29787 * config.in: Rebuilt. 29788 * configure: Rebuilt. 29789 * configure.ac: Test assembler for view support. 29790 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New. 29791 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare. 29792 * dwarf2out.c (var_loc_view): New typedef. 29793 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend. 29794 (dwarf2out_locviews_in_attribute): New. 29795 (dwarf2out_locviews_in_loclist): New. 29796 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists. 29797 (enum dw_line_info_opcode): Add LI_adv_address. 29798 (struct dw_line_info_table): Add view. 29799 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros. 29800 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default. 29801 (zero_view_p): New variable. 29802 (ZERO_VIEW_P): New macro. 29803 (output_asm_line_debug_info): New. 29804 (struct var_loc_node): Add view. 29805 (add_AT_view_list, AT_loc_list): New. 29806 (add_var_loc_to_decl): Add view param. Test it against last. 29807 (new_loc_list): Add view params. Record them. 29808 (AT_loc_list_ptr): Handle loc and view lists. 29809 (view_list_to_loc_list_val_node): New. 29810 (print_dw_val): Handle dw_val_class_view_list. 29811 (size_of_die): Likewise. 29812 (value_format): Likewise. 29813 (loc_list_has_views): New. 29814 (gen_llsym): Set vl_symbol too. 29815 (maybe_gen_llsym, skip_loc_list_entry): New. 29816 (dwarf2out_maybe_output_loclist_view_pair): New. 29817 (output_loc_list): Output view list or entries too. 29818 (output_view_list_offset): New. 29819 (output_die): Handle dw_val_class_view_list. 29820 (output_dwarf_version): New. 29821 (output_compilation_unit_header): Use it. 29822 (output_skeleton_debug_sections): Likewise. 29823 (output_rnglists, output_line_info): Likewise. 29824 (output_pubnames, output_aranges): Update version comments. 29825 (output_one_line_info_table): Output view numbers in asm comments. 29826 (dw_loc_list): Determine current endview, pass it to new_loc_list. 29827 Call maybe_gen_llsym. 29828 (loc_list_from_tree_1): Adjust. 29829 (add_AT_location_description): Create view list attribute if 29830 needed, check it's absent otherwise. 29831 (convert_cfa_to_fb_loc_list): Adjust. 29832 (maybe_emit_file): Call output_asm_line_debug_info for test. 29833 (dwarf2out_var_location): Reset views as needed. Precompute 29834 add_var_loc_to_decl args. Call get_attr_min_length only if we have the 29835 attribute. Set view. 29836 (new_line_info_table): Reset next view. 29837 (set_cur_line_info_table): Call output_asm_line_debug_info for test. 29838 (dwarf2out_source_line): Likewise. Output view resets and labels to 29839 the assembler, or select appropriate line info opcodes. 29840 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list. 29841 (optimize_string_length): Catch it. Adjust. 29842 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle 29843 dw_val_class_view_list, and remove it if no longer needed. 29844 (hash_loc_list): Hash view numbers. 29845 (loc_list_hasher::equal): Compare them. 29846 (optimize_location_lists): Check whether a view list symbol is 29847 needed, and whether the locview attribute is present, and 29848 whether they match. Remove the locview attribute if no longer 29849 needed. 29850 (index_location_lists): Call skip_loc_list_entry for test. 29851 (dwarf2out_finish): Call output_asm_line_debug_info for test. 29852 Use output_dwarf_version. 29853 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list. 29854 (struct dw_val_node): Add val_view_list. 29855 * final.c (SEEN_NEXT_VIEW): New. 29856 (set_next_view_needed): New. 29857 (clear_next_view_needed): New. 29858 (maybe_output_next_view): New. 29859 (final_start_function): Rename to... 29860 (final_start_function_1): ... this. Take pointer to FIRST, 29861 add SEEN parameter. Emit param bindings in the initial view. 29862 (final_start_function): Reintroduce SEEN-less interface. 29863 (final): Rename to... 29864 (final_1): ... this. Take SEEN parameter. Output final pending 29865 next view at the end. 29866 (final): Reintroduce seen-less interface. 29867 (final_scan_insn): Output pending next view before switching 29868 sections or ending a block. Mark the next view as needed when 29869 outputting variable locations. Notify debug backend of section 29870 changes, and of location view changes. 29871 (rest_of_handle_final): Adjust. 29872 * toplev.c (process_options): Autodetect value for debug variable 29873 location views option. Warn on incompat5 without -gdwarf-5. 29874 * doc/invoke.texi (gvariable-location-views): New. 29875 (gvariable-location-views=incompat5): New. 29876 (gno-variable-location-views): New. 29877 298782018-02-08 David Malcolm <dmalcolm@redhat.com> 29879 29880 PR tree-optimization/84136 29881 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion 29882 that the result of find_edge is non-NULL. 29883 298842018-02-08 Sergey Shalnov <sergey.shalnov@intel.com> 29885 29886 PR target/83008 29887 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of 29888 storing integer register in SImode. Fix cost of 256 and 512 29889 byte aligned SSE register store. 29890 298912018-02-08 Sergey Shalnov <sergey.shalnov@intel.com> 29892 29893 * config/i386/i386.c (ix86_multiplication_cost): Fix 29894 multiplication cost for TARGET_AVX512DQ. 29895 298962018-02-08 Marek Polacek <polacek@redhat.com> 29897 29898 PR tree-optimization/84238 29899 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of 29900 get_range_strlen. 29901 299022018-02-08 Richard Sandiford <richard.sandiford@linaro.org> 29903 29904 PR tree-optimization/84265 29905 * tree-vect-stmts.c (vectorizable_store): Don't treat 29906 VMAT_CONTIGUOUS accesses as grouped. 29907 (vectorizable_load): Likewise. 29908 299092018-02-08 Richard Sandiford <richard.sandiford@linaro.org> 29910 29911 PR tree-optimization/81635 29912 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare. 29913 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask) 29914 (test_round_for_mask): New functions. 29915 (wide_int_cc_tests): Call test_round_for_mask. 29916 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare. 29917 * tree-vrp.c (intersect_range_with_nonzero_bits): New function. 29918 * tree-data-ref.c (split_constant_offset_1): Use it to refine the 29919 range returned by get_range_info. 29920 299212018-02-08 Jan Hubicka <hubicka@ucw.cz> 29922 29923 PR ipa/81360 29924 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare 29925 * symtab.c: Include builtins.h 29926 (symtab_node::output_to_lto_symbol_table_p): Move here 29927 from lto-streamer-out.c:output_symbol_p. 29928 * lto-streamer-out.c (write_symbol): Turn early exit to assert. 29929 (output_symbol_p): Move all logic to symtab.c 29930 (produce_symtab): Update. 29931 299322018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com> 29933 29934 * config/s390/s390-opts.h (enum indirect_branch): Define. 29935 * config/s390/s390-protos.h (s390_return_addr_from_memory) 29936 (s390_indirect_branch_via_thunk) 29937 (s390_indirect_branch_via_inline_thunk): Add function prototypes. 29938 (enum s390_indirect_branch_type): Define. 29939 * config/s390/s390.c (struct s390_frame_layout, struct 29940 machine_function): Remove. 29941 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask) 29942 (indirect_branch_table_label_no, indirect_branch_table_name): 29943 Define variables. 29944 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro. 29945 (enum s390_indirect_branch_option): Define. 29946 (s390_return_addr_from_memory): New function. 29947 (s390_handle_string_attribute): New function. 29948 (s390_attribute_table): Add new attribute handler. 29949 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns. 29950 (s390_indirect_branch_via_thunk): New function. 29951 (s390_indirect_branch_via_inline_thunk): New function. 29952 (s390_function_ok_for_sibcall): When jumping via thunk disallow 29953 sibling call optimization for non z10 compiles. 29954 (s390_emit_call): Force indirect branch target to be a single 29955 register. Add r1 clobber for non-z10 compiles. 29956 (s390_emit_epilogue): Emit return jump via return_use expander. 29957 (s390_reorg): Handle JUMP_INSNs as execute targets. 29958 (s390_option_override_internal): Perform validity checks for the 29959 new command line options. 29960 (s390_indirect_branch_attrvalue): New function. 29961 (s390_indirect_branch_settings): New function. 29962 (s390_set_current_function): Invoke s390_indirect_branch_settings. 29963 (s390_output_indirect_thunk_function): New function. 29964 (s390_code_end): Implement target hook. 29965 (s390_case_values_threshold): Implement target hook. 29966 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target 29967 macros. 29968 * config/s390/s390.h (struct s390_frame_layout) 29969 (struct machine_function): Move here from s390.c. 29970 (TARGET_INDIRECT_BRANCH_NOBP_RET) 29971 (TARGET_INDIRECT_BRANCH_NOBP_JUMP) 29972 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK) 29973 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK) 29974 (TARGET_INDIRECT_BRANCH_NOBP_CALL) 29975 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE) 29976 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL) 29977 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX) 29978 (TARGET_INDIRECT_BRANCH_TABLE): Define macros. 29979 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP) 29980 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants. 29981 (mnemonic attribute): Add values which aren't recognized 29982 automatically. 29983 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable 29984 pattern for branch conversion. Fix mnemonic attribute. 29985 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit 29986 indirect branch via thunk if requested. 29987 ("indirect_jump", "<code>"): Expand patterns for branch conversion. 29988 ("*indirect_jump"): Disable for branch conversion using out of 29989 line thunks. 29990 ("indirect_jump_via_thunk<mode>_z10") 29991 ("indirect_jump_via_thunk<mode>") 29992 ("indirect_jump_via_inlinethunk<mode>_z10") 29993 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump") 29994 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>") 29995 ("casesi_jump_via_inlinethunk<mode>_z10") 29996 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10") 29997 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10") 29998 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern. 29999 ("*indirect2_jump"): Disable for branch conversion. 30000 ("casesi_jump"): Turn into expander and expand patterns for branch 30001 conversion. 30002 ("return_use"): New expander. 30003 ("*return"): Emit return via thunk and rename it to ... 30004 ("*return<mode>"): ... this one. 30005 * config/s390/s390.opt: Add new options and and enum for the 30006 option values. 30007 300082018-02-08 Richard Sandiford <richard.sandiford@linaro.org> 30009 30010 * lra-constraints.c (match_reload): Unconditionally use 30011 gen_lowpart_SUBREG, rather than selecting between that 30012 and equivalent gen_rtx_SUBREG code. 30013 300142018-02-08 Richard Biener <rguenther@suse.de> 30015 30016 PR tree-optimization/84233 30017 * tree-ssa-phiprop.c (propagate_with_phi): Use separate 30018 changed flag instead of boguously re-using phi_inserted. 30019 300202018-02-08 Martin Jambor <mjambor@suse.cz> 30021 30022 * hsa-gen.c (get_symbol_for_decl): Set program allocation for 30023 static local variables. 30024 300252018-02-08 Richard Biener <rguenther@suse.de> 30026 30027 PR tree-optimization/84278 30028 * tree-vect-stmts.c (vectorizable_store): When looking for 30029 smaller vector types to perform grouped strided loads/stores 30030 make sure the mode is supported by the target. 30031 (vectorizable_load): Likewise. 30032 300332018-02-08 Wilco Dijkstra <wdijkstr@arm.com> 30034 30035 * config/aarch64/aarch64.c (aarch64_components_for_bb): 30036 Increase LDP/STP opportunities by adding adjacent callee-saves. 30037 300382018-02-08 Wilco Dijkstra <wdijkstr@arm.com> 30039 30040 PR rtl-optimization/84068 30041 PR rtl-optimization/83459 30042 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting. 30043 300442018-02-08 Aldy Hernandez <aldyh@redhat.com> 30045 30046 PR tree-optimization/84224 30047 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert. 30048 * calls.c (gimple_alloca_call_p): Only return TRUE when we have 30049 non-zero arguments. 30050 300512018-02-07 Iain Sandoe <iain@codesourcery.com> 30052 30053 PR target/84113 30054 * config/rs6000/altivec.md (*restore_world): Remove LR use. 30055 * config/rs6000/predicates.md (restore_world_operation): Adjust op 30056 count, remove one USE. 30057 300582018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com> 30059 30060 * doc/install.texi (Configuration): Document the 30061 --with-long-double-format={ibm,ieee} PowerPC configuration 30062 options. 30063 30064 PR target/84154 30065 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2): 30066 Convert from define_expand to be define_insn_and_split. Rework 30067 float/double/_Float128 conversions to QI/HI/SImode to work with 30068 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where 30069 conversions to QI/HImode types did a store and then a load to 30070 truncate the value. For conversions to VSX registers, don't split 30071 the insn, instead emit the code directly. Use the code iterator 30072 any_fix to combine signed and unsigned conversions. 30073 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise. 30074 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise. 30075 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise. 30076 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise. 30077 (fix_<mode>di2_hw): Likewise. 30078 (fixuns_<mode>di2_hw): Likewise. 30079 (fix_<mode>si2_hw): Likewise. 30080 (fixuns_<mode>si2_hw): Likewise. 30081 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise. 30082 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise. 30083 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to 30084 fix<uns>_trunc<SFDF:mode>si2_p8. 30085 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used. 30086 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise. 30087 (fix<uns>_<mode>_mem): Likewise. 30088 (fctiw<u>z_<mode>_mem): Likewise. 30089 (fix<uns>_<mode>_mem): Likewise. 30090 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent 30091 the register allocator from doing a direct move to the GPRs to do 30092 a store, and instead use the ISA 3.0 store byte/half-word from 30093 vector register instruction. For IEEE 128-bit floating point, 30094 also optimize stores of 32-bit ints. 30095 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise. 30096 300972018-02-07 Alan Hayward <alan.hayward@arm.com> 30098 30099 * genextract.c (push_pathstr_operand): New function to support 30100 [a-zA-Z]. 30101 (walk_rtx): Call push_pathstr_operand. 30102 (print_path): Support [a-zA-Z]. 30103 301042018-02-07 Richard Biener <rguenther@suse.de> 30105 30106 PR tree-optimization/84037 30107 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member. 30108 (cse_and_gimplify_to_preheader): Declare. 30109 (vect_get_place_in_interleaving_chain): Likewise. 30110 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize 30111 ivexpr_map. 30112 (_loop_vec_info::~_loop_vec_info): Delete it. 30113 (cse_and_gimplify_to_preheader): New function. 30114 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export. 30115 * tree-vect-stmts.c (vectorizable_store): CSE base and steps. 30116 (vectorizable_load): Likewise. For grouped stores always base 30117 the IV on the first element. 30118 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning 30119 condition before gimplifying. 30120 301212018-02-07 Jakub Jelinek <jakub@redhat.com> 30122 30123 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for 30124 *DIV_EXPR and *MOD_EXPR. 30125 301262018-02-07 H.J. Lu <hongjiu.lu@intel.com> 30127 30128 PR target/84248 30129 * config/i386/i386.c (ix86_option_override_internal): Mask out 30130 the CF_SET bit when checking -fcf-protection. 30131 301322018-02-07 Tom de Vries <tom@codesourcery.com> 30133 30134 PR libgomp/84217 30135 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large 30136 enough. 30137 301382018-02-07 Richard Biener <rguenther@suse.de> 30139 30140 PR tree-optimization/84204 30141 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in 30142 this place. 30143 30144 PR tree-optimization/84205 30145 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also 30146 special-case isl_ast_op_zdiv_r. 30147 30148 PR tree-optimization/84223 30149 * graphite-scop-detection.c (gather_bbs::before_dom_children): 30150 Only add conditions from within the region. 30151 (gather_bbs::after_dom_children): Adjust. 30152 301532018-02-07 Georg-Johann Lay <avr@gjlay.de> 30154 30155 PR target/84209 30156 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros. 30157 * config/avr/avr.md: Only post-reload split REG-REG moves if 30158 either register is GENERAL_REG_P. 30159 301602018-02-07 Jakub Jelinek <jakub@redhat.com> 30161 30162 PR tree-optimization/84235 30163 * tree-ssa-scopedtables.c 30164 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt 30165 if the subtraction is performed in floating point type where NaNs are 30166 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't 30167 build 1. Formatting fix. 30168 301692018-02-06 Jakub Jelinek <jakub@redhat.com> 30170 30171 PR target/84146 30172 * config/i386/i386.c (rest_of_insert_endbranch): Only skip 30173 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else, 30174 and skip it regardless of bb boundaries. Use CALL_P macro, 30175 don't test INSN_P (insn) together with CALL_P or JUMP_P check 30176 unnecessarily, formatting fix. 30177 301782018-02-06 Michael Collison <michael.collison@arm.com> 30179 30180 * config/arm/thumb2.md: 30181 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it. 30182 (*thumb_mov_notscc): Ditto. 30183 301842018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com> 30185 30186 PR target/84154 30187 * config/rs6000/rs6000.md (su code attribute): Use "u" for 30188 unsigned_fix, not "s". 30189 301902018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE> 30191 30192 * configure.ac (gcc_fn_eh_frame_ro): New function. 30193 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for 30194 correct .eh_frame permissions. 30195 * configure: Regenerate. 30196 301972018-02-06 Andrew Jenner <andrew@codeourcery.com> 30198 30199 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove 30200 irrelevant options. 30201 302022018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com> 30203 30204 * config/rs6000/rs6000.c (rs6000_option_override_internal): 30205 Display warning message for -mno-speculate-indirect-jumps. 30206 302072018-02-06 Andrew Jenner <andrew@codesourcery.com> 30208 30209 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add 30210 Undocumented. 30211 * config/powerpcspe/sysv4.opt (mbit-align): Likewise. 30212 302132018-02-06 Aldy Hernandez <aldyh@redhat.com> 30214 30215 PR tree-optimization/84225 30216 * tree-eh.c (find_trapping_overflow): Only call 30217 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P. 30218 302192018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com> 30220 30221 PR target/84145 30222 * config/i386/i386.c: Reimplement the check of possible options 30223 -mibt/-mshstk conbination. Change error messages. 30224 * doc/invoke.texi: Fix a typo: remove extra '='. 30225 302262018-02-06 Marek Polacek <polacek@redhat.com> 30227 30228 PR tree-optimization/84228 30229 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements. 30230 302312018-02-06 Tamar Christina <tamar.christina@arm.com> 30232 30233 PR target/82641 30234 * config/arm/arm.c (arm_print_asm_arch_directives): Record already 30235 emitted arch directives. 30236 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and 30237 __ARM_FEATURE_COPROC before changing architectures. 30238 302392018-02-06 Richard Biener <rguenther@suse.de> 30240 30241 * config/i386/i386.c (print_reg): Fix typo. 30242 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll. 30243 302442018-02-06 Eric Botcazou <ebotcazou@adacore.com> 30245 30246 * configure: Regenerate. 30247 302482018-02-05 Martin Sebor <msebor@redhat.com> 30249 30250 PR tree-optimization/83369 30251 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print 30252 inlining context. 30253 302542018-02-05 Martin Liska <mliska@suse.cz> 30255 30256 * doc/invoke.texi: Cherry-pick upstream r323995. 30257 302582018-02-05 Richard Sandiford <richard.sandiford@linaro.org> 30259 30260 * ira.c (ira_init_register_move_cost): Adjust comment. 30261 302622018-02-05 Martin Liska <mliska@suse.cz> 30263 30264 PR gcov-profile/84137 30265 * doc/gcov.texi: Fix typo in documentation. 30266 302672018-02-05 Martin Liska <mliska@suse.cz> 30268 30269 PR gcov-profile/83879 30270 * doc/gcov.texi: Document necessity of --dynamic-list-data when 30271 using dlopen functionality. 30272 302732018-02-05 Olga Makhotina <olga.makhotina@intel.com> 30274 30275 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd, 30276 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss, 30277 _mm_maskz_range_ss, _mm_mask_range_round_ss, 30278 _mm_maskz_range_round_ss): New intrinsics. 30279 (__builtin_ia32_rangesd128_round) 30280 (__builtin_ia32_rangess128_round): Remove. 30281 (__builtin_ia32_rangesd128_mask_round, 30282 __builtin_ia32_rangess128_mask_round): New builtins. 30283 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round, 30284 __builtin_ia32_rangess128_round): Remove. 30285 (__builtin_ia32_rangesd128_mask_round, 30286 __builtin_ia32_rangess128_mask_round): New builtins. 30287 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ... 30288 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this. 30289 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>" 30290 "<round_saeonly_constraint>")): Changed to ... 30291 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>" 30292 "<round_saeonly_scalar_constraint>")): ... this. 30293 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0| 30294 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ... 30295 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2, 30296 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1, 30297 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this. 30298 302992018-02-02 Andrew Jenner <andrew@codesourcery.com> 30300 30301 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant 30302 options. 30303 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value): 30304 Remove all values except native, 8540 and 8548. 30305 303062018-02-02 H.J. Lu <hongjiu.lu@intel.com> 30307 30308 * config/i386/i386.c (ix86_output_function_return): Pass 30309 INVALID_REGNUM, instead of -1, as invalid register number to 30310 indirect_thunk_name and output_indirect_thunk. 30311 303122018-02-02 Julia Koval <julia.koval@intel.com> 30313 30314 * config.gcc: Add -march=icelake. 30315 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake. 30316 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake. 30317 * config/i386/i386.c (processor_costs): Add m_ICELAKE. 30318 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2, 30319 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New. 30320 (processor_target_table): Add icelake. 30321 (ix86_option_override_internal): Handle new PTAs. 30322 (get_builtin_code_for_version): Handle icelake. 30323 (M_INTEL_COREI7_ICELAKE): New. 30324 (fold_builtin_cpu): Handle icelake. 30325 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New. 30326 * doc/invoke.texi: Add -march=icelake. 30327 303282018-02-02 Julia Koval <julia.koval@intel.com> 30329 30330 * config/i386/i386.c (ix86_option_override_internal): Change flags type 30331 to wide_int_bitmask. 30332 * wide-int-bitmask.h: New. 30333 303342018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com> 30335 30336 PR target/84066 30337 * config/i386/i386.md: Replace Pmode with word_mode in 30338 builtin_setjmp_setup and builtin_longjmp to support x32. 30339 303402018-02-01 Peter Bergner <bergner@vnet.ibm.com> 30341 30342 PR target/56010 30343 PR target/83743 30344 * config/rs6000/driver-rs6000.c: #include "diagnostic.h". 30345 #include "opts.h". 30346 (rs6000_supported_cpu_names): New static variable. 30347 (linux_cpu_translation_table): Likewise. 30348 (elf_platform) <cpu>: Define new static variable and use it. 30349 Translate kernel AT_PLATFORM name to canonical name if needed. 30350 Error if platform name is unknown. 30351 303522018-02-01 Aldy Hernandez <aldyh@redhat.com> 30353 30354 PR target/84089 30355 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode. 30356 303572018-02-01 Jeff Law <law@redhat.com> 30358 30359 PR target/84128 30360 * config/i386/i386.c (release_scratch_register_on_entry): Add new 30361 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore 30362 the scratch if RELEASE_VIA_POP is false. 30363 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE. 30364 If we have to save a temporary register, decrement SIZE appropriately. 30365 Pass new arguments to release_scratch_register_on_entry. 30366 (ix86_adjust_stack_and_probe): Likewise. 30367 (ix86_emit_probe_stack_range): Pass new arguments to 30368 release_scratch_register_on_entry. 30369 303702018-02-01 Uros Bizjak <ubizjak@gmail.com> 30371 30372 PR rtl-optimization/84157 30373 * combine.c (change_zero_ext): Use REG_P predicate in 30374 front of HARD_REGISTER_P predicate. 30375 303762018-02-01 Georg-Johann Lay <avr@gjlay.de> 30377 30378 * config/avr/avr.c (avr_option_override): Move disabling of 30379 -fdelete-null-pointer-checks to... 30380 * common/config/avr/avr-common.c (avr_option_optimization_table): 30381 ...here. 30382 303832018-02-01 Richard Sandiford <richard.sandiford@linaro.org> 30384 30385 PR tree-optimization/81635 30386 * tree-data-ref.c (split_constant_offset_1): For types that 30387 wrap on overflow, try to use range info to prove that wrapping 30388 cannot occur. 30389 303902018-02-01 Renlin Li <renlin.li@arm.com> 30391 30392 PR target/83370 30393 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle 30394 TAILCALL_ADDR_REGS. 30395 (aarch64_register_move_cost): Likewise. 30396 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to 30397 TAILCALL_ADDR_REGS. 30398 (REG_CLASS_NAMES): Likewise. 30399 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to 30400 TAILCALL_ADDR_REGS. Remove IP registers. 30401 * config/aarch64/aarch64.md (Ucs): Update register constraint. 30402 304032018-02-01 Richard Biener <rguenther@suse.de> 30404 30405 * domwalk.h (dom_walker::dom_walker): Add additional constructor 30406 for specifying RPO order and allow NULL for that. 30407 * domwalk.c (dom_walker::dom_walker): Likewise. 30408 (dom_walker::walk): Handle NULL RPO order. 30409 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children 30410 in RPO order. 30411 (rewrite_update_dom_walker): Likewise. 30412 (mark_def_dom_walker): Likewise. 30413 304142018-02-01 Richard Sandiford <richard.sandiford@linaro.org> 30415 30416 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move) 30417 (aarch64_maybe_expand_sve_subreg_move): Declare. 30418 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec. 30419 * config/aarch64/predicates.md (aarch64_any_register_operand): New 30420 predicate. 30421 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves 30422 that are semantically a reverse operation. 30423 (*aarch64_sve_mov<mode>_subreg_be): New pattern. 30424 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move): 30425 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New 30426 functions. 30427 (aarch64_can_change_mode_class): For big-endian, forbid changes 30428 between two SVE modes if they have different element sizes. 30429 304302018-02-01 Richard Sandiford <richard.sandiford@linaro.org> 30431 30432 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer 30433 the TImode handling for big-endian targets. 30434 304352018-02-01 Richard Sandiford <richard.sandiford@linaro.org> 30436 30437 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with... 30438 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes, 30439 not just bytes. 30440 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate): 30441 Remove BSWAP handing for big-endian targets and use the form of 30442 LD1RQ appropariate for the mode. 30443 304442018-02-01 Richard Sandiford <richard.sandiford@linaro.org> 30445 30446 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle 30447 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single 30448 duplicated element. 30449 304502018-02-01 Richard Sandiford <richard.sandiford@linaro.org> 30451 30452 PR tearget/83845 30453 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten 30454 check for operands that need to go through aarch64_sve_reload_be. 30455 304562018-02-01 Jakub Jelinek <jakub@redhat.com> 30457 30458 PR tree-optimization/81661 30459 PR tree-optimization/84117 30460 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare. 30461 * tree-eh.c: Include gimplify.h. 30462 (find_trapping_overflow, replace_trapping_overflow, 30463 rewrite_to_non_trapping_overflow): New functions. 30464 * tree-vect-loop.c: Include tree-eh.h. 30465 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow. 30466 * tree-data-ref.c: Include tree-eh.h. 30467 (get_segment_min_max): Use rewrite_to_non_trapping_overflow. 30468 304692018-01-31 Uros Bizjak <ubizjak@gmail.com> 30470 30471 PR rtl-optimization/84123 30472 * combine.c (change_zero_ext): Check if hard register satisfies 30473 can_change_dest_mode before calling gen_lowpart_SUBREG. 30474 304752018-01-31 Vladimir Makarov <vmakarov@redhat.com> 30476 30477 PR target/82444 30478 * ira.c (ira_init_register_move_cost): Remove assert. 30479 304802018-01-31 Eric Botcazou <ebotcazou@adacore.com> 30481 30482 PR rtl-optimization/84071 30483 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case. 30484 * doc/tm.texi: Regenerate. 30485 304862018-01-31 Richard Biener <rguenther@suse.de> 30487 30488 PR tree-optimization/84132 30489 * tree-data-ref.c (analyze_miv_subscript): Properly 30490 check whether evolution_function_is_affine_multivariate_p 30491 before calling gcd_of_steps_may_divide_p. 30492 304932018-01-31 Julia Koval <julia.koval@intel.com> 30494 30495 PR target/83618 30496 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID. 30497 * config/i386/i386.md (rdpid_rex64) New. 30498 (rdpid): Make 32bit only. 30499 305002018-01-29 Aldy Hernandez <aldyh@redhat.com> 30501 30502 PR lto/84105 30503 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with 30504 an IDENTIFIER_NODE for FUNCTION_TYPE's. 30505 305062018-01-31 Eric Botcazou <ebotcazou@adacore.com> 30507 30508 Revert 30509 2018-01-12 Eric Botcazou <ebotcazou@adacore.com> 30510 30511 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register. 30512 305132018-01-31 Eric Botcazou <ebotcazou@adacore.com> 30514 30515 PR rtl-optimization/84071 30516 * combine.c (record_dead_and_set_regs_1): Record the source unmodified 30517 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target. 30518 305192018-01-31 Claudiu Zissulescu <claziss@synopsys.com> 30520 30521 * config/arc/arc.c (arc_handle_aux_attribute): New function. 30522 (arc_attribute_table): Add 'aux' attribute. 30523 (arc_in_small_data_p): Consider aux like variables. 30524 (arc_is_aux_reg_p): New function. 30525 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables. 30526 (arc_get_aux_arg): New function. 30527 (prepare_move_operands): Handle aux-register access. 30528 (arc_handle_aux_attribute): New function. 30529 * doc/extend.texi (ARC Variable attributes): Add subsection. 30530 305312018-01-31 Claudiu Zissulescu <claziss@synopsys.com> 30532 30533 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto. 30534 * config/arc/arc.c (arc_handle_uncached_attribute): New function. 30535 (arc_attribute_table): Add 'uncached' attribute. 30536 (arc_print_operand): Print '.di' flag for uncached memory 30537 accesses. 30538 (arc_in_small_data_p): Do not consider for small data the uncached 30539 types. 30540 (arc_is_uncached_mem_p): New function. 30541 * config/arc/predicates.md (compact_store_memory_operand): Check 30542 for uncached memory accesses. 30543 (nonvol_nonimm_operand): Likewise. 30544 * doc/extend.texi (ARC Type Attribute): New subsection. 30545 305462018-01-31 Jakub Jelinek <jakub@redhat.com> 30547 30548 PR c/84100 30549 * common.opt (falign-functions=, falign-jumps=, falign-labels=, 30550 falign-loops=): Add Optimization flag. 30551 305522018-01-30 Jeff Law <law@redhat.com> 30553 30554 PR target/84064 30555 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument 30556 INT_REGISTERS_SAVED. Check it prior to calling 30557 get_scratch_register_on_entry. 30558 (ix86_adjust_stack_and_probe): Similarly. 30559 (ix86_emit_probe_stack_range): Similarly. 30560 (ix86_expand_prologue): Corresponding changes. 30561 305622018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE> 30563 30564 PR target/40411 30565 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*, 30566 -std=iso9899:199409 instead of -pedantic to select values-Xc.o. 30567 305682018-01-30 Vladimir Makarov <vmakarov@redhat.com> 30569 30570 PR target/84112 30571 * lra-constraints.c (curr_insn_transform): Process AND in the 30572 address. 30573 305742018-01-30 Jakub Jelinek <jakub@redhat.com> 30575 30576 PR rtl-optimization/83986 30577 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti 30578 dependence against last_pending_memory_flush in addition to 30579 pending_jump_insns. 30580 305812018-01-30 Alexandre Oliva <aoliva@redhat.com> 30582 30583 PR tree-optimization/81611 30584 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening 30585 copies. 30586 305872018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com> 30588 30589 PR target/83758 30590 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return 30591 a reg rtx. 30592 305932018-01-30 Richard Biener <rguenther@suse.de> 30594 Jakub Jelinek <jakub@redhat.com> 30595 30596 PR tree-optimization/84111 30597 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip 30598 inner loops added during recursion, as they don't have up-to-date 30599 SSA form. 30600 306012018-01-30 Jan Hubicka <hubicka@ucw.cz> 30602 30603 PR ipa/81360 30604 * ipa-inline.c (can_inline_edge_p): Break out late tests to... 30605 (can_inline_edge_by_limits_p): ... here. 30606 (can_early_inline_edge_p, check_callers, 30607 update_caller_keys, update_callee_keys, recursive_inlining, 30608 add_new_edges_to_heap, speculation_useful_p, 30609 inline_small_functions, 30610 inline_small_functions, flatten_function, 30611 inline_to_all_callers_1): Update. 30612 306132018-01-30 Jan Hubicka <hubicka@ucw.cz> 30614 30615 * profile-count.c (profile_count::combine_with_ipa_count): Handle 30616 zeros correctly. 30617 306182018-01-30 Richard Biener <rguenther@suse.de> 30619 30620 PR tree-optimization/83008 30621 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost 30622 invariant and constant vector uses in stmts when they need 30623 more than one stmt. 30624 306252018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE> 30626 30627 PR bootstrap/84017 30628 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86. 30629 * configure: Regenerate. 30630 306312018-01-30 Richard Sandiford <richard.sandiford@linaro.org> 30632 30633 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New 30634 pattern. 30635 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number. 30636 Use gen_rtx_REG rather than gen_lowpart. 30637 306382018-01-30 Richard Sandiford <richard.sandiford@linaro.org> 30639 30640 * lra-constraints.c (match_reload): Use subreg_lowpart_offset 30641 rather than 0 when creating partial subregs. 30642 306432018-01-30 Richard Sandiford <richard.sandiford@linaro.org> 30644 30645 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples 30646 of usage. 30647 306482018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com> 30649 30650 PR target/81550 30651 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode 30652 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode, 30653 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY 30654 flags. This restores the settings used before the 2017-07-24. 30655 Turning off pre increment/decrement/modify allows IVOPTS to 30656 optimize DF/SF loops where the index is an int. 30657 306582018-01-29 Richard Biener <rguenther@suse.de> 30659 Kelvin Nilsen <kelvin@gcc.gnu.org> 30660 30661 PR bootstrap/80867 30662 * tree-vect-stmts.c (vectorizable_call): Don't call 30663 targetm.vectorize_builtin_md_vectorized_function if callee is 30664 NULL. 30665 306662018-01-22 Carl Love <cel@us.ibm.com> 30667 30668 * doc/extend.tex: Fix typo in second arg in 30669 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}. 30670 306712018-01-29 Richard Biener <rguenther@suse.de> 30672 30673 PR tree-optimization/84086 30674 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h. 30675 (flush_ssaname_freelist): When SSA names were released reset 30676 the SCEV hash table. 30677 306782018-01-29 Richard Biener <rguenther@suse.de> 30679 30680 PR tree-optimization/84057 30681 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already 30682 removed paths when removing edges. 30683 306842018-01-27 H.J. Lu <hongjiu.lu@intel.com> 30685 30686 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with 30687 -mfunction-return=@var{choice}. 30688 306892018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de> 30690 30691 PR diagnostic/84034 30692 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace): 30693 Handle CR like TAB. 30694 (layout::print_source_line): Likewise. 30695 (test_get_line_width_without_trailing_whitespace): Add test cases. 30696 306972018-01-27 Jakub Jelinek <jakub@redhat.com> 30698 30699 PR middle-end/84040 30700 * sched-deps.c (sched_macro_fuse_insns): Return immediately for 30701 debug insns. 30702 307032018-01-26 Jim Wilson <jimw@sifive.com> 30704 30705 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New. 30706 30707 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs 30708 specified. 30709 307102018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 30711 30712 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS 30713 and CMP + SUB-immediate -> SUBS. 30714 307152018-01-26 Martin Sebor <msebor@redhat.com> 30716 30717 PR tree-optimization/83896 30718 * tree-ssa-strlen.c (get_string_len): Rename... 30719 (get_string_cst_length): ...to this. Return HOST_WIDE_INT. 30720 Avoid assuming length is constant. 30721 (handle_char_store): Use HOST_WIDE_INT for string length. 30722 307232018-01-26 Uros Bizjak <ubizjak@gmail.com> 30724 30725 PR target/81763 30726 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber 30727 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives. 30728 307292018-01-26 Richard Biener <rguenther@suse.de> 30730 30731 PR rtl-optimization/84003 30732 * dse.c (record_store): Only record redundant stores when 30733 the earlier store aliases at least all accesses the later one does. 30734 307352018-01-26 Jakub Jelinek <jakub@redhat.com> 30736 30737 PR rtl-optimization/83985 30738 * dce.c (deletable_insn_p): Return false for separate shrink wrapping 30739 REG_CFA_RESTORE insns. 30740 (delete_unmarked_insns): Don't ignore separate shrink wrapping 30741 REG_CFA_RESTORE insns here. 30742 30743 PR c/83989 30744 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't 30745 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR. 30746 307472018-01-26 Claudiu Zissulescu <claziss@synopsys.com> 30748 30749 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3. 30750 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... . 30751 (arc_init): Likewise. 30752 (arc_override_options): Likewise. 30753 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune 30754 value. 30755 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn 30756 support. 30757 * config/arc/arc.h (TARGET_DBNZ): Define. 30758 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to 30759 properly set the tune attribute. 30760 (dbnz): Use TARGET_DBNZ guard. 30761 * config/arc/arc.opt (mtune): Add core3 option. 30762 307632018-01-26 Claudiu Zissulescu <claziss@synopsys.com> 30764 30765 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to 30766 recognize new pic like addresses. 30767 (arc_delegitimize_address): Clean up. 30768 307692018-01-26 Claudiu Zissulescu <claziss@synopsys.com> 30770 30771 * config/arc/arc-arches.def: Option mrf16 valid for all 30772 architectures. 30773 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro. 30774 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on. 30775 * config/arc/arc-options.def (FL_RF16): Add mrf16 option. 30776 * config/arc/arc-tables.opt: Regenerate. 30777 * config/arc/arc.c (arc_conditional_register_usage): Handle 30778 reduced register file case. 30779 (arc_file_start): Set must have build attributes. 30780 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using 30781 mrf16 option value. 30782 * config/arc/arc.opt (mrf16): Add new option. 30783 * config/arc/elf.h (ATTRIBUTE_PCS): Define. 30784 * config/arc/genmultilib.awk: Handle new mrf16 option. 30785 * config/arc/linux.h (ATTRIBUTE_PCS): Define. 30786 * config/arc/t-multilib: Regenerate. 30787 * doc/invoke.texi (ARC Options): Document mrf16 option. 30788 307892018-01-26 Claudiu Zissulescu <claziss@synopsys.com> 30790 30791 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto. 30792 * config/arc/arc.c (arc_handle_secure_attribute): New function. 30793 (arc_attribute_table): Add 'secure_call' attribute. 30794 (arc_print_operand): Print secure call operand. 30795 (arc_function_ok_for_sibcall): Don't optimize tail calls when 30796 secure. 30797 (arc_is_secure_call_p): New function. * config/arc/arc.md 30798 (call_i): Add support for sjli instruction. 30799 (call_value_i): Likewise. 30800 * config/arc/constraints.md (Csc): New constraint. 30801 308022018-01-26 Claudiu Zissulescu <claziss@synopsys.com> 30803 John Eric Martin <John.Martin@emmicro-us.com> 30804 30805 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto. 30806 * config/arc/arc.c (_arc_jli_section): New struct. 30807 (arc_jli_section): New type. 30808 (rc_jli_sections): New static variable. 30809 (arc_handle_jli_attribute): New function. 30810 (arc_attribute_table): Add jli_always and jli_fixed attribute. 30811 (arc_file_end): New function. 30812 (TARGET_ASM_FILE_END): Define. 30813 (arc_print_operand): Reuse 'S' letter for JLI output instruction. 30814 (arc_add_jli_section): New function. 30815 (jli_call_scan): Likewise. 30816 (arc_reorg): Call jli_call_scan. 30817 (arc_output_addsi): Remove 'S' from printing asm operand. 30818 (arc_is_jli_call_p): New function. 30819 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm 30820 operand. 30821 (movhi_insn): Likewise. 30822 (movsi_insn): Likewise. 30823 (movsi_set_cc_insn): Likewise. 30824 (loadqi_update): Likewise. 30825 (load_zeroextendqisi_update): Likewise. 30826 (load_signextendqisi_update): Likewise. 30827 (loadhi_update): Likewise. 30828 (load_zeroextendhisi_update): Likewise. 30829 (load_signextendhisi_update): Likewise. 30830 (loadsi_update): Likewise. 30831 (loadsf_update): Likewise. 30832 (movsicc_insn): Likewise. 30833 (bset_insn): Likewise. 30834 (bxor_insn): Likewise. 30835 (bclr_insn): Likewise. 30836 (bmsk_insn): Likewise. 30837 (bicsi3_insn): Likewise. 30838 (cmpsi_cc_c_insn): Likewise. 30839 (movsi_ne): Likewise. 30840 (movsi_cond_exec): Likewise. 30841 (clrsbsi2): Likewise. 30842 (norm_f): Likewise. 30843 (normw): Likewise. 30844 (swap): Likewise. 30845 (divaw): Likewise. 30846 (flag): Likewise. 30847 (sr): Likewise. 30848 (kflag): Likewise. 30849 (ffs): Likewise. 30850 (ffs_f): Likewise. 30851 (fls): Likewise. 30852 (call_i): Remove 'S' asm letter, add jli instruction. 30853 (call_value_i): Likewise. 30854 * config/arc/arc.op (mjli-always): New option. 30855 * config/arc/constraints.md (Cji): New constraint. 30856 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm 30857 operand. 30858 (subsf3_fpx): Likewise. 30859 (mulsf3_fpx): Likewise. 30860 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing 30861 asm operand. 30862 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed' 30863 function attrbutes. 30864 * doc/invoke.texi (ARC): Document mjli-always option. 30865 308662018-01-26 Sebastian Perta <sebastian.perta@renesas.com> 30867 30868 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const 30869 avoid addition with 0 and use incw and decw where possible. 30870 308712018-01-26 Richard Biener <rguenther@suse.de> 30872 30873 PR tree-optimization/81082 30874 * fold-const.c (fold_plusminus_mult_expr): Do not perform the 30875 association if it requires casting to unsigned. 30876 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived 30877 from fold_plusminus_mult_expr to catch important cases late when 30878 range info is available. 30879 308802018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE> 30881 30882 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove. 30883 * configure.ac (hidden_linkonce): New test. 30884 * configure: Regenerate. 30885 * config.in: Regenerate. 30886 308872018-01-26 Julia Koval <julia.koval@intel.com> 30888 30889 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask, 30890 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask, 30891 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask, 30892 _mm_mask_bitshuffle_epi64_mask): Fix type. 30893 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI, 30894 USI_FTYPE_V4DI_V4DI_USI): Remove. 30895 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask, 30896 __builtin_ia32_vpshufbitqmb256_mask, 30897 __builtin_ia32_vpshufbitqmb128_mask): Fix types. 30898 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types. 30899 * config/i386/sse.md (VI1_AVX512VLBW): Change types. 30900 309012018-01-26 Alan Modra <amodra@gmail.com> 30902 30903 PR target/84033 30904 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude 30905 UNSPEC_VBPERMQ. Sort other unspecs. 30906 309072018-01-25 David Edelsohn <dje.gcc@gmail.com> 30908 30909 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type. 30910 309112018-01-25 Jan Hubicka <hubicka@ucw.cz> 30912 30913 PR middle-end/83055 30914 * predict.c (drop_profile): Do not push/pop cfun; update also 30915 node->count. 30916 (handle_missing_profiles): Fix logic looking for zero profiles. 30917 309182018-01-25 Jakub Jelinek <jakub@redhat.com> 30919 30920 PR middle-end/83977 30921 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature 30922 on functions with #pragma omp declare simd or functions with simd 30923 attribute. 30924 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change. 30925 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen): 30926 Remove trailing \n from warning_at calls. 30927 309282018-01-25 Tom de Vries <tom@codesourcery.com> 30929 30930 PR target/84028 30931 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call 30932 for neutered workers. 30933 309342018-01-24 Joseph Myers <joseph@codesourcery.com> 30935 30936 PR target/68467 30937 * config/m68k/m68k.c (m68k_promote_function_mode): New function. 30938 (TARGET_PROMOTE_FUNCTION_MODE): New macro. 30939 309402018-01-24 Jeff Law <law@redhat.com> 30941 30942 PR target/83994 30943 * i386.c (get_probe_interval): Move to earlier point. 30944 (ix86_compute_frame_layout): If -fstack-clash-protection and 30945 the frame is larger than the probe interval, then use pushes 30946 to save registers rather than reg->mem moves. 30947 (ix86_expand_prologue): Remove conditional for int_registers_saved 30948 assertion. 30949 309502018-01-24 Vladimir Makarov <vmakarov@redhat.com> 30951 30952 PR target/84014 30953 * ira-build.c (setup_min_max_allocno_live_range_point): Set up 30954 min/max for never referenced object. 30955 309562018-01-24 Jakub Jelinek <jakub@redhat.com> 30957 30958 PR middle-end/83977 30959 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN 30960 here. 30961 * omp-low.c (create_omp_child_function): Remove "omp declare simd" 30962 attributes from DECL_ATTRIBUTES (decl) without affecting 30963 DECL_ATTRIBUTES (current_function_decl). 30964 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL 30965 functions with non-NULL DECL_ABSTRACT_ORIGIN. 30966 309672018-01-24 Richard Sandiford <richard.sandiford@linaro.org> 30968 30969 PR tree-optimization/83979 30970 * fold-const.c (fold_comparison): Use constant_boolean_node 30971 instead of boolean_{true,false}_node. 30972 309732018-01-24 Jan Hubicka <hubicka@ucw.cz> 30974 30975 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls 30976 with zero counts. 30977 309782018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com> 30979 30980 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>): 30981 Simplify the clause that sets the length attribute. 30982 (*call_value_indirect_nonlocal_sysv<mode>): Likewise. 30983 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the 30984 clause that sets the length attribute. 30985 (*sibcall_value_nonlocal_sysv<mode>): Likewise. 30986 309872018-01-24 Tom de Vries <tom@codesourcery.com> 30988 30989 PR target/83589 30990 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1. 30991 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c. 30992 Add strict parameter. 30993 (prevent_branch_around_nothing): Insert dummy insn between branch to 30994 label and label with no ptx insn inbetween. 30995 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn. 30996 309972018-01-24 Tom de Vries <tom@codesourcery.com> 30998 30999 PR target/81352 31000 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call 31001 for neutered threads in warp. 31002 * config/nvptx/nvptx.md (define_insn "exit"): New insn. 31003 310042018-01-24 Richard Biener <rguenther@suse.de> 31005 31006 PR tree-optimization/83176 31007 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. } 31008 operands. 31009 310102018-01-24 Richard Biener <rguenther@suse.de> 31011 31012 PR tree-optimization/82819 31013 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid 31014 code generating pluses that are no-ops in the target precision. 31015 310162018-01-24 Richard Biener <rguenther@suse.de> 31017 31018 PR middle-end/84000 31019 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind. 31020 310212018-01-23 Jan Hubicka <hubicka@ucw.cz> 31022 31023 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count 31024 to merge probabilities. 31025 * predict.c (probably_never_executed): Also mark as cold functions 31026 with global 0 profile and guessed local profile. 31027 * profile-count.c (profile_probability::combine_with_count): New 31028 member function. 31029 * profile-count.h (profile_probability::operator*, 31030 profile_probability::operator*=, profile_probability::operator/, 31031 profile_probability::operator/=): Reduce precision to adjusted 31032 and set value to guessed on contradictory divisions. 31033 (profile_probability::combine_with_freq): Remove. 31034 (profile_probability::combine_wiht_count): Declare. 31035 (profile_count::force_nonzero):: Set to adjusted. 31036 (profile_count::probability_in):: Set quality to adjusted. 31037 * tree-ssa-tail-merge.c (replace_block_by): Use 31038 combine_with_count. 31039 310402018-01-23 Andrew Waterman <andrew@sifive.com> 31041 Jim Wilson <jimw@sifive.com> 31042 31043 * config/riscv/riscv.c (riscv_stack_boundary): New. 31044 (riscv_option_override): Set riscv_stack_boundary. Handle 31045 riscv_preferred_stack_boundary_arg. 31046 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New. 31047 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY. 31048 (STACK_BOUNDARY): Set to riscv_stack_boundary. 31049 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY. 31050 * config/riscv/riscv.opt (mpreferred-stack-boundary): New. 31051 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary. 31052 310532018-01-23 H.J. Lu <hongjiu.lu@intel.com> 31054 31055 PR target/83905 31056 * config/i386/i386.c (ix86_expand_prologue): Use cost reference 31057 of struct ix86_frame. 31058 (ix86_expand_epilogue): Likewise. Add a local variable for 31059 the reg_save_offset field in struct ix86_frame. 31060 310612018-01-23 Bin Cheng <bin.cheng@arm.com> 31062 31063 PR tree-optimization/82604 31064 * tree-loop-distribution.c (enum partition_kind): New enum item 31065 PKIND_PARTIAL_MEMSET. 31066 (partition_builtin_p): Support above new enum item. 31067 (generate_code_for_partition): Ditto. 31068 (compute_access_range): Differentiate cases that equality can be 31069 proven at all loops, the innermost loops or no loops. 31070 (classify_builtin_st, classify_builtin_ldst): Adjust call to above 31071 function. Set PKIND_PARTIAL_MEMSET for partition appropriately. 31072 (finalize_partitions, distribute_loop): Don't fuse partition of 31073 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest. 31074 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if 31075 parloop is enabled. 31076 310772018-01-23 Martin Liska <mliska@suse.cz> 31078 31079 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in 31080 order to ignore the predictor. 31081 (PRED_POLYMORPHIC_CALL): Likewise. 31082 (PRED_RECURSIVE_CALL): Likewise. 31083 310842018-01-23 Martin Liska <mliska@suse.cz> 31085 31086 * tree-profile.c (tree_profiling): Print function header to 31087 aware reader which function we are working on. 31088 * value-prof.c (gimple_find_values_to_profile): Do not print 31089 not interesting value histograms. 31090 310912018-01-23 Martin Liska <mliska@suse.cz> 31092 31093 * profile-count.h (enum profile_quality): Add 31094 profile_uninitialized as the first value. Do not number values 31095 as they are zero based. 31096 (profile_count::verify): Update sanity check. 31097 (profile_probability::verify): Likewise. 31098 310992018-01-23 Nathan Sidwell <nathan@acm.org> 31100 31101 * doc/invoke.texi (ffor-scope): Deprecate. 31102 311032018-01-23 David Malcolm <dmalcolm@redhat.com> 31104 31105 PR tree-optimization/83510 31106 * domwalk.c (set_all_edges_as_executable): New function. 31107 (dom_walker::dom_walker): Convert bool param 31108 "skip_unreachable_blocks" to enum reachability. Move setup of 31109 edge flags to set_all_edges_as_executable and only do it when 31110 reachability is REACHABLE_BLOCKS. 31111 * domwalk.h (enum dom_walker::reachability): New enum. 31112 (dom_walker::dom_walker): Convert bool param 31113 "skip_unreachable_blocks" to enum reachability. 31114 (set_all_edges_as_executable): New decl. 31115 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert 31116 from false for "skip_unreachable_blocks" to ALL_BLOCKS for 31117 "reachability". 31118 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise, 31119 but converting true to REACHABLE_BLOCKS. 31120 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise. 31121 * tree-vrp.c 31122 (check_array_bounds_dom_walker::check_array_bounds_dom_walker): 31123 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS. 31124 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to 31125 REACHABLE_BLOCKS. 31126 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable 31127 if check_all_array_refs will be called. 31128 311292018-01-23 David Malcolm <dmalcolm@redhat.com> 31130 31131 * tree.c (selftest::test_location_wrappers): Add more test 31132 coverage. 31133 311342018-01-23 David Malcolm <dmalcolm@redhat.com> 31135 31136 * sbitmap.c (selftest::test_set_range): Fix memory leaks. 31137 (selftest::test_bit_in_range): Likewise. 31138 311392018-01-23 Richard Sandiford <richard.sandiford@linaro.org> 31140 31141 PR testsuite/83888 31142 * doc/sourcebuild.texi (vect_float): Say that the selector 31143 only describes the situation when -funsafe-math-optimizations is on. 31144 (vect_float_strict): Document. 31145 311462018-01-23 Richard Sandiford <richard.sandiford@linaro.org> 31147 31148 PR tree-optimization/83965 31149 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function. 31150 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it 31151 instead of checking only for a reduction. 31152 (vect_recog_widen_sum_pattern): Likewise. 31153 311542018-01-23 Jan Hubicka <hubicka@ucw.cz> 31155 31156 * predict.c (probably_never_executed): Only use precise profile info. 31157 (compute_function_frequency): Skip after inlining hack since we now 31158 have quality checking. 31159 311602018-01-23 Jan Hubicka <hubicka@ucw.cz> 31161 31162 * profile-count.h (profile_probability::very_unlikely, 31163 profile_probability::unlikely, profile_probability::even): Set 31164 precision to guessed. 31165 311662018-01-23 Richard Biener <rguenther@suse.de> 31167 31168 PR tree-optimization/83963 31169 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region): 31170 Properly terminate dominator walk when crossing the exit edge not 31171 when visiting its source block. 31172 311732018-01-23 Jakub Jelinek <jakub@redhat.com> 31174 31175 PR c++/83918 31176 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than 31177 VIEW_CONVERT_EXPR to wrap CONST_DECLs. 31178 311792018-01-22 Jakub Jelinek <jakub@redhat.com> 31180 31181 PR tree-optimization/83957 31182 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove 31183 semicolon after for body surrounded by braces. 31184 31185 PR tree-optimization/83081 31186 * profile-count.h (profile_probability::split): New method. 31187 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>: 31188 Use profile_probability::split. 31189 (do_compare_rtx_and_jump): Fix adjustment of probabilities 31190 when splitting a single conditional jump into 2. 31191 311922018-01-22 David Malcolm <dmalcolm@redhat.com> 31193 31194 PR tree-optimization/69452 31195 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove 31196 decl. 31197 311982018-01-22 Sebastian Perta <sebastian.perta@renesas.com> 31199 31200 * config/rl78/rl78-expand.md (bswaphi2): New define_expand. 31201 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn. 31202 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn. 31203 312042018-01-22 Sebastian Perta <sebastian.perta@renesas.com> 31205 31206 * config/rl78/rl78-protos.h (rl78_split_movdi): New function 31207 declaration. 31208 * config/rl78/rl78.md (movdi): New define_expand. 31209 * config/rl78/rl78.c (rl78_split_movdi): New function. 31210 312112018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com> 31212 31213 PR target/83862 31214 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete, 31215 no longer used. 31216 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise. 31217 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE 31218 128-bit to produce an UNSPEC move to get the double word with the 31219 signbit and then a shift directly to do signbit. 31220 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit 31221 implementation with a new version that just does either a direct 31222 move or a regular move. Move memory interface to separate insns. 31223 Move insns so they are next to the expander. 31224 (signbit<mode>2_dm_mem_be): New combiner insns to combine load 31225 with signbit move. Split big and little endian case. 31226 (signbit<mode>2_dm_mem_le): Likewise. 31227 (signbit<mode>2_dm_<su>ext): Delete, no longer used. 31228 (signbit<mode>2_dm2): Likewise. 31229 312302018-01-22 Sebastian Perta <sebastian.perta@renesas.com> 31231 31232 * config/rl78/rl78.md (anddi3): New define_expand. 31233 312342018-01-22 Sebastian Perta <sebastian.perta@renesas.com> 31235 31236 * config/rl78/rl78.md (umindi3): New define_expand. 31237 312382018-01-22 Sebastian Perta <sebastian.perta@renesas.com> 31239 31240 * config/rl78/rl78.md (smindi3): New define_expand. 31241 312422018-01-22 Sebastian Perta <sebastian.perta@renesas.com> 31243 31244 * config/rl78/rl78.md (smaxdi3): New define_expand. 31245 312462018-01-22 Carl Love <cel@us.ibm.com> 31247 31248 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI, 31249 LVX_V1TI): Add macro expansion. 31250 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument 31251 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST, 31252 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins. 31253 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p); 31254 Change check to determine if the instruction is a byte reversing 31255 entry. Fix typo in comment. 31256 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry 31257 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI. 31258 Add def_builtin calls for new builtins. 31259 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti): 31260 Add define_insn expansion. 31261 312622018-01-22 Sebastian Perta <sebastian.perta@renesas.com> 31263 31264 * config/rl78/rl78.md (umaxdi3): New define_expand. 31265 312662018-01-22 Sebastian Perta <sebastian.perta@renesas.com> 31267 31268 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check 31269 for non-QImode registers. 31270 312712018-01-22 Richard Biener <rguenther@suse.de> 31272 31273 PR tree-optimization/83963 31274 * graphite-scop-detection.c (scop_detection::get_sese): Delay 31275 including the loop exit block. 31276 (scop_detection::merge_sese): Likewise. 31277 (scop_detection::add_scop): Do it here instead. 31278 312792018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 31280 31281 * doc/sourcebuild.texi (arm_softfloat): Document. 31282 312832018-01-21 John David Anglin <danglin@gcc.gnu.org> 31284 31285 PR gcc/77734 31286 * config/pa/pa.c (pa_function_ok_for_sibcall): Use 31287 targetm.binds_local_p instead of TREE_PUBLIC to check local binding. 31288 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check. 31289 312902018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com> 31291 David Edelsohn <dje.gcc@gmail.com> 31292 31293 PR target/83946 31294 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>): 31295 Change "crset eq" to "crset 2". 31296 (*call_value_indirect_nonlocal_sysv<mode>): Likewise. 31297 (*call_indirect_aix<mode>_nospec): Likewise. 31298 (*call_value_indirect_aix<mode>_nospec): Likewise. 31299 (*call_indirect_elfv2<mode>_nospec): Likewise. 31300 (*call_value_indirect_elfv2<mode>_nospec): Likewise. 31301 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2"; 31302 change assembly output from . to $. 31303 (*sibcall_value_nonlocal_sysv<mode>): Likewise. 31304 (indirect_jump<mode>_nospec): Change assembly output from . to $. 31305 (*tablejump<mode>_internal1_nospec): Likewise. 31306 313072018-01-21 Oleg Endo <olegendo@gcc.gnu.org> 31308 31309 PR target/80870 31310 * config/sh/sh_optimize_sett_clrt.cc: 31311 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes. 31312 313132018-01-20 Richard Sandiford <richard.sandiford@linaro.org> 31314 31315 PR tree-optimization/83940 31316 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set 31317 offset_dt to vect_constant_def rather than vect_unknown_def_type. 31318 (vect_check_load_store_mask): Add a mask_dt_out parameter and 31319 use it to pass back the definition type. 31320 (vect_check_store_rhs): Likewise rhs_dt_out. 31321 (vect_build_gather_load_calls): Add a mask_dt argument and use 31322 it instead of a call to vect_is_simple_use. 31323 (vectorizable_store): Update calls to vect_check_load_store_mask 31324 and vect_check_store_rhs. Use the dt returned by the latter instead 31325 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt 31326 instead of calls to vect_is_simple_use. Pass the scalar rather 31327 than the vector operand to vect_is_simple_use when handling 31328 second and subsequent copies of an rhs value. 31329 (vectorizable_load): Update calls to vect_check_load_store_mask 31330 and vect_build_gather_load_calls. Use the cached mask_dt and 31331 gs_info.offset_dt instead of calls to vect_is_simple_use. 31332 313332018-01-20 Jakub Jelinek <jakub@redhat.com> 31334 31335 PR middle-end/83945 31336 * tree-emutls.c: Include gimplify.h. 31337 (lower_emutls_2): New function. 31338 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree 31339 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr 31340 it before further processing. 31341 31342 PR target/83930 31343 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use 31344 UINTVAL (trueop1) instead of INTVAL (op1). 31345 313462018-01-19 Jakub Jelinek <jakub@redhat.com> 31347 31348 PR debug/81570 31349 PR debug/83728 31350 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to 31351 INCOMING_FRAME_SP_OFFSET if not defined. 31352 (scan_trace): Add ENTRY argument. If true and 31353 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET, 31354 emit a note to adjust the CFA offset. 31355 (create_cfi_notes): Adjust scan_trace callers. 31356 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than 31357 INCOMING_FRAME_SP_OFFSET in the CIE. 31358 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define. 31359 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): 31360 Likewise. 31361 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document. 31362 * doc/tm.texi: Regenerated. 31363 313642018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com> 31365 31366 PR rtl-optimization/83147 31367 * lra-constraints.c (remove_inheritance_pseudos): Use 31368 lra_substitute_pseudo_within_insn. 31369 313702018-01-19 Tom de Vries <tom@codesourcery.com> 31371 Cesar Philippidis <cesar@codesourcery.com> 31372 31373 PR target/83920 31374 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround. 31375 313762018-01-19 Cesar Philippidis <cesar@codesourcery.com> 31377 31378 PR target/83790 31379 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address 31380 spaces for function labels. 31381 313822018-01-19 Martin Liska <mliska@suse.cz> 31383 31384 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89. 31385 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78. 31386 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67. 31387 (PRED_OPCODE_POSITIVE): Change from 64 to 59. 31388 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59. 31389 (PRED_CONST_RETURN): Change from 69 to 65. 31390 (PRED_NULL_RETURN): Change from 91 to 71. 31391 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64. 31392 (PRED_LOOP_GUARD): Change from 66 to 73. 31393 313942018-01-19 Martin Liska <mliska@suse.cz> 31395 31396 * predict.c (predict_insn_def): Add new assert. 31397 (struct branch_predictor): Change type to signed integer. 31398 (test_prediction_value_range): Amend test to cover 31399 PROB_UNINITIALIZED. 31400 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant. 31401 (PRED_LOOP_ITERATIONS_GUESSED): Likewise. 31402 (PRED_LOOP_ITERATIONS_MAX): Likewise. 31403 (PRED_LOOP_IV_COMPARE): Likewise. 31404 * predict.h (PROB_UNINITIALIZED): Define new constant. 31405 314062018-01-19 Martin Liska <mliska@suse.cz> 31407 31408 * predict.c (dump_prediction): Add new format for 31409 analyze_brprob.py script which is enabled with -details 31410 suboption. 31411 * profile-count.h (precise_p): New function. 31412 314132018-01-19 Richard Sandiford <richard.sandiford@linaro.org> 31414 31415 PR tree-optimization/83922 31416 * tree-vect-loop.c (vect_verify_full_masking): Return false if 31417 there are no statements that need masking. 31418 (vect_active_double_reduction_p): New function. 31419 (vect_analyze_loop_operations): Use it when handling phis that 31420 are not in the loop header. 31421 314222018-01-19 Richard Sandiford <richard.sandiford@linaro.org> 31423 31424 PR tree-optimization/83914 31425 * tree-vect-loop.c (vectorizable_induction): Don't convert 31426 init_expr or apply the peeling adjustment for inductions 31427 that are nested within the vectorized loop. 31428 314292018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 31430 31431 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic 31432 instead of NEG. 31433 314342018-01-18 Jakub Jelinek <jakub@redhat.com> 31435 31436 PR sanitizer/81715 31437 PR testsuite/83882 31438 * function.h (gimplify_parameters): Add gimple_seq * argument. 31439 * function.c: Include gimple.h and options.h. 31440 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts 31441 for the added local temporaries if needed. 31442 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller, 31443 if there are any parameter cleanups, wrap whole body into a 31444 try/finally with the cleanups. 31445 314462018-01-18 Wilco Dijkstra <wdijkstr@arm.com> 31447 31448 PR target/82964 31449 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): 31450 Use GET_MODE_CLASS for scalar floating point. 31451 314522018-01-18 Jan Hubicka <hubicka@ucw.cz> 31453 31454 PR ipa/82256 31455 patch by PaX Team 31456 * cgraphclones.c (cgraph_node::create_version_clone_with_body): 31457 Fix call of call_cgraph_insertion_hooks. 31458 314592018-01-18 Martin Sebor <msebor@redhat.com> 31460 31461 * doc/invoke.texi (-Wclass-memaccess): Tweak text. 31462 314632018-01-18 Jan Hubicka <hubicka@ucw.cz> 31464 31465 PR ipa/83619 31466 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge 31467 frequencies. 31468 314692018-01-18 Boris Kolpackov <boris@codesynthesis.com> 31470 31471 PR other/70268 31472 * common.opt: (-ffile-prefix-map): New option. 31473 * opts.c (common_handle_option): Defer it. 31474 * opts-global.c (handle_common_deferred_options): Handle it. 31475 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to... 31476 * file-prefix-map.h: New file. 31477 (remap_debug_filename, add_debug_prefix_map): ...here. 31478 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New. 31479 * final.c (debug_prefix_map, add_debug_prefix_map 31480 remap_debug_filename): Move to... 31481 * file-prefix-map.c: New file. 31482 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename, 31483 generalize, get rid of alloca(), use strrchr() instead of strchr(). 31484 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map): 31485 Implement in terms of add_prefix_map(). 31486 (remap_macro_filename, remap_debug_filename): Implement in term of 31487 remap_filename(). 31488 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files. 31489 * builtins.c (fold_builtin_FILE): Call remap_macro_filename(). 31490 * dbxout.c: Include file-prefix-map.h. 31491 * varasm.c: Likewise. 31492 * vmsdbgout.c: Likewise. 31493 * xcoffout.c: Likewise. 31494 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer. 31495 * doc/cppopts.texi (-fmacro-prefix-map): Document. 31496 * doc/invoke.texi (-ffile-prefix-map): Document. 31497 (-fdebug-prefix-map): Update description. 31498 314992018-01-18 Martin Liska <mliska@suse.cz> 31500 31501 * config/i386/i386.c (indirect_thunk_name): Document that also 31502 lfence is emitted. 31503 (output_indirect_thunk): Document why both instructions 31504 (pause and lfence) are generated. 31505 315062018-01-18 Richard Biener <rguenther@suse.de> 31507 31508 PR tree-optimization/83887 31509 * graphite-scop-detection.c 31510 (scop_detection::get_nearest_dom_with_single_entry): Remove. 31511 (scop_detection::get_nearest_pdom_with_single_exit): Likewise. 31512 (scop_detection::merge_sese): Re-implement with a flood-fill 31513 algorithm that properly finds a SESE region if it exists. 31514 315152018-01-18 Jakub Jelinek <jakub@redhat.com> 31516 31517 PR c/61240 31518 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For 31519 pointer_diff optimizations use view_convert instead of convert. 31520 315212018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com> 31522 31523 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>): 31524 Generate different code for -mno-speculate-indirect-jumps. 31525 (*call_value_indirect_nonlocal_sysv<mode>): Likewise. 31526 (*call_indirect_aix<mode>): Disable for 31527 -mno-speculate-indirect-jumps. 31528 (*call_indirect_aix<mode>_nospec): New define_insn. 31529 (*call_value_indirect_aix<mode>): Disable for 31530 -mno-speculate-indirect-jumps. 31531 (*call_value_indirect_aix<mode>_nospec): New define_insn. 31532 (*sibcall_nonlocal_sysv<mode>): Generate different code for 31533 -mno-speculate-indirect-jumps. 31534 (*sibcall_value_nonlocal_sysv<mode>): Likewise. 31535 315362018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com> 31537 31538 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a 31539 long double type, set the flags for noting the default long double 31540 type, even if we don't pass or return a long double type. 31541 315422018-01-17 Jan Hubicka <hubicka@ucw.cz> 31543 31544 PR ipa/83051 31545 * ipa-inline.c (flatten_function): Do not overwrite final inlining 31546 failure. 31547 315482018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com> 31549 31550 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding 31551 support for merge[hl]. 31552 (fold_mergehl_helper): New helper function. 31553 (tree-vector-builder.h): New #include for tree_vector_builder usage. 31554 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn. 31555 (altivec_vmrglw_direct): Add xxmrglw insn. 31556 315572018-01-17 Andrew Waterman <andrew@sifive.com> 31558 31559 * config/riscv/riscv.c (riscv_conditional_register_usage): If 31560 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs. 31561 315622018-01-17 David Malcolm <dmalcolm@redhat.com> 31563 31564 PR lto/83121 31565 * ipa-devirt.c (add_type_duplicate): When comparing memory layout, 31566 call the lto_location_cache before reading the 31567 DECL_SOURCE_LOCATION of the types. 31568 315692018-01-17 Wilco Dijkstra <wdijkstr@arm.com> 31570 Richard Sandiford <richard.sandiford@linaro.org> 31571 31572 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint. 31573 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function. 31574 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE 31575 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates. 31576 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate): 31577 Add declaration. 31578 * config/aarch64/constraints.md (aarch64_movti_operand): 31579 Limit immediates. 31580 * config/aarch64/predicates.md (Uti): Add new constraint. 31581 315822018-01-17 Carl Love <cel@us.ibm.com> 31583 31584 * config/rs6000/vsx.md (define_expand xl_len_r, 31585 define_expand stxvl, define_expand *stxvl): Add match_dup argument. 31586 (define_insn): Add, match_dup 1 argument to define_insn stxvll and 31587 lxvll. 31588 (define_expand, define_insn): Move the shift left from the 31589 define_insn to the define_expand for lxvl and stxvl instructions. 31590 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL 31591 and XL_LEN_R definitions to PURE. 31592 315932018-01-17 Uros Bizjak <ubizjak@gmail.com> 31594 31595 * config/i386/i386.c (indirect_thunk_name): Declare regno 31596 as unsigned int. Compare regno with INVALID_REGNUM. 31597 (output_indirect_thunk): Ditto. 31598 (output_indirect_thunk_function): Ditto. 31599 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM 31600 in the call to output_indirect_thunk_function. 31601 316022018-01-17 Richard Sandiford <richard.sandiford@linaro.org> 31603 31604 PR middle-end/83884 31605 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0) 31606 rather than the size of inner_type to determine the stack slot size 31607 when handling VIEW_CONVERT_EXPRs on strict-alignment targets. 31608 316092018-01-16 Sebastian Peryt <sebastian.peryt@intel.com> 31610 31611 PR target/83546 31612 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND 31613 to PTA_SILVERMONT. 31614 316152018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com> 31616 31617 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little 31618 endian Linux systems to optionally enable multilibs for selecting 31619 the long double type if the user configured an explicit type. 31620 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we 31621 have no long double multilibs if not defined. 31622 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not 31623 warn if the user used -mabi={ieee,ibm}longdouble and we built 31624 multilibs for long double. 31625 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the 31626 appropriate multilib option. 31627 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default 31628 multilib options. 31629 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files 31630 for building long double multilibs. 31631 * config/rs6000/t-ldouble-linux64le-ieee: Likewise. 31632 316332018-01-16 John David Anglin <danglin@gcc.gnu.org> 31634 31635 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller 31636 copies. 31637 31638 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to 31639 64 bits. 31640 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to 31641 128 bits. 31642 31643 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode 31644 variables. 31645 31646 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE 31647 return value. 31648 316492018-01-16 Eric Botcazou <ebotcazou@adacore.com> 31650 31651 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an 31652 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice. 31653 316542018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org> 31655 31656 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate 31657 different rtl trees depending on TARGET_64BIT. 31658 (rs6000_gen_lvx): Likewise. 31659 316602018-01-16 Eric Botcazou <ebotcazou@adacore.com> 31661 31662 * config/visium/visium.md (nop): Tweak comment. 31663 (hazard_nop): Likewise. 31664 316652018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com> 31666 31667 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for 31668 -mspeculate-indirect-jumps. 31669 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable 31670 for -mno-speculate-indirect-jumps. 31671 (*call_indirect_elfv2<mode>_nospec): New define_insn. 31672 (*call_value_indirect_elfv2<mode>): Disable for 31673 -mno-speculate-indirect-jumps. 31674 (*call_value_indirect_elfv2<mode>_nospec): New define_insn. 31675 (indirect_jump): Emit different RTL for 31676 -mno-speculate-indirect-jumps. 31677 (*indirect_jump<mode>): Disable for 31678 -mno-speculate-indirect-jumps. 31679 (*indirect_jump<mode>_nospec): New define_insn. 31680 (tablejump): Emit different RTL for 31681 -mno-speculate-indirect-jumps. 31682 (tablejumpsi): Disable for -mno-speculate-indirect-jumps. 31683 (tablejumpsi_nospec): New define_expand. 31684 (tablejumpdi): Disable for -mno-speculate-indirect-jumps. 31685 (tablejumpdi_nospec): New define_expand. 31686 (*tablejump<mode>_internal1): Disable for 31687 -mno-speculate-indirect-jumps. 31688 (*tablejump<mode>_internal1_nospec): New define_insn. 31689 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New 31690 option. 31691 316922018-01-16 Artyom Skrobov tyomitch@gmail.com 31693 31694 * caller-save.c (insert_save): Drop unnecessary parameter. All 31695 callers updated. 31696 316972018-01-16 Jakub Jelinek <jakub@redhat.com> 31698 Richard Biener <rguenth@suse.de> 31699 31700 PR libgomp/83590 31701 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr) 31702 return early, inline manually is_gimple_sizepos. Make sure if we 31703 call gimplify_expr we don't end up with a gimple constant. 31704 * tree.c (variably_modified_type_p): Don't return true for 31705 is_gimple_constant (_t). Inline manually is_gimple_sizepos. 31706 * gimplify.h (is_gimple_sizepos): Remove. 31707 317082018-01-16 Richard Sandiford <richard.sandiford@linaro.org> 31709 31710 PR tree-optimization/83857 31711 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call 31712 vectorizable_live_operation for pure SLP statements. 31713 (vectorizable_live_operation): Handle PHIs. 31714 317152018-01-16 Richard Biener <rguenther@suse.de> 31716 31717 PR tree-optimization/83867 31718 * tree-vect-stmts.c (vect_transform_stmt): Precompute 31719 nested_in_vect_loop_p since the scalar stmt may get invalidated. 31720 317212018-01-16 Jakub Jelinek <jakub@redhat.com> 31722 31723 PR c/83844 31724 * stor-layout.c (handle_warn_if_not_align): Use byte_position and 31725 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check. 31726 If off is not INTEGER_CST, issue a may not be aligned warning 31727 rather than isn't aligned. Use isn%'t rather than isn't. 31728 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through 31729 into MULT_EXPR. 31730 <case MULT_EXPR>: Improve the case when bottom and one of the 31731 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that 31732 operand, in that case check if the other operand is multiple of 31733 bottom divided by the INTEGER_CST operand. 31734 317352018-01-16 Richard Sandiford <richard.sandiford@linaro.org> 31736 31737 PR target/83858 31738 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete. 31739 * config/pa/pa-protos.h (pa_function_arg_size): Declare. 31740 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use 31741 pa_function_arg_size instead of FUNCTION_ARG_SIZE. 31742 * config/pa/pa.c (pa_function_arg_advance): Likewise. 31743 (pa_function_arg, pa_arg_partial_bytes): Likewise. 31744 (pa_function_arg_size): New function. 31745 317462018-01-16 Richard Sandiford <richard.sandiford@linaro.org> 31747 31748 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices 31749 in a separate statement. 31750 317512018-01-16 Richard Sandiford <richard.sandiford@linaro.org> 31752 31753 PR tree-optimization/83847 31754 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't 31755 group gathers and scatters. 31756 317572018-01-16 Jakub Jelinek <jakub@redhat.com> 31758 31759 PR rtl-optimization/86620 31760 * params.def (max-sched-ready-insns): Bump minimum value to 1. 31761 31762 PR rtl-optimization/83213 31763 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn 31764 to last if both are JUMP_INSNs. 31765 31766 PR tree-optimization/83843 31767 * gimple-ssa-store-merging.c 31768 (imm_store_chain_info::output_merged_store): Handle bit_not_p on 31769 store_immediate_info for bswap/nop orig_stores. 31770 317712018-01-15 Andrew Waterman <andrew@sifive.com> 31772 31773 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if 31774 !TARGET_MUL. 31775 <UDIV>: Increase cost if !TARGET_DIV. 31776 317772018-01-15 Segher Boessenkool <segher@kernel.crashing.org> 31778 31779 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr. 31780 (define_attr "cr_logical_3op"): New. 31781 (cceq_ior_compare): Adjust. 31782 (cceq_ior_compare_complement): Adjust. 31783 (*cceq_rev_compare): Adjust. 31784 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust. 31785 (is_cracked_insn): Adjust. 31786 (insn_must_be_first_in_group): Adjust. 31787 * config/rs6000/40x.md: Adjust. 31788 * config/rs6000/440.md: Adjust. 31789 * config/rs6000/476.md: Adjust. 31790 * config/rs6000/601.md: Adjust. 31791 * config/rs6000/603.md: Adjust. 31792 * config/rs6000/6xx.md: Adjust. 31793 * config/rs6000/7450.md: Adjust. 31794 * config/rs6000/7xx.md: Adjust. 31795 * config/rs6000/8540.md: Adjust. 31796 * config/rs6000/cell.md: Adjust. 31797 * config/rs6000/e300c2c3.md: Adjust. 31798 * config/rs6000/e500mc.md: Adjust. 31799 * config/rs6000/e500mc64.md: Adjust. 31800 * config/rs6000/e5500.md: Adjust. 31801 * config/rs6000/e6500.md: Adjust. 31802 * config/rs6000/mpc.md: Adjust. 31803 * config/rs6000/power4.md: Adjust. 31804 * config/rs6000/power5.md: Adjust. 31805 * config/rs6000/power6.md: Adjust. 31806 * config/rs6000/power7.md: Adjust. 31807 * config/rs6000/power8.md: Adjust. 31808 * config/rs6000/power9.md: Adjust. 31809 * config/rs6000/rs64.md: Adjust. 31810 * config/rs6000/titan.md: Adjust. 31811 318122018-01-15 H.J. Lu <hongjiu.lu@intel.com> 31813 31814 * config/i386/predicates.md (indirect_branch_operand): Rewrite 31815 ix86_indirect_branch_register logic. 31816 318172018-01-15 H.J. Lu <hongjiu.lu@intel.com> 31818 31819 * config/i386/constraints.md (Bs): Update 31820 ix86_indirect_branch_register check. Don't check 31821 ix86_indirect_branch_register with GOT_memory_operand. 31822 (Bw): Likewise. 31823 * config/i386/predicates.md (GOT_memory_operand): Don't check 31824 ix86_indirect_branch_register here. 31825 (GOT32_symbol_operand): Likewise. 31826 318272018-01-15 H.J. Lu <hongjiu.lu@intel.com> 31828 31829 * config/i386/predicates.md (constant_call_address_operand): 31830 Rewrite ix86_indirect_branch_register logic. 31831 (sibcall_insn_operand): Likewise. 31832 318332018-01-15 H.J. Lu <hongjiu.lu@intel.com> 31834 31835 * config/i386/constraints.md (Bs): Replace 31836 ix86_indirect_branch_thunk_register with 31837 ix86_indirect_branch_register. 31838 (Bw): Likewise. 31839 * config/i386/i386.md (indirect_jump): Likewise. 31840 (tablejump): Likewise. 31841 (*sibcall_memory): Likewise. 31842 (*sibcall_value_memory): Likewise. 31843 Peepholes of indirect call and jump via memory: Likewise. 31844 * config/i386/i386.opt: Likewise. 31845 * config/i386/predicates.md (indirect_branch_operand): Likewise. 31846 (GOT_memory_operand): Likewise. 31847 (call_insn_operand): Likewise. 31848 (sibcall_insn_operand): Likewise. 31849 (GOT32_symbol_operand): Likewise. 31850 318512018-01-15 Jakub Jelinek <jakub@redhat.com> 31852 31853 PR middle-end/83837 31854 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val 31855 type rather than type addr's type points to. 31856 (expand_omp_atomic_mutex): Likewise. 31857 (expand_omp_atomic): Likewise. 31858 318592018-01-15 H.J. Lu <hongjiu.lu@intel.com> 31860 31861 PR target/83839 31862 * config/i386/i386.c (output_indirect_thunk_function): Use 31863 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO 31864 for __x86_return_thunk. 31865 318662018-01-15 Richard Biener <rguenther@suse.de> 31867 31868 PR middle-end/83850 31869 * expmed.c (extract_bit_field_1): Fix typo. 31870 318712018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 31872 31873 PR target/83687 31874 * config/arm/iterators.md (VF): New mode iterator. 31875 * config/arm/neon.md (neon_vabd<mode>_2): Use the above. 31876 Remove integer-related logic from pattern. 31877 (neon_vabd<mode>_3): Likewise. 31878 318792018-01-15 Jakub Jelinek <jakub@redhat.com> 31880 31881 PR middle-end/82694 31882 * common.opt (fstrict-overflow): No longer an alias. 31883 (fwrapv-pointer): New option. 31884 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define 31885 also for pointer types based on flag_wrapv_pointer. 31886 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set 31887 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if 31888 opts->x_flag_wrapv got set. 31889 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01 31890 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of 31891 POINTER_TYPE_OVERFLOW_UNDEFINED. 31892 * match.pd: Likewise in address comparison pattern. 31893 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow. 31894 318952018-01-15 Richard Biener <rguenther@suse.de> 31896 31897 PR lto/83804 31898 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs 31899 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization. 31900 Reset type names to their identifier if their TYPE_DECL doesn't 31901 have linkage (and thus is used for ODR and devirt). 31902 (save_debug_info_for_decl): Remove. 31903 (save_debug_info_for_type): Likewise. 31904 (add_tree_to_fld_list): Adjust. 31905 * tree-pretty-print.c (dump_generic_node): Make dumping of 31906 type names more robust. 31907 319082018-01-15 Richard Biener <rguenther@suse.de> 31909 31910 * BASE-VER: Bump to 8.0.1. 31911 319122018-01-14 Martin Sebor <msebor@redhat.com> 31913 31914 PR other/83508 31915 * builtins.c (check_access): Avoid warning when the no-warning bit 31916 is set. 31917 319182018-01-14 Cory Fields <cory-nospam-@coryfields.com> 31919 31920 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort. 31921 * ira-color (allocno_hard_regs_compare): Likewise. 31922 319232018-01-14 Nathan Rossi <nathan@nathanrossi.com> 31924 31925 PR target/83013 31926 * config/microblaze/microblaze.c (microblaze_asm_output_ident): 31927 Use .pushsection/.popsection. 31928 319292018-01-14 Martin Sebor <msebor@redhat.com> 31930 31931 PR c++/81327 31932 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting. 31933 319342018-01-14 Jakub Jelinek <jakub@redhat.com> 31935 31936 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h 31937 entry from extra_headers. 31938 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from 31939 extra_headers, make the list bitwise identical to the i?86-*-* one. 31940 319412018-01-14 H.J. Lu <hongjiu.lu@intel.com> 31942 31943 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow 31944 -mcmodel=large with -mindirect-branch=thunk, 31945 -mindirect-branch=thunk-extern, -mfunction-return=thunk and 31946 -mfunction-return=thunk-extern. 31947 * doc/invoke.texi: Document -mcmodel=large is incompatible with 31948 -mindirect-branch=thunk, -mindirect-branch=thunk-extern, 31949 -mfunction-return=thunk and -mfunction-return=thunk-extern. 31950 319512018-01-14 H.J. Lu <hongjiu.lu@intel.com> 31952 31953 * config/i386/i386.c (print_reg): Print the name of the full 31954 integer register without '%'. 31955 (ix86_print_operand): Handle 'V'. 31956 * doc/extend.texi: Document 'V' modifier. 31957 319582018-01-14 H.J. Lu <hongjiu.lu@intel.com> 31959 31960 * config/i386/constraints.md (Bs): Disallow memory operand for 31961 -mindirect-branch-register. 31962 (Bw): Likewise. 31963 * config/i386/predicates.md (indirect_branch_operand): Likewise. 31964 (GOT_memory_operand): Likewise. 31965 (call_insn_operand): Likewise. 31966 (sibcall_insn_operand): Likewise. 31967 (GOT32_symbol_operand): Likewise. 31968 * config/i386/i386.md (indirect_jump): Call convert_memory_address 31969 for -mindirect-branch-register. 31970 (tablejump): Likewise. 31971 (*sibcall_memory): Likewise. 31972 (*sibcall_value_memory): Likewise. 31973 Disallow peepholes of indirect call and jump via memory for 31974 -mindirect-branch-register. 31975 (*call_pop): Replace m with Bw. 31976 (*call_value_pop): Likewise. 31977 (*sibcall_pop_memory): Replace m with Bs. 31978 * config/i386/i386.opt (mindirect-branch-register): New option. 31979 * doc/invoke.texi: Document -mindirect-branch-register option. 31980 319812018-01-14 H.J. Lu <hongjiu.lu@intel.com> 31982 31983 * config/i386/i386-protos.h (ix86_output_function_return): New. 31984 * config/i386/i386.c (ix86_set_indirect_branch_type): Also 31985 set function_return_type. 31986 (indirect_thunk_name): Add ret_p to indicate thunk for function 31987 return. 31988 (output_indirect_thunk_function): Pass false to 31989 indirect_thunk_name. 31990 (ix86_output_indirect_branch_via_reg): Likewise. 31991 (ix86_output_indirect_branch_via_push): Likewise. 31992 (output_indirect_thunk_function): Create alias for function 31993 return thunk if regno < 0. 31994 (ix86_output_function_return): New function. 31995 (ix86_handle_fndecl_attribute): Handle function_return. 31996 (ix86_attribute_table): Add function_return. 31997 * config/i386/i386.h (machine_function): Add 31998 function_return_type. 31999 * config/i386/i386.md (simple_return_internal): Use 32000 ix86_output_function_return. 32001 (simple_return_internal_long): Likewise. 32002 * config/i386/i386.opt (mfunction-return=): New option. 32003 (indirect_branch): Mention -mfunction-return=. 32004 * doc/extend.texi: Document function_return function attribute. 32005 * doc/invoke.texi: Document -mfunction-return= option. 32006 320072018-01-14 H.J. Lu <hongjiu.lu@intel.com> 32008 32009 * config/i386/i386-opts.h (indirect_branch): New. 32010 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise. 32011 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone 32012 with local indirect jump when converting indirect call and jump. 32013 (ix86_set_indirect_branch_type): New. 32014 (ix86_set_current_function): Call ix86_set_indirect_branch_type. 32015 (indirectlabelno): New. 32016 (indirect_thunk_needed): Likewise. 32017 (indirect_thunk_bnd_needed): Likewise. 32018 (indirect_thunks_used): Likewise. 32019 (indirect_thunks_bnd_used): Likewise. 32020 (INDIRECT_LABEL): Likewise. 32021 (indirect_thunk_name): Likewise. 32022 (output_indirect_thunk): Likewise. 32023 (output_indirect_thunk_function): Likewise. 32024 (ix86_output_indirect_branch_via_reg): Likewise. 32025 (ix86_output_indirect_branch_via_push): Likewise. 32026 (ix86_output_indirect_branch): Likewise. 32027 (ix86_output_indirect_jmp): Likewise. 32028 (ix86_code_end): Call output_indirect_thunk_function if needed. 32029 (ix86_output_call_insn): Call ix86_output_indirect_branch if 32030 needed. 32031 (ix86_handle_fndecl_attribute): Handle indirect_branch. 32032 (ix86_attribute_table): Add indirect_branch. 32033 * config/i386/i386.h (machine_function): Add indirect_branch_type 32034 and has_local_indirect_jump. 32035 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump 32036 to true. 32037 (tablejump): Likewise. 32038 (*indirect_jump): Use ix86_output_indirect_jmp. 32039 (*tablejump_1): Likewise. 32040 (simple_return_indirect_internal): Likewise. 32041 * config/i386/i386.opt (mindirect-branch=): New option. 32042 (indirect_branch): New. 32043 (keep): Likewise. 32044 (thunk): Likewise. 32045 (thunk-inline): Likewise. 32046 (thunk-extern): Likewise. 32047 * doc/extend.texi: Document indirect_branch function attribute. 32048 * doc/invoke.texi: Document -mindirect-branch= option. 32049 320502018-01-14 Jan Hubicka <hubicka@ucw.cz> 32051 32052 PR ipa/83051 32053 * ipa-inline.c (edge_badness): Tolerate roundoff errors. 32054 320552018-01-14 Richard Sandiford <richard.sandiford@linaro.org> 32056 32057 * ipa-inline.c (want_inline_small_function_p): Return false if 32058 inlining has already failed with CIF_FINAL_ERROR. 32059 (update_caller_keys): Call want_inline_small_function_p before 32060 can_inline_edge_p. 32061 (update_callee_keys): Likewise. 32062 320632018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org> 32064 32065 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p): 32066 New function. 32067 (rs6000_quadword_masked_address_p): Likewise. 32068 (quad_aligned_load_p): Likewise. 32069 (quad_aligned_store_p): Likewise. 32070 (const_load_sequence_p): Add comment to describe the outer-most loop. 32071 (mimic_memory_attributes_and_flags): New function. 32072 (rs6000_gen_stvx): Likewise. 32073 (replace_swapped_aligned_store): Likewise. 32074 (rs6000_gen_lvx): Likewise. 32075 (replace_swapped_aligned_load): Likewise. 32076 (replace_swapped_load_constant): Capitalize argument name in 32077 comment describing this function. 32078 (rs6000_analyze_swaps): Add a third pass to search for vector loads 32079 and stores that access quad-word aligned addresses and replace 32080 with stvx or lvx instructions when appropriate. 32081 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p): 32082 New function prototype. 32083 (rs6000_quadword_masked_address_p): Likewise. 32084 (rs6000_gen_lvx): Likewise. 32085 (rs6000_gen_stvx): Likewise. 32086 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes 32087 VSX_D (V2DF, V2DI), modify this split to select lvx instruction 32088 when memory address is aligned. 32089 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify 32090 this split to select lvx instruction when memory address is aligned. 32091 (*vsx_le_perm_load_v8hi): Modify this split to select lvx 32092 instruction when memory address is aligned. 32093 (*vsx_le_perm_load_v16qi): Likewise. 32094 (four unnamed splitters): Modify to select the stvx instruction 32095 when memory is aligned. 32096 320972018-01-13 Jan Hubicka <hubicka@ucw.cz> 32098 32099 * predict.c (determine_unlikely_bbs): Handle correctly BBs 32100 which appears in the queue multiple times. 32101 321022018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32103 Alan Hayward <alan.hayward@arm.com> 32104 David Sherwood <david.sherwood@arm.com> 32105 32106 * tree-vectorizer.h (vec_lower_bound): New structure. 32107 (_loop_vec_info): Add check_nonzero and lower_bounds. 32108 (LOOP_VINFO_CHECK_NONZERO): New macro. 32109 (LOOP_VINFO_LOWER_BOUNDS): Likewise. 32110 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too. 32111 * tree-data-ref.h (dr_with_seg_len): Add access_size and align 32112 fields. Make seg_len the distance travelled, not including the 32113 access size. 32114 (dr_direction_indicator): Declare. 32115 (dr_zero_step_indicator): Likewise. 32116 (dr_known_forward_stride_p): Likewise. 32117 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and 32118 tree-ssanames.h. 32119 (runtime_alias_check_p): Allow runtime alias checks with 32120 variable strides. 32121 (operator ==): Compare access_size and align. 32122 (prune_runtime_alias_test_list): Rework for new distinction between 32123 the access_size and seg_len. 32124 (create_intersect_range_checks_index): Likewise. Cope with polynomial 32125 segment lengths. 32126 (get_segment_min_max): New function. 32127 (create_intersect_range_checks): Use it. 32128 (dr_step_indicator): New function. 32129 (dr_direction_indicator): Likewise. 32130 (dr_zero_step_indicator): Likewise. 32131 (dr_known_forward_stride_p): Likewise. 32132 * tree-loop-distribution.c (data_ref_segment_size): Return 32133 DR_STEP * (niters - 1). 32134 (compute_alias_check_pairs): Update call to the dr_with_seg_len 32135 constructor. 32136 * tree-vect-data-refs.c (vect_check_nonzero_value): New function. 32137 (vect_preserves_scalar_order_p): New function, split out from... 32138 (vect_analyze_data_ref_dependence): ...here. Check for zero steps. 32139 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1). 32140 (vect_vfa_access_size): New function. 32141 (vect_vfa_align): Likewise. 32142 (vect_compile_time_alias): Take access_size_a and access_b arguments. 32143 (dump_lower_bound): New function. 32144 (vect_check_lower_bound): Likewise. 32145 (vect_small_gap_p): Likewise. 32146 (vectorizable_with_step_bound_p): Likewise. 32147 (vect_prune_runtime_alias_test_list): Ignore cross-iteration 32148 depencies if the vectorization factor is 1. Convert the checks 32149 for nonzero steps into checks on the bounds of DR_STEP. Try using 32150 a bunds check for variable steps if the minimum required step is 32151 relatively small. Update calls to the dr_with_seg_len 32152 constructor and to vect_compile_time_alias. 32153 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New 32154 function. 32155 (vect_loop_versioning): Call it. 32156 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS 32157 when retrying. 32158 (vect_estimate_min_profitable_iters): Account for any bounds checks. 32159 321602018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32161 Alan Hayward <alan.hayward@arm.com> 32162 David Sherwood <david.sherwood@arm.com> 32163 32164 * doc/sourcebuild.texi (vect_scatter_store): Document. 32165 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New 32166 optabs. 32167 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}): 32168 Document. 32169 * genopinit.c (main): Add supports_vec_scatter_store and 32170 supports_vec_scatter_store_cached to target_optabs. 32171 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and 32172 IFN_MASK_SCATTER_STORE. 32173 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal 32174 functions. 32175 * internal-fn.h (internal_store_fn_p): Declare. 32176 (internal_fn_stored_value_index): Likewise. 32177 * internal-fn.c (scatter_store_direct): New macro. 32178 (expand_scatter_store_optab_fn): New function. 32179 (direct_scatter_store_optab_supported_p): New macro. 32180 (internal_store_fn_p): New function. 32181 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and 32182 IFN_MASK_SCATTER_STORE. 32183 (internal_fn_mask_index): Likewise. 32184 (internal_fn_stored_value_index): New function. 32185 (internal_gather_scatter_fn_supported_p): Adjust operand numbers 32186 for scatter stores. 32187 * optabs-query.h (supports_vec_scatter_store_p): Declare. 32188 * optabs-query.c (supports_vec_scatter_store_p): New function. 32189 * tree-vectorizer.h (vect_get_store_rhs): Declare. 32190 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return 32191 true for scatter stores. 32192 (vect_gather_scatter_fn_p): Handle scatter stores too. 32193 (vect_check_gather_scatter): Consider using scatter stores if 32194 supports_vec_scatter_store_p. 32195 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle 32196 scatter stores too. 32197 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use 32198 internal_fn_stored_value_index. 32199 (check_load_store_masking): Handle scatter stores too. 32200 (vect_get_store_rhs): Make public. 32201 (vectorizable_call): Use internal_store_fn_p. 32202 (vectorizable_store): Handle scatter store internal functions. 32203 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE 32204 when deciding whether the end of the group has been reached. 32205 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec. 32206 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander. 32207 (mask_scatter_store<mode>): New insns. 32208 322092018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32210 Alan Hayward <alan.hayward@arm.com> 32211 David Sherwood <david.sherwood@arm.com> 32212 32213 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare. 32214 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public. 32215 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New 32216 function. 32217 (vect_use_strided_gather_scatters_p): Take a masked_p argument. 32218 Use vect_truncate_gather_scatter_offset if we can't treat the 32219 operation as a normal gather load or scatter store. 32220 (get_group_load_store_type): Take the gather_scatter_info 32221 as argument. Try using a gather load or scatter store for 32222 single-element groups. 32223 (get_load_store_type): Update calls to get_group_load_store_type 32224 and vect_use_strided_gather_scatters_p. 32225 322262018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32227 Alan Hayward <alan.hayward@arm.com> 32228 David Sherwood <david.sherwood@arm.com> 32229 32230 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra 32231 optional tree argument. 32232 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for 32233 null target hooks. 32234 (vect_create_data_ref_ptr): Take the iv_step as an optional argument, 32235 but continue to use the current value as a fallback. 32236 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare 32237 to compare the updates. 32238 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function. 32239 (get_load_store_type): Use it when handling a strided access. 32240 (vect_get_strided_load_store_ops): New function. 32241 (vect_get_data_ptr_increment): Likewise. 32242 (vectorizable_load): Handle strided gather loads. Always pass 32243 a step to vect_create_data_ref_ptr and bump_vector_ptr. 32244 322452018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32246 Alan Hayward <alan.hayward@arm.com> 32247 David Sherwood <david.sherwood@arm.com> 32248 32249 * doc/md.texi (gather_load@var{m}): Document. 32250 (mask_gather_load@var{m}): Likewise. 32251 * genopinit.c (main): Add supports_vec_gather_load and 32252 supports_vec_gather_load_cached to target_optabs. 32253 * optabs-tree.c (init_tree_optimization_optabs): Use 32254 ggc_cleared_alloc to allocate target_optabs. 32255 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs. 32256 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal 32257 functions. 32258 * internal-fn.h (internal_load_fn_p): Declare. 32259 (internal_gather_scatter_fn_p): Likewise. 32260 (internal_fn_mask_index): Likewise. 32261 (internal_gather_scatter_fn_supported_p): Likewise. 32262 * internal-fn.c (gather_load_direct): New macro. 32263 (expand_gather_load_optab_fn): New function. 32264 (direct_gather_load_optab_supported_p): New macro. 32265 (direct_internal_fn_optab): New function. 32266 (internal_load_fn_p): Likewise. 32267 (internal_gather_scatter_fn_p): Likewise. 32268 (internal_fn_mask_index): Likewise. 32269 (internal_gather_scatter_fn_supported_p): Likewise. 32270 * optabs-query.c (supports_at_least_one_mode_p): New function. 32271 (supports_vec_gather_load_p): Likewise. 32272 * optabs-query.h (supports_vec_gather_load_p): Declare. 32273 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type 32274 and memory_type field. 32275 (NUM_PATTERNS): Bump to 15. 32276 * tree-vect-data-refs.c: Include internal-fn.h. 32277 (vect_gather_scatter_fn_p): New function. 32278 (vect_describe_gather_scatter_call): Likewise. 32279 (vect_check_gather_scatter): Try using internal functions for 32280 gather loads. Recognize existing calls to a gather load function. 32281 (vect_analyze_data_refs): Consider using gather loads if 32282 supports_vec_gather_load_p. 32283 * tree-vect-patterns.c (vect_get_load_store_mask): New function. 32284 (vect_get_gather_scatter_offset_type): Likewise. 32285 (vect_convert_mask_for_vectype): Likewise. 32286 (vect_add_conversion_to_patterm): Likewise. 32287 (vect_try_gather_scatter_pattern): Likewise. 32288 (vect_recog_gather_scatter_pattern): New pattern recognizer. 32289 (vect_vect_recog_func_ptrs): Add it. 32290 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use 32291 internal_fn_mask_index and internal_gather_scatter_fn_p. 32292 (check_load_store_masking): Take the gather_scatter_info as an 32293 argument and handle gather loads. 32294 (vect_get_gather_scatter_ops): New function. 32295 (vectorizable_call): Check internal_load_fn_p. 32296 (vectorizable_load): Likewise. Handle gather load internal 32297 functions. 32298 (vectorizable_store): Update call to check_load_store_masking. 32299 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec. 32300 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators. 32301 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w) 32302 (aarch64_gather_scale_operand_d): New predicates. 32303 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander. 32304 (mask_gather_load<mode>): New insns. 32305 323062018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32307 Alan Hayward <alan.hayward@arm.com> 32308 David Sherwood <david.sherwood@arm.com> 32309 32310 * optabs.def (fold_left_plus_optab): New optab. 32311 * doc/md.texi (fold_left_plus_@var{m}): Document. 32312 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function. 32313 * internal-fn.c (fold_left_direct): Define. 32314 (expand_fold_left_optab_fn): Likewise. 32315 (direct_fold_left_optab_supported_p): Likewise. 32316 * fold-const-call.c (fold_const_fold_left): New function. 32317 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS. 32318 * tree-parloops.c (valid_reduction_p): New function. 32319 (gather_scalar_reductions): Use it. 32320 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type. 32321 (vect_finish_replace_stmt): Declare. 32322 * tree-vect-loop.c (fold_left_reduction_fn): New function. 32323 (needs_fold_left_reduction_p): New function, split out from... 32324 (vect_is_simple_reduction): ...here. Accept reductions that 32325 forbid reassociation, but give them type FOLD_LEFT_REDUCTION. 32326 (vect_force_simple_reduction): Also store the reduction type in 32327 the assignment's STMT_VINFO_REDUC_TYPE. 32328 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION. 32329 (merge_with_identity): New function. 32330 (vect_expand_fold_left): Likewise. 32331 (vectorize_fold_left_reduction): Likewise. 32332 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the 32333 scalar phi in place for it. Check for target support and reject 32334 cases that would reassociate the operation. Defer the transform 32335 phase to vectorize_fold_left_reduction. 32336 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec. 32337 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander. 32338 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns. 32339 323402018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32341 32342 * tree-if-conv.c (predicate_mem_writes): Remove redundant 32343 call to ifc_temp_var. 32344 323452018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32346 Alan Hayward <alan.hayward@arm.com> 32347 David Sherwood <david.sherwood@arm.com> 32348 32349 * target.def (legitimize_address_displacement): Take the original 32350 offset as a poly_int. 32351 * targhooks.h (default_legitimize_address_displacement): Update 32352 accordingly. 32353 * targhooks.c (default_legitimize_address_displacement): Likewise. 32354 * doc/tm.texi: Regenerate. 32355 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement 32356 as an argument, moving assert of ad->disp == ad->disp_term to... 32357 (process_address_1): ...here. Update calls to base_plus_disp_to_reg. 32358 Try calling targetm.legitimize_address_displacement before expanding 32359 the address rather than afterwards, and adjust for the new interface. 32360 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement): 32361 Match the new hook interface. Handle SVE addresses. 32362 * config/sh/sh.c (sh_legitimize_address_displacement): Make the 32363 new hook interface. 32364 323652018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32366 32367 * Makefile.in (OBJS): Add early-remat.o. 32368 * target.def (select_early_remat_modes): New hook. 32369 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook. 32370 * doc/tm.texi: Regenerate. 32371 * targhooks.h (default_select_early_remat_modes): Declare. 32372 * targhooks.c (default_select_early_remat_modes): New function. 32373 * timevar.def (TV_EARLY_REMAT): New timevar. 32374 * passes.def (pass_early_remat): New pass. 32375 * tree-pass.h (make_pass_early_remat): Declare. 32376 * early-remat.c: New file. 32377 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New 32378 function. 32379 (TARGET_SELECT_EARLY_REMAT_MODES): Define. 32380 323812018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32382 Alan Hayward <alan.hayward@arm.com> 32383 David Sherwood <david.sherwood@arm.com> 32384 32385 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace 32386 vfm1 with a bound_epilog parameter. 32387 (vect_do_peeling): Update calls accordingly, and move the prologue 32388 call earlier in the function. Treat the base bound_epilog as 0 for 32389 fully-masked loops and retain vf - 1 for other loops. Add 1 to 32390 this base when peeling for gaps. 32391 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps 32392 with fully-masked loops. 32393 (vect_estimate_min_profitable_iters): Handle the single peeled 32394 iteration in that case. 32395 323962018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32397 Alan Hayward <alan.hayward@arm.com> 32398 David Sherwood <david.sherwood@arm.com> 32399 32400 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow 32401 single-element interleaving even if the size is not a power of 2. 32402 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise 32403 accesses for single-element interleaving if the group size is 32404 not a power of 2. 32405 324062018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32407 Alan Hayward <alan.hayward@arm.com> 32408 David Sherwood <david.sherwood@arm.com> 32409 32410 * doc/md.texi (fold_extract_last_@var{m}): Document. 32411 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise. 32412 * optabs.def (fold_extract_last_optab): New optab. 32413 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function. 32414 * internal-fn.c (fold_extract_direct): New macro. 32415 (expand_fold_extract_optab_fn): Likewise. 32416 (direct_fold_extract_optab_supported_p): Likewise. 32417 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type. 32418 * tree-vect-loop.c (vect_model_reduction_cost): Handle 32419 EXTRACT_LAST_REDUCTION. 32420 (get_initial_def_for_reduction): Do not create an initial vector 32421 for EXTRACT_LAST_REDUCTION reductions. 32422 (vectorizable_reduction): Leave the scalar phi in place for 32423 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION 32424 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an 32425 epilogue code for EXTRACT_LAST_REDUCTION and defer the 32426 transform phase to vectorizable_condition. 32427 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function, 32428 split out from... 32429 (vect_finish_stmt_generation): ...here. 32430 (vect_finish_replace_stmt): New function. 32431 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION. 32432 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New 32433 pattern. 32434 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec. 32435 324362018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32437 Alan Hayward <alan.hayward@arm.com> 32438 David Sherwood <david.sherwood@arm.com> 32439 32440 * doc/md.texi (extract_last_@var{m}): Document. 32441 * optabs.def (extract_last_optab): New optab. 32442 * internal-fn.def (EXTRACT_LAST): New internal function. 32443 * internal-fn.c (cond_unary_direct): New macro. 32444 (expand_cond_unary_optab_fn): Likewise. 32445 (direct_cond_unary_optab_supported_p): Likewise. 32446 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked 32447 loops using EXTRACT_LAST. 32448 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to... 32449 (extract_last_<mode>): ...this optab. 32450 (vec_extract<mode><Vel>): Update accordingly. 32451 324522018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32453 Alan Hayward <alan.hayward@arm.com> 32454 David Sherwood <david.sherwood@arm.com> 32455 32456 * target.def (empty_mask_is_expensive): New hook. 32457 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook. 32458 * doc/tm.texi: Regenerate. 32459 * targhooks.h (default_empty_mask_is_expensive): Declare. 32460 * targhooks.c (default_empty_mask_is_expensive): New function. 32461 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores 32462 if the target says that empty masks are expensive. 32463 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive): 32464 New function. 32465 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine. 32466 324672018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32468 Alan Hayward <alan.hayward@arm.com> 32469 David Sherwood <david.sherwood@arm.com> 32470 32471 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field. 32472 (LOOP_VINFO_MASK_SKIP_NITERS): New macro. 32473 (vect_use_loop_mask_for_alignment_p): New function. 32474 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare. 32475 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an 32476 niters_skip argument. Make sure that the first niters_skip elements 32477 of the first iteration are inactive. 32478 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS. 32479 Update call to vect_set_loop_masks_directly. 32480 (get_misalign_in_elems): New function, split out from... 32481 (vect_gen_prolog_loop_niters): ...here. 32482 (vect_update_init_of_dr): Take a code argument that specifies whether 32483 the adjustment should be added or subtracted. 32484 (vect_update_init_of_drs): Likewise. 32485 (vect_prepare_for_masked_peels): New function. 32486 (vect_do_peeling): Skip prologue peeling if we're using a mask 32487 instead. Update call to vect_update_inits_of_drs. 32488 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize 32489 mask_skip_niters. 32490 (vect_analyze_loop_2): Allow fully-masked loops with peeling for 32491 alignment. Do not include the number of peeled iterations in 32492 the minimum threshold in that case. 32493 (vectorizable_induction): Adjust the start value down by 32494 LOOP_VINFO_MASK_SKIP_NITERS iterations. 32495 (vect_transform_loop): Call vect_prepare_for_masked_peels. 32496 Take the number of skipped iterations into account when calculating 32497 the loop bounds. 32498 * tree-vect-stmts.c (vect_gen_while_not): New function. 32499 325002018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32501 Alan Hayward <alan.hayward@arm.com> 32502 David Sherwood <david.sherwood@arm.com> 32503 32504 * doc/sourcebuild.texi (vect_fully_masked): Document. 32505 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and 32506 default value to 0. 32507 * tree-vect-loop.c (vect_analyze_loop_costing): New function, 32508 split out from... 32509 (vect_analyze_loop_2): ...here. Don't check the vectorization 32510 factor against the number of loop iterations if the loop is 32511 fully-masked. 32512 325132018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32514 Alan Hayward <alan.hayward@arm.com> 32515 David Sherwood <david.sherwood@arm.com> 32516 32517 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into... 32518 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types. 32519 (dump_groups): Update accordingly. 32520 (iv_use::mem_type): New member variable. 32521 (address_p): New function. 32522 (record_use): Add a mem_type argument and initialize the new 32523 mem_type field. 32524 (record_group_use): Add a mem_type argument. Use address_p. 32525 Remove obsolete null checks of base_object. Update call to record_use. 32526 (find_interesting_uses_op): Update call to record_group_use. 32527 (find_interesting_uses_cond): Likewise. 32528 (find_interesting_uses_address): Likewise. 32529 (get_mem_type_for_internal_fn): New function. 32530 (find_address_like_use): Likewise. 32531 (find_interesting_uses_stmt): Try find_address_like_use before 32532 calling find_interesting_uses_op. 32533 (addr_offset_valid_p): Use the iv mem_type field as the type 32534 of the addressed memory. 32535 (add_autoinc_candidates): Likewise. 32536 (get_address_cost): Likewise. 32537 (split_small_address_groups_p): Use address_p. 32538 (split_address_groups): Likewise. 32539 (add_iv_candidate_for_use): Likewise. 32540 (autoinc_possible_for_pair): Likewise. 32541 (rewrite_groups): Likewise. 32542 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS. 32543 (determine_group_iv_cost): Update after split of USE_ADDRESS. 32544 (get_alias_ptr_type_for_ptr_address): New function. 32545 (rewrite_use_address): Rewrite address uses in calls that were 32546 identified by find_address_like_use. 32547 325482018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32549 Alan Hayward <alan.hayward@arm.com> 32550 David Sherwood <david.sherwood@arm.com> 32551 32552 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of 32553 TARGET_MEM_REFs. 32554 * gimple-expr.h (is_gimple_addressable: Likewise. 32555 * gimple-expr.c (is_gimple_address): Likewise. 32556 * internal-fn.c (expand_call_mem_ref): New function. 32557 (expand_mask_load_optab_fn): Use it. 32558 (expand_mask_store_optab_fn): Likewise. 32559 325602018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32561 Alan Hayward <alan.hayward@arm.com> 32562 David Sherwood <david.sherwood@arm.com> 32563 32564 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode}) 32565 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode}) 32566 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode}) 32567 (cond_umax@var{mode}): Document. 32568 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab) 32569 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab) 32570 (cond_umin_optab, cond_umax_optab): New optabs. 32571 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND) 32572 (COND_IOR, COND_XOR): New internal functions. 32573 * internal-fn.h (get_conditional_internal_fn): Declare. 32574 * internal-fn.c (cond_binary_direct): New macro. 32575 (expand_cond_binary_optab_fn): Likewise. 32576 (direct_cond_binary_optab_supported_p): Likewise. 32577 (get_conditional_internal_fn): New function. 32578 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops. 32579 Cope with reduction statements that are vectorized as calls rather 32580 than assignments. 32581 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns. 32582 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB) 32583 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN) 32584 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR) 32585 (UNSPEC_COND_EOR): New unspecs. 32586 (optab): Add mappings for them. 32587 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators. 32588 (sve_int_op, sve_fp_op): New int attributes. 32589 325902018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32591 Alan Hayward <alan.hayward@arm.com> 32592 David Sherwood <david.sherwood@arm.com> 32593 32594 * optabs.def (while_ult_optab): New optab. 32595 * doc/md.texi (while_ult@var{m}@var{n}): Document. 32596 * internal-fn.def (WHILE_ULT): New internal function. 32597 * internal-fn.h (direct_internal_fn_supported_p): New override 32598 that takes two types as argument. 32599 * internal-fn.c (while_direct): New macro. 32600 (expand_while_optab_fn): New function. 32601 (convert_optab_supported_p): Likewise. 32602 (direct_while_optab_supported_p): New macro. 32603 * wide-int.h (wi::udiv_ceil): New function. 32604 * tree-vectorizer.h (rgroup_masks): New structure. 32605 (vec_loop_masks): New typedef. 32606 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p 32607 and fully_masked_p. 32608 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P) 32609 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros. 32610 (vect_max_vf): New function. 32611 (slpeel_make_loop_iterate_ntimes): Delete. 32612 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while) 32613 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare. 32614 (vect_record_loop_mask, vect_get_loop_mask): Likewise. 32615 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h, 32616 internal-fn.h, stor-layout.h and optabs-query.h. 32617 (vect_set_loop_mask): New function. 32618 (add_preheader_seq): Likewise. 32619 (add_header_seq): Likewise. 32620 (interleave_supported_p): Likewise. 32621 (vect_maybe_permute_loop_masks): Likewise. 32622 (vect_set_loop_masks_directly): Likewise. 32623 (vect_set_loop_condition_masked): Likewise. 32624 (vect_set_loop_condition_unmasked): New function, split out from 32625 slpeel_make_loop_iterate_ntimes. 32626 (slpeel_make_loop_iterate_ntimes): Rename to.. 32627 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked 32628 for fully-masked loops and vect_set_loop_condition_unmasked otherwise. 32629 (vect_do_peeling): Update call accordingly. 32630 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked 32631 loops. 32632 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize 32633 mask_compare_type, can_fully_mask_p and fully_masked_p. 32634 (release_vec_loop_masks): New function. 32635 (_loop_vec_info): Use it to free the loop masks. 32636 (can_produce_all_loop_masks_p): New function. 32637 (vect_get_max_nscalars_per_iter): Likewise. 32638 (vect_verify_full_masking): Likewise. 32639 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around 32640 retries, and free the mask rgroups before retrying. Check loop-wide 32641 reasons for disallowing fully-masked loops. Make the final decision 32642 about whether use a fully-masked loop or not. 32643 (vect_estimate_min_profitable_iters): Do not assume that peeling 32644 for the number of iterations will be needed for fully-masked loops. 32645 (vectorizable_reduction): Disable fully-masked loops. 32646 (vectorizable_live_operation): Likewise. 32647 (vect_halve_mask_nunits): New function. 32648 (vect_double_mask_nunits): Likewise. 32649 (vect_record_loop_mask): Likewise. 32650 (vect_get_loop_mask): Likewise. 32651 (vect_transform_loop): Handle the case in which the final loop 32652 iteration might handle a partial vector. Call vect_set_loop_condition 32653 instead of slpeel_make_loop_iterate_ntimes. 32654 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h. 32655 (check_load_store_masking): New function. 32656 (prepare_load_store_mask): Likewise. 32657 (vectorizable_store): Handle fully-masked loops. 32658 (vectorizable_load): Likewise. 32659 (supportable_widening_operation): Use vect_halve_mask_nunits for 32660 booleans. 32661 (supportable_narrowing_operation): Likewise vect_double_mask_nunits. 32662 (vect_gen_while): New function. 32663 * config/aarch64/aarch64.md (umax<mode>3): New expander. 32664 (aarch64_uqdec<mode>): New insn. 32665 326662018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32667 Alan Hayward <alan.hayward@arm.com> 32668 David Sherwood <david.sherwood@arm.com> 32669 32670 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab) 32671 (reduc_xor_scal_optab): New optabs. 32672 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m}) 32673 (reduc_xor_scal_@var{m}): Document. 32674 * doc/sourcebuild.texi (vect_logical_reduc): Likewise. 32675 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New 32676 internal functions. 32677 * fold-const-call.c (fold_const_call): Handle them. 32678 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new 32679 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR. 32680 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>): 32681 (*reduc_<bit_reduc>_scal_<mode>): New patterns. 32682 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV) 32683 (UNSPEC_XORV): New unspecs. 32684 (optab): Add entries for them. 32685 (BITWISEV): New int iterator. 32686 (bit_reduc_op): New int attributes. 32687 326882018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32689 Alan Hayward <alan.hayward@arm.com> 32690 David Sherwood <david.sherwood@arm.com> 32691 32692 * doc/md.texi (vec_shl_insert_@var{m}): New optab. 32693 * internal-fn.def (VEC_SHL_INSERT): New internal function. 32694 * optabs.def (vec_shl_insert_optab): New optab. 32695 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare. 32696 (duplicate_and_interleave): Likewise. 32697 * tree-vect-loop.c: Include internal-fn.h. 32698 (neutral_op_for_slp_reduction): New function, split out from 32699 get_initial_defs_for_reduction. 32700 (get_initial_def_for_reduction): Handle option 2 for variable-length 32701 vectors by loading the neutral value into a vector and then shifting 32702 the initial value into element 0. 32703 (get_initial_defs_for_reduction): Replace the code argument with 32704 the neutral value calculated by neutral_op_for_slp_reduction. 32705 Use gimple_build_vector for constant-length vectors. 32706 Use IFN_VEC_SHL_INSERT for variable-length vectors if all 32707 but the first group_size elements have a neutral value. 32708 Use duplicate_and_interleave otherwise. 32709 (vect_create_epilog_for_reduction): Take a neutral_op parameter. 32710 Update call to get_initial_defs_for_reduction. Handle SLP 32711 reductions for variable-length vectors by creating one vector 32712 result for each scalar result, with the elements associated 32713 with other scalar results stubbed out with the neutral value. 32714 (vectorizable_reduction): Call neutral_op_for_slp_reduction. 32715 Require IFN_VEC_SHL_INSERT for double reductions on 32716 variable-length vectors, or SLP reductions that have 32717 a neutral value. Require can_duplicate_and_interleave_p 32718 support for variable-length unchained SLP reductions if there 32719 is no neutral value, such as for MIN/MAX reductions. Also require 32720 the number of vector elements to be a multiple of the number of 32721 SLP statements when doing variable-length unchained SLP reductions. 32722 Update call to vect_create_epilog_for_reduction. 32723 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public 32724 and remove initial values. 32725 (duplicate_and_interleave): Make public. 32726 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec. 32727 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn. 32728 327292018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32730 Alan Hayward <alan.hayward@arm.com> 32731 David Sherwood <david.sherwood@arm.com> 32732 32733 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h 32734 (can_duplicate_and_interleave_p): New function. 32735 (vect_get_and_check_slp_defs): Take the vector of statements 32736 rather than just the current one. Remove excess parentheses. 32737 Restriction rejectinon of vect_constant_def and vect_external_def 32738 for variable-length vectors to boolean types, or types for which 32739 can_duplicate_and_interleave_p is false. 32740 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs. 32741 (duplicate_and_interleave): New function. 32742 (vect_get_constant_vectors): Use gimple_build_vector for 32743 constant-length vectors and suitable variable-length constant 32744 vectors. Use duplicate_and_interleave for other variable-length 32745 vectors. Don't defer the update when inserting new statements. 32746 327472018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32748 Alan Hayward <alan.hayward@arm.com> 32749 David Sherwood <david.sherwood@arm.com> 32750 32751 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure 32752 min_profitable_iters doesn't go negative. 32753 327542018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32755 Alan Hayward <alan.hayward@arm.com> 32756 David Sherwood <david.sherwood@arm.com> 32757 32758 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document. 32759 (vec_mask_store_lanes@var{m}@var{n}): Likewise. 32760 * optabs.def (vec_mask_load_lanes_optab): New optab. 32761 (vec_mask_store_lanes_optab): Likewise. 32762 * internal-fn.def (MASK_LOAD_LANES): New internal function. 32763 (MASK_STORE_LANES): Likewise. 32764 * internal-fn.c (mask_load_lanes_direct): New macro. 32765 (mask_store_lanes_direct): Likewise. 32766 (expand_mask_load_optab_fn): Handle masked operations. 32767 (expand_mask_load_lanes_optab_fn): New macro. 32768 (expand_mask_store_optab_fn): Handle masked operations. 32769 (expand_mask_store_lanes_optab_fn): New macro. 32770 (direct_mask_load_lanes_optab_supported_p): Likewise. 32771 (direct_mask_store_lanes_optab_supported_p): Likewise. 32772 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p 32773 parameter. 32774 (vect_load_lanes_supported): Likewise. 32775 * tree-vect-data-refs.c (strip_conversion): New function. 32776 (can_group_stmts_p): Likewise. 32777 (vect_analyze_data_ref_accesses): Use it instead of checking 32778 for a pair of assignments. 32779 (vect_store_lanes_supported): Take a masked_p parameter. 32780 (vect_load_lanes_supported): Likewise. 32781 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to 32782 vect_store_lanes_supported and vect_load_lanes_supported. 32783 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise. 32784 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p 32785 parameter. Don't allow gaps for masked accesses. 32786 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported 32787 and vect_load_lanes_supported. 32788 (get_load_store_type): Take a masked_p parameter and update 32789 call to get_group_load_store_type. 32790 (vectorizable_store): Update call to get_load_store_type. 32791 Handle IFN_MASK_STORE_LANES. 32792 (vectorizable_load): Update call to get_load_store_type. 32793 Handle IFN_MASK_LOAD_LANES. 32794 327952018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32796 Alan Hayward <alan.hayward@arm.com> 32797 David Sherwood <david.sherwood@arm.com> 32798 32799 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector 32800 modes for SVE. 32801 * config/aarch64/aarch64-protos.h 32802 (aarch64_sve_struct_memory_operand_p): Declare. 32803 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator. 32804 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes. 32805 (VPRED, vpred): Handle SVE structure modes. 32806 * config/aarch64/constraints.md (Utx): New constraint. 32807 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand) 32808 (aarch64_sve_struct_nonimmediate_operand): New predicates. 32809 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs. 32810 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le) 32811 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for 32812 structure modes. Split into pieces after RA. 32813 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>) 32814 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>): 32815 New patterns. 32816 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle 32817 SVE structure modes. 32818 (aarch64_classify_address): Likewise. 32819 (sizetochar): Move earlier in file. 32820 (aarch64_print_operand): Handle SVE register lists. 32821 (aarch64_array_mode): New function. 32822 (aarch64_sve_struct_memory_operand_p): Likewise. 32823 (TARGET_ARRAY_MODE): Redefine. 32824 328252018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32826 Alan Hayward <alan.hayward@arm.com> 32827 David Sherwood <david.sherwood@arm.com> 32828 32829 * target.def (array_mode): New target hook. 32830 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook. 32831 * doc/tm.texi: Regenerate. 32832 * hooks.h (hook_optmode_mode_uhwi_none): Declare. 32833 * hooks.c (hook_optmode_mode_uhwi_none): New function. 32834 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use 32835 targetm.array_mode. 32836 * stor-layout.c (mode_for_array): Likewise. Support polynomial 32837 type sizes. 32838 328392018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32840 Alan Hayward <alan.hayward@arm.com> 32841 David Sherwood <david.sherwood@arm.com> 32842 32843 * fold-const.c (fold_binary_loc): Check the argument types 32844 rather than the result type when testing for a vector operation. 32845 328462018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32847 32848 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document. 32849 * doc/tm.texi: Regenerate. 32850 328512018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 32852 Alan Hayward <alan.hayward@arm.com> 32853 David Sherwood <david.sherwood@arm.com> 32854 32855 * doc/invoke.texi (-msve-vector-bits=): Document new option. 32856 (sve): Document new AArch64 extension. 32857 * doc/md.texi (w): Extend the description of the AArch64 32858 constraint to include SVE vectors. 32859 (Upl, Upa): Document new AArch64 predicate constraints. 32860 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New 32861 enum. 32862 * config/aarch64/aarch64.opt (sve_vector_bits): New enum. 32863 (msve-vector-bits=): New option. 32864 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable 32865 SVE when these are disabled. 32866 (sve): New extension. 32867 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate 32868 modes. Adjust their number of units based on aarch64_sve_vg. 32869 (MAX_BITSIZE_MODE_ANY_MODE): Define. 32870 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New 32871 aarch64_addr_query_type. 32872 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode) 32873 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p) 32874 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries) 32875 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate) 32876 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate) 32877 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare. 32878 (aarch64_simd_imm_zero_p): Delete. 32879 (aarch64_check_zero_based_sve_index_immediate): Declare. 32880 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p) 32881 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p) 32882 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p) 32883 (aarch64_sve_float_mul_immediate_p): Likewise. 32884 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT 32885 rather than an rtx. 32886 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare. 32887 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback. 32888 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare. 32889 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float) 32890 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare. 32891 (aarch64_regmode_natural_size): Likewise. 32892 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro. 32893 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift 32894 left one place. 32895 (AARCH64_ISA_SVE, TARGET_SVE): New macros. 32896 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries 32897 for VG and the SVE predicate registers. 32898 (V_ALIASES): Add a "z"-prefixed alias. 32899 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1. 32900 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros. 32901 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise. 32902 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes. 32903 (REG_CLASS_NAMES): Add entries for them. 32904 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG 32905 and the predicate registers. 32906 (aarch64_sve_vg): Declare. 32907 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED) 32908 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros. 32909 (REGMODE_NATURAL_SIZE): Define. 32910 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle 32911 SVE macros. 32912 * config/aarch64/aarch64.c: Include cfgrtl.h. 32913 (simd_immediate_info): Add a constructor for series vectors, 32914 and an associated step field. 32915 (aarch64_sve_vg): New variable. 32916 (aarch64_dbx_register_number): Handle VG and the predicate registers. 32917 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete. 32918 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE) 32919 (VEC_ANY_DATA, VEC_STRUCT): New constants. 32920 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p) 32921 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p) 32922 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode) 32923 (aarch64_get_mask_mode): New functions. 32924 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS 32925 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS. 32926 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE 32927 predicate modes and predicate registers. Explicitly restrict 32928 GPRs to modes of 16 bytes or smaller. Only allow FP registers 32929 to store a vector mode if it is recognized by 32930 aarch64_classify_vector_mode. 32931 (aarch64_regmode_natural_size): New function. 32932 (aarch64_hard_regno_caller_save_mode): Return the original mode 32933 for predicates. 32934 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate) 32935 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl) 32936 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate) 32937 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New 32938 functions. 32939 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1 32940 does not overlap dest if the function is frame-related. Handle 32941 SVE constants. 32942 (aarch64_split_add_offset): New function. 32943 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass 32944 them aarch64_add_offset. 32945 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter 32946 and update call to aarch64_sub_sp. 32947 (aarch64_add_cfa_expression): New function. 32948 (aarch64_expand_prologue): Pass extra temporary registers to the 32949 functions above. Handle the case in which we need to emit new 32950 DW_CFA_expressions for registers that were originally saved 32951 relative to the stack pointer, but now have to be expressed 32952 relative to the frame pointer. 32953 (aarch64_output_mi_thunk): Pass extra temporary registers to the 32954 functions above. 32955 (aarch64_expand_epilogue): Likewise. Prevent inheritance of 32956 IP0 and IP1 values for SVE frames. 32957 (aarch64_expand_vec_series): New function. 32958 (aarch64_expand_sve_widened_duplicate): Likewise. 32959 (aarch64_expand_sve_const_vector): Likewise. 32960 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter. 32961 Handle SVE constants. Use emit_move_insn to move a force_const_mem 32962 into the register, rather than emitting a SET directly. 32963 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move) 32964 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p) 32965 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p) 32966 (offset_9bit_signed_scaled_p): New functions. 32967 (aarch64_replicate_bitmask_imm): New function. 32968 (aarch64_bitmask_imm): Use it. 32969 (aarch64_cannot_force_const_mem): Reject expressions involving 32970 a CONST_POLY_INT. Update call to aarch64_classify_symbol. 32971 (aarch64_classify_index): Handle SVE indices, by requiring 32972 a plain register index with a scale that matches the element size. 32973 (aarch64_classify_address): Handle SVE addresses. Assert that 32974 the mode of the address is VOIDmode or an integer mode. 32975 Update call to aarch64_classify_symbol. 32976 (aarch64_classify_symbolic_expression): Update call to 32977 aarch64_classify_symbol. 32978 (aarch64_const_vec_all_in_range_p): New function. 32979 (aarch64_print_vector_float_operand): Likewise. 32980 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than 32981 "vN" for FP registers with SVE modes. Handle (const ...) vectors 32982 and the FP immediates 1.0 and 0.5. 32983 (aarch64_print_address_internal): Handle SVE addresses. 32984 (aarch64_print_operand_address): Use ADDR_QUERY_ANY. 32985 (aarch64_regno_regclass): Handle predicate registers. 32986 (aarch64_secondary_reload): Handle big-endian reloads of SVE 32987 data modes. 32988 (aarch64_class_max_nregs): Handle SVE modes and predicate registers. 32989 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions. 32990 (aarch64_convert_sve_vector_bits): New function. 32991 (aarch64_override_options): Use it to handle -msve-vector-bits=. 32992 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT 32993 rather than an rtx. 32994 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode. 32995 Handle SVE vector and predicate modes. Accept VL-based constants 32996 that need only one temporary register, and VL offsets that require 32997 no temporary registers. 32998 (aarch64_conditional_register_usage): Mark the predicate registers 32999 as fixed if SVE isn't available. 33000 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode. 33001 Return true for SVE vector and predicate modes. 33002 (aarch64_simd_container_mode): Take the number of bits as a poly_int64 33003 rather than an unsigned int. Handle SVE modes. 33004 (aarch64_preferred_simd_mode): Update call accordingly. Handle 33005 SVE modes. 33006 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR 33007 if SVE is enabled. 33008 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p) 33009 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p) 33010 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p) 33011 (aarch64_sve_float_mul_immediate_p): New functions. 33012 (aarch64_sve_valid_immediate): New function. 33013 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors. 33014 Explicitly reject structure modes. Check for INDEX constants. 33015 Handle PTRUE and PFALSE constants. 33016 (aarch64_check_zero_based_sve_index_immediate): New function. 33017 (aarch64_simd_imm_zero_p): Delete. 33018 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for 33019 vector modes. Accept constants in the range of CNT[BHWD]. 33020 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly 33021 ask for an Advanced SIMD mode. 33022 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions. 33023 (aarch64_simd_vector_alignment): Handle SVE predicates. 33024 (aarch64_vectorize_preferred_vector_alignment): New function. 33025 (aarch64_simd_vector_alignment_reachable): Use it instead of 33026 the vector size. 33027 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p. 33028 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New 33029 functions. 33030 (MAX_VECT_LEN): Delete. 33031 (expand_vec_perm_d): Add a vec_flags field. 33032 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions. 33033 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip) 33034 (aarch64_evpc_ext): Don't apply a big-endian lane correction 33035 for SVE modes. 33036 (aarch64_evpc_rev): Rename to... 33037 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE. 33038 (aarch64_evpc_rev_global): New function. 33039 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP. 33040 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of 33041 MAX_VECT_LEN. 33042 (aarch64_evpc_sve_tbl): New function. 33043 (aarch64_expand_vec_perm_const_1): Update after rename of 33044 aarch64_evpc_rev. Handle SVE permutes too, trying 33045 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather 33046 than aarch64_evpc_tbl. 33047 (aarch64_vectorize_vec_perm_const): Initialize vec_flags. 33048 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code) 33049 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int) 33050 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or) 33051 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float) 33052 (aarch64_expand_sve_vcond): New functions. 33053 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead 33054 of aarch64_vector_mode_p. 33055 (aarch64_dwarf_poly_indeterminate_value): New function. 33056 (aarch64_compute_pressure_classes): Likewise. 33057 (aarch64_can_change_mode_class): Likewise. 33058 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine. 33059 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise. 33060 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise. 33061 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise. 33062 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise. 33063 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise. 33064 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr) 33065 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New 33066 constraints. 33067 (Dn, Dl, Dr): Accept const as well as const_vector. 33068 (Dz): Likewise. Compare against CONST0_RTX. 33069 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead 33070 of "vector" where appropriate. 33071 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD) 33072 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators. 33073 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT) 33074 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE) 33075 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS) 33076 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs. 33077 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV) 33078 (v_int_equiv): Extend to SVE modes. 33079 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New 33080 mode attributes. 33081 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators. 33082 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt. 33083 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs. 33084 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP) 33085 (SVE_COND_FP_CMP): New int iterators. 33086 (perm_hilo): Handle the new unpack unspecs. 33087 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int 33088 attributes. 33089 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate) 33090 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate) 33091 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand) 33092 (aarch64_equality_operator, aarch64_constant_vector_operand) 33093 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates. 33094 (aarch64_sve_nonimmediate_operand): Likewise. 33095 (aarch64_sve_general_operand): Likewise. 33096 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise. 33097 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate) 33098 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise. 33099 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise. 33100 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise. 33101 (aarch64_sve_float_arith_immediate): Likewise. 33102 (aarch64_sve_float_arith_with_sub_immediate): Likewise. 33103 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise. 33104 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise. 33105 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise. 33106 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise. 33107 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise. 33108 (aarch64_sve_float_arith_operand): Likewise. 33109 (aarch64_sve_float_arith_with_sub_operand): Likewise. 33110 (aarch64_sve_float_mul_operand): Likewise. 33111 (aarch64_sve_vec_perm_operand): Likewise. 33112 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate. 33113 (aarch64_mov_operand): Accept const_poly_int and const_vector. 33114 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const 33115 as well as const_vector. 33116 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier 33117 in file. Use CONST0_RTX and CONSTM1_RTX. 33118 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes. 33119 (aarch64_simd_reg_or_zero): Accept const as well as const_vector. 33120 Use aarch64_simd_imm_zero. 33121 * config/aarch64/aarch64-sve.md: New file. 33122 * config/aarch64/aarch64.md: Include it. 33123 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers. 33124 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE) 33125 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI) 33126 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK) 33127 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants. 33128 (sve): New attribute. 33129 (enabled): Disable instructions with the sve attribute unless 33130 TARGET_SVE. 33131 (movqi, movhi): Pass CONST_POLY_INT operaneds through 33132 aarch64_expand_mov_immediate. 33133 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle 33134 CNT[BHSD] immediates. 33135 (movti): Split CONST_POLY_INT moves into two halves. 33136 (add<mode>3): Accept aarch64_pluslong_or_poly_operand. 33137 Split additions that need a temporary here if the destination 33138 is the stack pointer. 33139 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates. 33140 (*add<mode>3_poly_1): New instruction. 33141 (set_clobber_cc): New expander. 33142 331432018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 33144 33145 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes 33146 parameter and use it instead of GET_MODE_SIZE (innermode). Use 33147 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode). 33148 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of 33149 GET_MODE_NUNITS (innermode). Also add a first_elem parameter. 33150 Change innermode from fixed_mode_size to machine_mode. 33151 (simplify_subreg): Update call accordingly. Handle a constant-sized 33152 subreg of a variable-length CONST_VECTOR. 33153 331542018-01-13 Richard Sandiford <richard.sandiford@linaro.org> 33155 Alan Hayward <alan.hayward@arm.com> 33156 David Sherwood <david.sherwood@arm.com> 33157 33158 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function. 33159 (add_offset_to_base): New function, split out from... 33160 (create_mem_ref): ...here. When handling a scale other than 1, 33161 check first whether the address is valid without the offset. 33162 Add it into the base if so, leaving the index and scale as-is. 33163 331642018-01-12 Jakub Jelinek <jakub@redhat.com> 33165 33166 PR c++/83778 33167 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call 33168 fold_for_warn before checking if arg2 is INTEGER_CST. 33169 331702018-01-12 Segher Boessenkool <segher@kernel.crashing.org> 33171 33172 * config/rs6000/predicates.md (load_multiple_operation): Delete. 33173 (store_multiple_operation): Delete. 33174 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING. 33175 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete. 33176 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything 33177 guarded by TARGET_STRING. 33178 (rs6000_output_load_multiple): Delete. 33179 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete 33180 OPTION_MASK_STRING / TARGET_STRING handling. 33181 (print_operand) <'N', 'O'>: Add comment that these are unused now. 33182 (const rs6000_opt_masks) <"string">: Change mask to 0. 33183 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING. 33184 (MASK_STRING): Delete. 33185 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING 33186 parts. Simplify. 33187 (load_multiple): Delete. 33188 (*ldmsi8): Delete. 33189 (*ldmsi7): Delete. 33190 (*ldmsi6): Delete. 33191 (*ldmsi5): Delete. 33192 (*ldmsi4): Delete. 33193 (*ldmsi3): Delete. 33194 (store_multiple): Delete. 33195 (*stmsi8): Delete. 33196 (*stmsi7): Delete. 33197 (*stmsi6): Delete. 33198 (*stmsi5): Delete. 33199 (*stmsi4): Delete. 33200 (*stmsi3): Delete. 33201 (movmemsi_8reg): Delete. 33202 (corresponding unnamed define_insn): Delete. 33203 (movmemsi_6reg): Delete. 33204 (corresponding unnamed define_insn): Delete. 33205 (movmemsi_4reg): Delete. 33206 (corresponding unnamed define_insn): Delete. 33207 (movmemsi_2reg): Delete. 33208 (corresponding unnamed define_insn): Delete. 33209 (movmemsi_1reg): Delete. 33210 (corresponding unnamed define_insn): Delete. 33211 * config/rs6000/rs6000.opt (mno-string): New. 33212 (mstring): Replace by deprecation warning stub. 33213 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring. 33214 332152018-01-12 Jakub Jelinek <jakub@redhat.com> 33216 33217 * regrename.c (regrename_do_replace): If replacing the same 33218 reg multiple times, try to reuse last created gen_raw_REG. 33219 33220 PR debug/81155 33221 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition 33222 main to workaround a bug in GDB. 33223 332242018-01-12 Tom de Vries <tom@codesourcery.com> 33225 33226 PR target/83737 33227 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap. 33228 332292018-01-12 Vladimir Makarov <vmakarov@redhat.com> 33230 33231 PR rtl-optimization/80481 33232 * ira-color.c (get_cap_member): New function. 33233 (allocnos_conflict_by_live_ranges_p): Use it. 33234 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert. 33235 (setup_slot_coalesced_allocno_live_ranges): Ditto. 33236 332372018-01-12 Uros Bizjak <ubizjak@gmail.com> 33238 33239 PR target/83628 33240 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern. 33241 (*saddl_se_1): Ditto. 33242 (*ssubsi_1): Ditto. 33243 (*ssubl_se_1): Ditto. 33244 332452018-01-12 Richard Sandiford <richard.sandiford@linaro.org> 33246 33247 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest 33248 rather than wi::to_widest for DR_INITs. 33249 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use 33250 wi::to_poly_offset rather than wi::to_offset for DR_INIT. 33251 (vect_analyze_data_ref_accesses): Require both DR_INITs to be 33252 INTEGER_CSTs. 33253 (vect_analyze_group_access_1): Note that here. 33254 332552018-01-12 Richard Sandiford <richard.sandiford@linaro.org> 33256 33257 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle 33258 polynomial type sizes. 33259 332602018-01-12 Richard Sandiford <richard.sandiford@linaro.org> 33261 33262 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a 33263 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size. 33264 (gimple_add_tmp_var): Likewise. 33265 332662018-01-12 Martin Liska <mliska@suse.cz> 33267 33268 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int. 33269 (gimple_alloc_sizes): Likewise. 33270 (dump_gimple_statistics): Use PRIu64 in printf format. 33271 * gimple.h: Change uint64_t to int. 33272 332732018-01-12 Martin Liska <mliska@suse.cz> 33274 33275 * tree-core.h: Use uint64_t instead of int. 33276 * tree.c (tree_node_counts): Likewise. 33277 (tree_node_sizes): Likewise. 33278 (dump_tree_statistics): Use PRIu64 in printf format. 33279 332802018-01-12 Martin Liska <mliska@suse.cz> 33281 33282 * Makefile.in: As qsort_chk is implemented in vec.c, add 33283 vec.o to linkage of gencfn-macros. 33284 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's 33285 passing the info to record_node_allocation_statistics. 33286 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration 33287 and pass the info. 33288 * ggc-common.c (struct ggc_usage): Add operator== and use 33289 it in operator< and compare function. 33290 * mem-stats.h (struct mem_usage): Likewise. 33291 * vec.c (struct vec_usage): Remove operator< and compare 33292 function. Can be simply inherited. 33293 332942018-01-12 Martin Jambor <mjambor@suse.cz> 33295 33296 PR target/81616 33297 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS. 33298 * tree-ssa-math-opts.c: Include domwalk.h. 33299 (convert_mult_to_fma_1): New function. 33300 (fma_transformation_info): New type. 33301 (fma_deferring_state): Likewise. 33302 (cancel_fma_deferring): New function. 33303 (result_of_phi): Likewise. 33304 (last_fma_candidate_feeds_initial_phi): Likewise. 33305 (convert_mult_to_fma): Added deferring logic, split actual 33306 transformation to convert_mult_to_fma_1. 33307 (math_opts_dom_walker): New type. 33308 (math_opts_dom_walker::after_dom_children): New method, body moved 33309 here from pass_optimize_widening_mul::execute, added deferring logic 33310 bits. 33311 (pass_optimize_widening_mul::execute): Moved most of code to 33312 math_opts_dom_walker::after_dom_children. 33313 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New. 33314 * config/i386/i386.c (ix86_option_override_internal): Added 33315 maybe_setting of PARAM_AVOID_FMA_MAX_BITS. 33316 333172018-01-12 Richard Biener <rguenther@suse.de> 33318 33319 PR debug/83157 33320 * dwarf2out.c (gen_variable_die): Do not reset old_die for 33321 inline instance vars. 33322 333232018-01-12 Oleg Endo <olegendo@gcc.gnu.org> 33324 33325 PR target/81819 33326 * config/rx/rx.c (rx_is_restricted_memory_address): 33327 Handle SUBREG case. 33328 333292018-01-12 Richard Biener <rguenther@suse.de> 33330 33331 PR tree-optimization/80846 33332 * target.def (split_reduction): New target hook. 33333 * targhooks.c (default_split_reduction): New function. 33334 * targhooks.h (default_split_reduction): Declare. 33335 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the 33336 target requests first reduce vectors by combining low and high 33337 parts. 33338 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust. 33339 (get_vectype_for_scalar_type_and_size): Export. 33340 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare. 33341 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document. 33342 * doc/tm.texi: Regenerate. 33343 * config/i386/i386.c (ix86_split_reduction): Implement 33344 TARGET_VECTORIZE_SPLIT_REDUCTION. 33345 333462018-01-12 Eric Botcazou <ebotcazou@adacore.com> 33347 33348 PR target/83368 33349 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM 33350 in PIC mode except for TARGET_VXWORKS_RTP. 33351 * config/sparc/sparc.c: Include cfgrtl.h. 33352 (TARGET_INIT_PIC_REG): Define. 33353 (TARGET_USE_PSEUDO_PIC_REG): Likewise. 33354 (sparc_pic_register_p): New predicate. 33355 (sparc_legitimate_address_p): Use it. 33356 (sparc_legitimize_pic_address): Likewise. 33357 (sparc_delegitimize_address): Likewise. 33358 (sparc_mode_dependent_address_p): Likewise. 33359 (gen_load_pcrel_sym): Remove 4th parameter. 33360 (load_got_register): Adjust call to above. Remove obsolete stuff. 33361 (sparc_expand_prologue): Do not call load_got_register here. 33362 (sparc_flat_expand_prologue): Likewise. 33363 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object. 33364 (sparc_use_pseudo_pic_reg): New function. 33365 (sparc_init_pic_reg): Likewise. 33366 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register. 33367 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP. 33368 333692018-01-12 Christophe Lyon <christophe.lyon@linaro.org> 33370 33371 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes): 33372 Add item for branch_cost. 33373 333742018-01-12 Eric Botcazou <ebotcazou@adacore.com> 33375 33376 PR rtl-optimization/83565 33377 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do 33378 not extend the result to a larger mode for rotate operations. 33379 (num_sign_bit_copies1): Likewise. 33380 333812018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE> 33382 33383 PR target/40411 33384 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or 33385 -symbolic. 33386 Use values-Xc.o for -pedantic. 33387 Link with values-xpg4.o for C90, values-xpg6.o otherwise. 33388 333892018-01-12 Martin Liska <mliska@suse.cz> 33390 33391 PR ipa/83054 33392 * ipa-devirt.c (final_warning_record::grow_type_warnings): 33393 New function. 33394 (possible_polymorphic_call_targets): Use it. 33395 (ipa_devirt): Likewise. 33396 333972018-01-12 Martin Liska <mliska@suse.cz> 33398 33399 * profile-count.h (enum profile_quality): Use 0 as invalid 33400 enum value of profile_quality. 33401 334022018-01-12 Chung-Ju Wu <jasonwucj@gmail.com> 33403 33404 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and 33405 -mext-string options. 33406 334072018-01-12 Richard Biener <rguenther@suse.de> 33408 33409 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process 33410 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P. 33411 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers): 33412 Likewise. 33413 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise. 33414 334152018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com> 33416 33417 * configure.ac (--with-long-double-format): Add support for the 33418 configuration option to change the default long double format on 33419 PowerPC systems. 33420 * config.gcc (powerpc*-linux*-*): Likewise. 33421 * configure: Regenerate. 33422 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long 33423 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be 33424 used without modification. 33425 334262018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com> 33427 33428 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define. 33429 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X. 33430 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle 33431 MISC_BUILTIN_SPEC_BARRIER. 33432 (rs6000_init_builtins): Likewise. 33433 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV 33434 enum value. 33435 (speculation_barrier): New define_insn. 33436 * doc/extend.texi: Document __builtin_speculation_barrier. 33437 334382018-01-11 Jakub Jelinek <jakub@redhat.com> 33439 33440 PR target/83203 33441 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var 33442 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0. 33443 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode 33444 iterators. 33445 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for 33446 integral modes instead of "ss" and "sd". 33447 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit 33448 vectors with 32-bit and 64-bit elements. 33449 (vecdupssescalarmodesuffix): New mode attribute. 33450 (vec_dup<mode>): Use it. 33451 334522018-01-11 H.J. Lu <hongjiu.lu@intel.com> 33453 33454 PR target/83330 33455 * config/i386/i386.c (ix86_compute_frame_layout): Align stack 33456 frame if argument is passed on stack. 33457 334582018-01-11 Jakub Jelinek <jakub@redhat.com> 33459 33460 PR target/82682 33461 * ree.c (combine_reaching_defs): Optimize also 33462 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into 33463 reg2=any_extend(exp); reg1=reg2;, formatting fix. 33464 334652018-01-11 Jan Hubicka <hubicka@ucw.cz> 33466 33467 PR middle-end/83189 33468 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update. 33469 334702018-01-11 Jan Hubicka <hubicka@ucw.cz> 33471 33472 PR middle-end/83718 33473 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling 33474 after they are computed. 33475 334762018-01-11 Bin Cheng <bin.cheng@arm.com> 33477 33478 PR tree-optimization/83695 33479 * gimple-loop-linterchange.cc 33480 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to 33481 reset cached scev information after interchange. 33482 (pass_linterchange::execute): Remove call to scev_reset_htab. 33483 334842018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 33485 33486 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32, 33487 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32, 33488 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32, 33489 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32, 33490 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32, 33491 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define. 33492 * config/arm/arm_neon_builtins.def (vfmal_lane_low, 33493 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high, 33494 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low, 33495 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high, 33496 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins. 33497 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes. 33498 (V_lane_reg): Likewise. 33499 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>): 33500 New define_expand. 33501 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise. 33502 (vfmal_lane_low<mode>_intrinsic, 33503 vfmal_lane_low<vfmlsel2><mode>_intrinsic, 33504 vfmal_lane_high<vfmlsel2><mode>_intrinsic, 33505 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic, 33506 vfmsl_lane_low<vfmlsel2><mode>_intrinsic, 33507 vfmsl_lane_high<vfmlsel2><mode>_intrinsic, 33508 vfmsl_lane_high<mode>_intrinsic): New define_insns. 33509 335102018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 33511 33512 * config/arm/arm-cpus.in (fp16fml): New feature. 33513 (ALL_SIMD): Add fp16fml. 33514 (armv8.2-a): Add fp16fml as an option. 33515 (armv8.3-a): Likewise. 33516 (armv8.4-a): Add fp16fml as part of fp16. 33517 * config/arm/arm.h (TARGET_FP16FML): Define. 33518 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML 33519 when appropriate. 33520 * config/arm/arm-modes.def (V2HF): Define. 33521 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32, 33522 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32, 33523 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define. 33524 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high, 33525 vfmsl_low, vfmsl_high): New set of builtins. 33526 * config/arm/iterators.md (PLUSMINUS): New code iterator. 33527 (vfml_op): New code attribute. 33528 (VFMLHALVES): New int iterator. 33529 (VFML, VFMLSEL): New mode attributes. 33530 (V_reg): Define mapping for V2HF. 33531 (V_hi, V_lo): New mode attributes. 33532 (VF_constraint): Likewise. 33533 (vfml_half, vfml_half_selector): New int attributes. 33534 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New 33535 define_expand. 33536 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic, 33537 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic): 33538 New define_insn. 33539 * config/arm/t-arm-elf (v8_fps): Add fp16fml. 33540 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml. 33541 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs. 33542 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a 33543 documentation. 33544 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon): 33545 Document new effective target and option set. 33546 335472018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 33548 33549 * config/arm/arm-cpus.in (armv8_4): New feature. 33550 (ARMv8_4a): New fgroup. 33551 (armv8.4-a): New arch. 33552 * config/arm/arm-tables.opt: Regenerate. 33553 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a. 33554 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a. 33555 * config/arm/t-multilib (v8_4_a_simd_variants): New variable. 33556 Add matching rules for -march=armv8.4-a and extensions. 33557 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a. 33558 335592018-01-11 Oleg Endo <olegendo@gcc.gnu.org> 33560 33561 PR target/81821 33562 * config/rx/rx.md (BW): New mode attribute. 33563 (sync_lock_test_and_setsi): Add mode suffix to insn output. 33564 335652018-01-11 Richard Biener <rguenther@suse.de> 33566 33567 PR tree-optimization/83435 33568 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges. 33569 * graphite-scop-detection.c (scop_detection::get_sese): Likewise. 33570 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear. 33571 335722018-01-11 Richard Sandiford <richard.sandiford@linaro.org> 33573 Alan Hayward <alan.hayward@arm.com> 33574 David Sherwood <david.sherwood@arm.com> 33575 33576 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset 33577 field. 33578 (aarch64_classify_address): Initialize it. Track polynomial offsets. 33579 (aarch64_print_address_internal): Use it to check for a zero offset. 33580 335812018-01-11 Richard Sandiford <richard.sandiford@linaro.org> 33582 Alan Hayward <alan.hayward@arm.com> 33583 David Sherwood <david.sherwood@arm.com> 33584 33585 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2. 33586 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset): 33587 Return a poly_int64 rather than a HOST_WIDE_INT. 33588 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64 33589 rather than a HOST_WIDE_INT. 33590 * config/aarch64/aarch64.h (aarch64_frame): Protect with 33591 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset, 33592 hard_fp_offset, frame_size, initial_adjust, callee_offset and 33593 final_offset from HOST_WIDE_INT to poly_int64. 33594 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use 33595 to_constant when getting the number of units in an Advanced SIMD 33596 mode. 33597 (aarch64_builtin_vectorized_function): Check for a constant number 33598 of units. 33599 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial 33600 GET_MODE_SIZE. 33601 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits 33602 attribute instead of GET_MODE_NUNITS. 33603 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs) 33604 (aarch64_class_max_nregs): Use the constant_lowest_bound of the 33605 GET_MODE_SIZE for fixed-size registers. 33606 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p. 33607 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index) 33608 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address) 33609 (aarch64_legitimize_address_displacement, aarch64_secondary_reload) 33610 (aarch64_print_operand, aarch64_print_address_internal) 33611 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost) 33612 (aarch64_short_vector_p, aapcs_vfp_sub_candidate) 33613 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp): 33614 Handle polynomial GET_MODE_SIZE. 33615 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes 33616 wider than SImode without modification. 33617 (tls_symbolic_operand_type): Use strip_offset instead of split_const. 33618 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward) 33619 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle 33620 passing and returning SVE modes. 33621 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode 33622 rather than GEN_INT. 33623 (aarch64_emit_probe_stack_range): Take the size as a poly_int64 33624 rather than a HOST_WIDE_INT, but call sorry if it isn't constant. 33625 (aarch64_allocate_and_probe_stack_space): Likewise. 33626 (aarch64_layout_frame): Cope with polynomial offsets. 33627 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the 33628 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track 33629 polynomial offsets. 33630 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p) 33631 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a 33632 poly_int64 rather than a HOST_WIDE_INT. 33633 (aarch64_get_separate_components, aarch64_process_components) 33634 (aarch64_expand_prologue, aarch64_expand_epilogue) 33635 (aarch64_use_return_insn_p): Handle polynomial frame offsets. 33636 (aarch64_anchor_offset): New function, split out from... 33637 (aarch64_legitimize_address): ...here. 33638 (aarch64_builtin_vectorization_cost): Handle polynomial 33639 TYPE_VECTOR_SUBPARTS. 33640 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial 33641 GET_MODE_NUNITS. 33642 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the 33643 number of elements from the PARALLEL rather than the mode. 33644 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE 33645 rather than GET_MODE_BITSIZE. 33646 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext) 33647 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip) 33648 (aarch64_expand_vec_perm_const_1): Handle polynomial 33649 d->perm.length () and d->perm elements. 33650 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS. 33651 Apply to_constant to d->perm elements. 33652 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle 33653 polynomial CONST_VECTOR_NUNITS. 33654 (aarch64_move_pointer): Take amount as a poly_int64 rather 33655 than an int. 33656 (aarch64_progress_pointer): Avoid temporary variable. 33657 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use 33658 the mode attribute instead of GET_MODE. 33659 336602018-01-11 Richard Sandiford <richard.sandiford@linaro.org> 33661 Alan Hayward <alan.hayward@arm.com> 33662 David Sherwood <david.sherwood@arm.com> 33663 33664 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that 33665 x exists before using it. 33666 (aarch64_add_constant_internal): Rename to... 33667 (aarch64_add_offset_1): ...this. Replace regnum with separate 33668 src and dest rtxes. Handle the case in which they're different, 33669 including when the offset is zero. Replace scratchreg with an rtx. 33670 Use 2 additions if there is no spare register into which we can 33671 move a 16-bit constant. 33672 (aarch64_add_constant): Delete. 33673 (aarch64_add_offset): Replace reg with separate src and dest 33674 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT. 33675 Use aarch64_add_offset_1. 33676 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as 33677 an rtx rather than an int. Take the delta as a poly_int64 33678 rather than a HOST_WIDE_INT. Use aarch64_add_offset. 33679 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset. 33680 (aarch64_expand_prologue): Update calls to aarch64_sub_sp, 33681 aarch64_allocate_and_probe_stack_space and aarch64_add_offset. 33682 (aarch64_expand_epilogue): Update calls to aarch64_add_offset 33683 and aarch64_add_sp. 33684 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than 33685 aarch64_add_constant. 33686 336872018-01-11 Richard Sandiford <richard.sandiford@linaro.org> 33688 33689 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int): 33690 Use scalar_float_mode. 33691 336922018-01-11 Richard Sandiford <richard.sandiford@linaro.org> 33693 33694 * config/aarch64/aarch64-simd.md 33695 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS. 33696 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise. 33697 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise. 33698 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise. 33699 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise. 33700 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise. 33701 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise. 33702 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise. 33703 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise. 33704 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise. 33705 337062018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org> 33707 33708 PR target/83514 33709 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if 33710 targ_options->x_arm_arch_string is non NULL. 33711 337122018-01-11 Tamar Christina <tamar.christina@arm.com> 33713 33714 * config/aarch64/aarch64.h 33715 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD. 33716 337172018-01-11 Sudakshina Das <sudi.das@arm.com> 33718 33719 PR target/82096 33720 * expmed.c (emit_store_flag_force): Swap if const op0 33721 and change VOIDmode to mode of op0. 33722 337232018-01-11 Richard Sandiford <richard.sandiford@linaro.org> 33724 33725 PR rtl-optimization/83761 33726 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather 33727 than bytes to mode_for_size. 33728 337292018-01-10 Jan Hubicka <hubicka@ucw.cz> 33730 33731 PR middle-end/83189 33732 * gfortran.fortran-torture/compile/pr83189.f90: New testcase. 33733 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero 33734 profile. 33735 337362018-01-10 Jan Hubicka <hubicka@ucw.cz> 33737 33738 PR middle-end/83575 33739 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition 33740 when in layout mode. 33741 (cfg_layout_finalize): Do not verify cfg before we are out of layout. 33742 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing 33743 partition fixup. 33744 337452018-01-10 Michael Collison <michael.collison@arm.com> 33746 33747 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE. 33748 * config/aarch64/aarch64-option-extension.def: Add 33749 AARCH64_OPT_EXTENSION of 'fp16fml'. 33750 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): 33751 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true. 33752 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate. 33753 * config/aarch64/constraints.md (Ui7): New constraint. 33754 * config/aarch64/iterators.md (VFMLA_W): New mode iterator. 33755 (VFMLA_SEL_W): Ditto. 33756 (f16quad): Ditto. 33757 (f16mac1): Ditto. 33758 (VFMLA16_LOW): New int iterator. 33759 (VFMLA16_HIGH): Ditto. 33760 (UNSPEC_FMLAL): New unspec. 33761 (UNSPEC_FMLSL): Ditto. 33762 (UNSPEC_FMLAL2): Ditto. 33763 (UNSPEC_FMLSL2): Ditto. 33764 (f16mac): New code attribute. 33765 * config/aarch64/aarch64-simd-builtins.def 33766 (aarch64_fmlal_lowv2sf): Ditto. 33767 (aarch64_fmlsl_lowv2sf): Ditto. 33768 (aarch64_fmlalq_lowv4sf): Ditto. 33769 (aarch64_fmlslq_lowv4sf): Ditto. 33770 (aarch64_fmlal_highv2sf): Ditto. 33771 (aarch64_fmlsl_highv2sf): Ditto. 33772 (aarch64_fmlalq_highv4sf): Ditto. 33773 (aarch64_fmlslq_highv4sf): Ditto. 33774 (aarch64_fmlal_lane_lowv2sf): Ditto. 33775 (aarch64_fmlsl_lane_lowv2sf): Ditto. 33776 (aarch64_fmlal_laneq_lowv2sf): Ditto. 33777 (aarch64_fmlsl_laneq_lowv2sf): Ditto. 33778 (aarch64_fmlalq_lane_lowv4sf): Ditto. 33779 (aarch64_fmlsl_lane_lowv4sf): Ditto. 33780 (aarch64_fmlalq_laneq_lowv4sf): Ditto. 33781 (aarch64_fmlsl_laneq_lowv4sf): Ditto. 33782 (aarch64_fmlal_lane_highv2sf): Ditto. 33783 (aarch64_fmlsl_lane_highv2sf): Ditto. 33784 (aarch64_fmlal_laneq_highv2sf): Ditto. 33785 (aarch64_fmlsl_laneq_highv2sf): Ditto. 33786 (aarch64_fmlalq_lane_highv4sf): Ditto. 33787 (aarch64_fmlsl_lane_highv4sf): Ditto. 33788 (aarch64_fmlalq_laneq_highv4sf): Ditto. 33789 (aarch64_fmlsl_laneq_highv4sf): Ditto. 33790 * config/aarch64/aarch64-simd.md: 33791 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern. 33792 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto. 33793 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto. 33794 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto. 33795 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto. 33796 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto. 33797 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto. 33798 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto. 33799 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto. 33800 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto. 33801 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto. 33802 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto. 33803 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto. 33804 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto. 33805 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto. 33806 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto. 33807 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto. 33808 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto. 33809 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto. 33810 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto. 33811 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic. 33812 (vfmlsl_low_u32): Ditto. 33813 (vfmlalq_low_u32): Ditto. 33814 (vfmlslq_low_u32): Ditto. 33815 (vfmlal_high_u32): Ditto. 33816 (vfmlsl_high_u32): Ditto. 33817 (vfmlalq_high_u32): Ditto. 33818 (vfmlslq_high_u32): Ditto. 33819 (vfmlal_lane_low_u32): Ditto. 33820 (vfmlsl_lane_low_u32): Ditto. 33821 (vfmlal_laneq_low_u32): Ditto. 33822 (vfmlsl_laneq_low_u32): Ditto. 33823 (vfmlalq_lane_low_u32): Ditto. 33824 (vfmlslq_lane_low_u32): Ditto. 33825 (vfmlalq_laneq_low_u32): Ditto. 33826 (vfmlslq_laneq_low_u32): Ditto. 33827 (vfmlal_lane_high_u32): Ditto. 33828 (vfmlsl_lane_high_u32): Ditto. 33829 (vfmlal_laneq_high_u32): Ditto. 33830 (vfmlsl_laneq_high_u32): Ditto. 33831 (vfmlalq_lane_high_u32): Ditto. 33832 (vfmlslq_lane_high_u32): Ditto. 33833 (vfmlalq_laneq_high_u32): Ditto. 33834 (vfmlslq_laneq_high_u32): Ditto. 33835 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag. 33836 (AARCH64_FL_FOR_ARCH8_4): New. 33837 (AARCH64_ISA_F16FML): New ISA flag. 33838 (TARGET_F16FML): New feature flag for fp16fml. 33839 (doc/invoke.texi): Document new fp16fml option. 33840 338412018-01-10 Michael Collison <michael.collison@arm.com> 33842 33843 * config/aarch64/aarch64-builtins.c: 33844 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New. 33845 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): 33846 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true. 33847 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags. 33848 (AARCH64_ISA_SHA3): New ISA flag. 33849 (TARGET_SHA3): New feature flag for sha3. 33850 * config/aarch64/iterators.md (sha512_op): New int attribute. 33851 (CRYPTO_SHA512): New int iterator. 33852 (UNSPEC_SHA512H): New unspec. 33853 (UNSPEC_SHA512H2): Ditto. 33854 (UNSPEC_SHA512SU0): Ditto. 33855 (UNSPEC_SHA512SU1): Ditto. 33856 * config/aarch64/aarch64-simd-builtins.def 33857 (aarch64_crypto_sha512hqv2di): New builtin. 33858 (aarch64_crypto_sha512h2qv2di): Ditto. 33859 (aarch64_crypto_sha512su0qv2di): Ditto. 33860 (aarch64_crypto_sha512su1qv2di): Ditto. 33861 (aarch64_eor3qv8hi): Ditto. 33862 (aarch64_rax1qv2di): Ditto. 33863 (aarch64_xarqv2di): Ditto. 33864 (aarch64_bcaxqv8hi): Ditto. 33865 * config/aarch64/aarch64-simd.md: 33866 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern. 33867 (aarch64_crypto_sha512su0qv2di): Ditto. 33868 (aarch64_crypto_sha512su1qv2di): Ditto. 33869 (aarch64_eor3qv8hi): Ditto. 33870 (aarch64_rax1qv2di): Ditto. 33871 (aarch64_xarqv2di): Ditto. 33872 (aarch64_bcaxqv8hi): Ditto. 33873 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic. 33874 (vsha512h2q_u64): Ditto. 33875 (vsha512su0q_u64): Ditto. 33876 (vsha512su1q_u64): Ditto. 33877 (veor3q_u16): Ditto. 33878 (vrax1q_u64): Ditto. 33879 (vxarq_u64): Ditto. 33880 (vbcaxq_u16): Ditto. 33881 * config/arm/types.md (crypto_sha512): New type attribute. 33882 (crypto_sha3): Ditto. 33883 (doc/invoke.texi): Document new sha3 option. 33884 338852018-01-10 Michael Collison <michael.collison@arm.com> 33886 33887 * config/aarch64/aarch64-builtins.c: 33888 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New. 33889 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): 33890 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true. 33891 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true. 33892 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags. 33893 (AARCH64_ISA_SM4): New ISA flag. 33894 (TARGET_SM4): New feature flag for sm4. 33895 * config/aarch64/aarch64-simd-builtins.def 33896 (aarch64_sm3ss1qv4si): Ditto. 33897 (aarch64_sm3tt1aq4si): Ditto. 33898 (aarch64_sm3tt1bq4si): Ditto. 33899 (aarch64_sm3tt2aq4si): Ditto. 33900 (aarch64_sm3tt2bq4si): Ditto. 33901 (aarch64_sm3partw1qv4si): Ditto. 33902 (aarch64_sm3partw2qv4si): Ditto. 33903 (aarch64_sm4eqv4si): Ditto. 33904 (aarch64_sm4ekeyqv4si): Ditto. 33905 * config/aarch64/aarch64-simd.md: 33906 (aarch64_sm3ss1qv4si): Ditto. 33907 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto. 33908 (aarch64_sm3partw<sm3part_op>qv4si): Ditto. 33909 (aarch64_sm4eqv4si): Ditto. 33910 (aarch64_sm4ekeyqv4si): Ditto. 33911 * config/aarch64/iterators.md (sm3tt_op): New int iterator. 33912 (sm3part_op): Ditto. 33913 (CRYPTO_SM3TT): Ditto. 33914 (CRYPTO_SM3PART): Ditto. 33915 (UNSPEC_SM3SS1): New unspec. 33916 (UNSPEC_SM3TT1A): Ditto. 33917 (UNSPEC_SM3TT1B): Ditto. 33918 (UNSPEC_SM3TT2A): Ditto. 33919 (UNSPEC_SM3TT2B): Ditto. 33920 (UNSPEC_SM3PARTW1): Ditto. 33921 (UNSPEC_SM3PARTW2): Ditto. 33922 (UNSPEC_SM4E): Ditto. 33923 (UNSPEC_SM4EKEY): Ditto. 33924 * config/aarch64/constraints.md (Ui2): New constraint. 33925 * config/aarch64/predicates.md (aarch64_imm2): New predicate. 33926 * config/arm/types.md (crypto_sm3): New type attribute. 33927 (crypto_sm4): Ditto. 33928 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic. 33929 (vsm3tt1aq_u32): Ditto. 33930 (vsm3tt1bq_u32): Ditto. 33931 (vsm3tt2aq_u32): Ditto. 33932 (vsm3tt2bq_u32): Ditto. 33933 (vsm3partw1q_u32): Ditto. 33934 (vsm3partw2q_u32): Ditto. 33935 (vsm4eq_u32): Ditto. 33936 (vsm4ekeyq_u32): Ditto. 33937 (doc/invoke.texi): Document new sm4 option. 33938 339392018-01-10 Michael Collison <michael.collison@arm.com> 33940 33941 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture. 33942 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag. 33943 (AARCH64_FL_FOR_ARCH8_4): New. 33944 (AARCH64_FL_V8_4): New flag. 33945 (doc/invoke.texi): Document new armv8.4-a option. 33946 339472018-01-10 Michael Collison <michael.collison@arm.com> 33948 33949 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): 33950 (__ARM_FEATURE_AES): Define if TARGET_AES is true. 33951 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true. 33952 * config/aarch64/aarch64-option-extension.def: Add 33953 AARCH64_OPT_EXTENSION of 'sha2'. 33954 (aes): Add AARCH64_OPT_EXTENSION of 'aes'. 33955 (crypto): Disable sha2 and aes if crypto disabled. 33956 (crypto): Enable aes and sha2 if enabled. 33957 (simd): Disable sha2 and aes if simd disabled. 33958 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2): 33959 New flags. 33960 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags. 33961 (TARGET_SHA2): New feature flag for sha2. 33962 (TARGET_AES): New feature flag for aes. 33963 * config/aarch64/aarch64-simd.md: 33964 (aarch64_crypto_aes<aes_op>v16qi): Make pattern 33965 conditional on TARGET_AES. 33966 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto. 33967 (aarch64_crypto_sha1hsi): Make pattern conditional 33968 on TARGET_SHA2. 33969 (aarch64_crypto_sha1hv4si): Ditto. 33970 (aarch64_be_crypto_sha1hv4si): Ditto. 33971 (aarch64_crypto_sha1su1v4si): Ditto. 33972 (aarch64_crypto_sha1<sha1_op>v4si): Ditto. 33973 (aarch64_crypto_sha1su0v4si): Ditto. 33974 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto. 33975 (aarch64_crypto_sha256su0v4si): Ditto. 33976 (aarch64_crypto_sha256su1v4si): Ditto. 33977 (doc/invoke.texi): Document new aes and sha2 options. 33978 339792018-01-10 Martin Sebor <msebor@redhat.com> 33980 33981 PR tree-optimization/83781 33982 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers 33983 as string arrays. 33984 339852018-01-11 Martin Sebor <msebor@gmail.com> 33986 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org> 33987 33988 PR tree-optimization/83501 33989 PR tree-optimization/81703 33990 33991 * tree-ssa-strlen.c (get_string_cst): Rename... 33992 (get_string_len): ...to this. Handle global constants. 33993 (handle_char_store): Adjust. 33994 339952018-01-10 Kito Cheng <kito.cheng@gmail.com> 33996 Jim Wilson <jimw@sifive.com> 33997 33998 * config/riscv/riscv-protos.h (riscv_output_return): New. 33999 * config/riscv/riscv.c (struct machine_function): New naked_p field. 34000 (riscv_attribute_table, riscv_output_return), 34001 (riscv_handle_fndecl_attribute, riscv_naked_function_p), 34002 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New. 34003 (riscv_compute_frame_info): Only compute frame->mask if not a naked 34004 function. 34005 (riscv_expand_prologue): Add early return for naked function. 34006 (riscv_expand_epilogue): Likewise. 34007 (riscv_function_ok_for_sibcall): Return false for naked function. 34008 (riscv_set_current_function): New. 34009 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS), 34010 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New. 34011 * config/riscv/riscv.md (simple_return): Call riscv_output_return. 34012 * doc/extend.texi (RISC-V Function Attributes): New. 34013 340142018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com> 34015 34016 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly 34017 check for 128-bit long double before checking TCmode. 34018 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for 34019 128-bit long doubles before checking TFmode or TCmode. 34020 (FLOAT128_IBM_P): Likewise. 34021 340222018-01-10 Martin Sebor <msebor@redhat.com> 34023 34024 PR tree-optimization/83671 34025 * builtins.c (c_strlen): Unconditionally return zero for the empty 34026 string. 34027 Use -Warray-bounds for warnings. 34028 * gimple-fold.c (get_range_strlen): Handle non-constant lengths 34029 for non-constant array indices with COMPONENT_REF, arrays of 34030 arrays, and pointers to arrays. 34031 (gimple_fold_builtin_strlen): Determine and set length range for 34032 non-constant character arrays. 34033 340342018-01-10 Aldy Hernandez <aldyh@redhat.com> 34035 34036 PR middle-end/81897 34037 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip 34038 empty blocks. 34039 340402018-01-10 Eric Botcazou <ebotcazou@adacore.com> 34041 34042 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf. 34043 340442018-01-10 Peter Bergner <bergner@vnet.ibm.com> 34045 34046 PR target/83399 34047 * config/rs6000/rs6000.c (print_operand) <'y'>: Use 34048 VECTOR_MEM_ALTIVEC_OR_VSX_P. 34049 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use 34050 indexed_or_indirect_operand predicate. 34051 (*vsx_le_perm_load_<mode> for VSX_W): Likewise. 34052 (*vsx_le_perm_load_v8hi): Likewise. 34053 (*vsx_le_perm_load_v16qi): Likewise. 34054 (*vsx_le_perm_store_<mode> for VSX_D): Likewise. 34055 (*vsx_le_perm_store_<mode> for VSX_W): Likewise. 34056 (*vsx_le_perm_store_v8hi): Likewise. 34057 (*vsx_le_perm_store_v16qi): Likewise. 34058 (eight unnamed splitters): Likewise. 34059 340602018-01-10 Peter Bergner <bergner@vnet.ibm.com> 34061 34062 * config/rs6000/x86intrin.h: Change #warning to #error. Update message. 34063 * config/rs6000/emmintrin.h: Likewise. 34064 * config/rs6000/mmintrin.h: Likewise. 34065 * config/rs6000/xmmintrin.h: Likewise. 34066 340672018-01-10 David Malcolm <dmalcolm@redhat.com> 34068 34069 PR c++/43486 34070 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of 34071 "public_flag". 34072 * tree.c (tree_nop_conversion): Return true for location wrapper 34073 nodes. 34074 (maybe_wrap_with_location): New function. 34075 (selftest::check_strip_nops): New function. 34076 (selftest::test_location_wrappers): New function. 34077 (selftest::tree_c_tests): Call it. 34078 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro. 34079 (maybe_wrap_with_location): New decl. 34080 (EXPR_LOCATION_WRAPPER_P): New macro. 34081 (location_wrapper_p): New inline function. 34082 (tree_strip_any_location_wrapper): New inline function. 34083 340842018-01-10 H.J. Lu <hongjiu.lu@intel.com> 34085 34086 PR target/83735 34087 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust 34088 stack_realign_offset for the largest alignment of stack slot 34089 actually used. 34090 (ix86_find_max_used_stack_alignment): New function. 34091 (ix86_finalize_stack_frame_flags): Use it. Set 34092 max_used_stack_alignment if we don't realign stack. 34093 * config/i386/i386.h (machine_function): Add 34094 max_used_stack_alignment. 34095 340962018-01-10 Christophe Lyon <christophe.lyon@linaro.org> 34097 34098 * config/arm/arm.opt (-mbranch-cost): New option. 34099 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into 34100 account. 34101 341022018-01-10 Segher Boessenkool <segher@kernel.crashing.org> 34103 34104 PR target/83629 34105 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b, 34106 load_toc_v4_PIC_3c): Wrap const term in CONST RTL. 34107 341082018-01-10 Richard Biener <rguenther@suse.de> 34109 34110 PR debug/83765 34111 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration 34112 early out so it also covers the case where we have a non-NULL 34113 origin. 34114 341152018-01-10 Richard Sandiford <richard.sandiford@linaro.org> 34116 34117 PR tree-optimization/83753 34118 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS 34119 for non-strided grouped accesses if the number of elements is 1. 34120 341212018-01-10 Jan Hubicka <hubicka@ucw.cz> 34122 34123 PR target/81616 34124 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER. 34125 * i386.h (TARGET_USE_GATHER): Define. 34126 * x86-tune.def (X86_TUNE_USE_GATHER): New. 34127 341282018-01-10 Martin Liska <mliska@suse.cz> 34129 34130 PR bootstrap/82831 34131 * basic-block.h (CLEANUP_NO_PARTITIONING): New define. 34132 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up 34133 partitioning. 34134 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if 34135 CLEANUP_NO_PARTITIONING is not set. 34136 341372018-01-10 Richard Sandiford <richard.sandiford@linaro.org> 34138 34139 * doc/rtl.texi: Remove documentation of (const ...) wrappers 34140 for vectors, as a partial revert of r254296. 34141 * rtl.h (const_vec_p): Delete. 34142 (const_vec_duplicate_p): Don't test for vector CONSTs. 34143 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise. 34144 * expmed.c (make_tree): Likewise. 34145 34146 Revert: 34147 * common.md (E, F): Use CONSTANT_P instead of checking for 34148 CONST_VECTOR. 34149 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of 34150 checking for CONST_VECTOR. 34151 341522018-01-09 Jan Hubicka <hubicka@ucw.cz> 34153 34154 PR middle-end/83575 34155 * predict.c (force_edge_cold): Handle in more sane way edges 34156 with no prediction. 34157 341582018-01-09 Carl Love <cel@us.ibm.com> 34159 34160 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF, 34161 V4SI, V4SF types. 34162 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types. 34163 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF, 34164 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF, 34165 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW. 34166 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2, 34167 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions. 34168 * config/rs6000/rs6000-protos.h: Add extern defition for 34169 rs6000_generate_float2_double_code. 34170 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add 34171 function. 34172 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn. 34173 (float2_v2df): Add define_expand. 34174 341752018-01-09 Uros Bizjak <ubizjak@gmail.com> 34176 34177 PR target/83628 34178 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of 34179 op_mode in the force_to_mode call. 34180 341812018-01-09 Richard Sandiford <richard.sandiford@linaro.org> 34182 34183 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p 34184 instead of checking each element individually. 34185 (aarch64_evpc_uzp): Likewise. 34186 (aarch64_evpc_zip): Likewise. 34187 (aarch64_evpc_ext): Likewise. 34188 (aarch64_evpc_rev): Likewise. 34189 (aarch64_evpc_dup): Test the encoding for a single duplicated element, 34190 instead of checking each element individually. Return true without 34191 generating rtl if 34192 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test 34193 whether all selected elements come from the same input, instead of 34194 checking each element individually. Remove calls to gen_rtx_REG, 34195 start_sequence and end_sequence and instead assert that no rtl is 34196 generated. 34197 341982018-01-09 Richard Sandiford <richard.sandiford@linaro.org> 34199 34200 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix 34201 order of HIGH and CONST checks. 34202 342032018-01-09 Richard Sandiford <richard.sandiford@linaro.org> 34204 34205 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable 34206 if the destination isn't an SSA_NAME. 34207 342082018-01-09 Richard Biener <rguenther@suse.de> 34209 34210 PR tree-optimization/83668 34211 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument, 34212 move prologue... 34213 (canonicalize_loop_form): ... here, renamed from ... 34214 (canonicalize_loop_closed_ssa_form): ... this and amended to 34215 swap successor edges for loop exit blocks to make us use 34216 the RPO order we need for initial schedule generation. 34217 342182018-01-09 Joseph Myers <joseph@codesourcery.com> 34219 34220 PR tree-optimization/64811 34221 * match.pd: When optimizing comparisons with Inf, avoid 34222 introducing or losing exceptions from comparisons with NaN. 34223 342242018-01-09 Martin Liska <mliska@suse.cz> 34225 34226 PR sanitizer/82517 34227 * asan.c (shadow_mem_size): Add gcc_assert. 34228 342292018-01-09 Georg-Johann Lay <avr@gjlay.de> 34230 34231 Don't save registers in main(). 34232 34233 PR target/83738 34234 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it. 34235 * config/avr/avr.opt (-mmain-is-OS_task): New target option. 34236 * config/avr/avr.c (avr_set_current_function): Don't error if 34237 naked, OS_task or OS_main are specified at the same time. 34238 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task, 34239 OS_main. 34240 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task 34241 attribute. 34242 * common/config/avr/avr-common.c (avr_option_optimization_table): 34243 Switch on -mmain-is-OS_task for optimizing compilations. 34244 342452018-01-09 Richard Biener <rguenther@suse.de> 34246 34247 PR tree-optimization/83572 34248 * graphite.c: Include cfganal.h. 34249 (graphite_transform_loops): Connect infinite loops to exit 34250 and remove fake edges at the end. 34251 342522018-01-09 Jan Hubicka <hubicka@ucw.cz> 34253 34254 * ipa-inline.c (edge_badness): Revert accidental checkin. 34255 342562018-01-09 Jan Hubicka <hubicka@ucw.cz> 34257 34258 PR ipa/80763 34259 * ipa-comdats.c (set_comdat_group): Only set comdat group of real 34260 symbols; not inline clones. 34261 342622018-01-09 Jakub Jelinek <jakub@redhat.com> 34263 34264 PR target/83507 34265 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move 34266 hard registers. Formatting fixes. 34267 34268 PR preprocessor/83722 34269 * gcc.c (try_generate_repro): Pass 34270 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than 34271 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to 34272 do_report_bug. 34273 342742018-01-08 Monk Chiang <sh.chiang04@gmail.com> 34275 Kito Cheng <kito.cheng@gmail.com> 34276 34277 * config/riscv/riscv.c (machine_function::is_leaf): Remove field. 34278 (riscv_leaf_function_p): Delete. 34279 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE. 34280 342812018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com> 34282 34283 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New 34284 function. 34285 (do_ifelse): New function. 34286 (do_isel): New function. 34287 (do_sub3): New function. 34288 (do_add3): New function. 34289 (do_load_mask_compare): New function. 34290 (do_overlap_load_compare): New function. 34291 (expand_compare_loop): New function. 34292 (expand_block_compare): Call expand_compare_loop() when appropriate. 34293 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change 34294 option description. 34295 (-mblock-compare-inline-loop-limit): New option. 34296 342972018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com> 34298 34299 PR target/83677 34300 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal): 34301 Reverse order of second and third operands in first alternative. 34302 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order 34303 of first and second elements in UNSPEC_VPERMR vector. 34304 (altivec_expand_vec_perm_le): Likewise. 34305 343062018-01-08 Jeff Law <law@redhat.com> 34307 34308 PR rtl-optimizatin/81308 34309 * tree-switch-conversion.c (cfg_altered): New file scoped static. 34310 (process_switch): If group_case_labels makes a change, then set 34311 cfg_altered. 34312 (pass_convert_switch::execute): If a switch is converted, then 34313 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true. 34314 34315 PR rtl-optimization/81308 34316 * recog.c (split_all_insns): Conditionally cleanup the CFG after 34317 splitting insns. 34318 343192018-01-08 Vidya Praveen <vidyapraveen@arm.com> 34320 34321 PR target/83663 - Revert r255946 34322 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code 34323 generation for cases where splatting a value is not useful. 34324 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge 34325 across a vec_duplicate and a paradoxical subreg forming a vector 34326 mode to a vec_concat. 34327 343282018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com> 34329 34330 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for 34331 -march=armv8.3-a variants. 34332 * config/arm/t-multilib: Likewise. 34333 * config/arm/t-arm-elf: Likewise. Handle dotprod extension. 34334 343352018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com> 34336 34337 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it 34338 to generate rtl. 34339 (cceq_ior_compare_complement): Give it a name so I can use it, and 34340 change boolean_or_operator predicate to boolean_operator so it can 34341 be used to generate a crand. 34342 (eqne): New code iterator. 34343 (bd/bd_neg): New code_attrs. 34344 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into 34345 a single define_insn. 34346 (<bd>tf_<mode>): A new insn pattern for the conditional form branch 34347 decrement (bdnzt/bdnzf/bdzt/bdzf). 34348 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated 34349 with the new names of the branch decrement patterns, and added the 34350 names of the branch decrement conditional patterns. 34351 343522018-01-08 Richard Biener <rguenther@suse.de> 34353 34354 PR tree-optimization/83563 34355 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV 34356 cache. 34357 343582018-01-08 Richard Biener <rguenther@suse.de> 34359 34360 PR middle-end/83713 34361 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks. 34362 343632018-01-08 Richard Biener <rguenther@suse.de> 34364 34365 PR tree-optimization/83685 34366 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert 34367 references to abnormals. 34368 343692018-01-08 Richard Biener <rguenther@suse.de> 34370 34371 PR lto/83719 34372 * dwarf2out.c (output_indirect_strings): Handle empty 34373 skeleton_debug_str_hash. 34374 (dwarf2out_early_finish): Index strings for -gsplit-dwarf. 34375 343762018-01-08 Claudiu Zissulescu <claziss@synopsys.com> 34377 34378 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete. 34379 (emit_store_direct): Likewise. 34380 (arc_trampoline_adjust_address): Likewise. 34381 (arc_asm_trampoline_template): New function. 34382 (arc_initialize_trampoline): Use asm_trampoline_template. 34383 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define. 34384 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16. 34385 * config/arc/arc.md (flush_icache): Delete pattern. 34386 343872018-01-08 Claudiu Zissulescu <claziss@synopsys.com> 34388 34389 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define. 34390 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using 34391 munaligned-access. 34392 343932018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de> 34394 34395 PR target/83681 34396 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard 34397 by not USED_FOR_TARGET. 34398 (make_pass_resolve_sw_modes): Likewise. 34399 344002018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de> 34401 34402 * config/nios2/nios2.h (nios2_section_threshold): Guard by not 34403 USED_FOR_TARGET. 34404 344052018-01-08 Richard Biener <rguenther@suse.de> 34406 34407 PR middle-end/83580 34408 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS. 34409 344102018-01-08 Richard Biener <rguenther@suse.de> 34411 34412 PR middle-end/83517 34413 * match.pd ((t * 2) / 2) -> t): Add missing :c. 34414 344152018-01-06 Aldy Hernandez <aldyh@redhat.com> 34416 34417 PR middle-end/81897 34418 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on 34419 basic blocks with a small number of successors. 34420 (convert_control_dep_chain_into_preds): Improve handling of 34421 forwarder blocks. 34422 (dump_predicates): Split apart into... 34423 (dump_pred_chain): ...here... 34424 (dump_pred_info): ...and here. 34425 (can_one_predicate_be_invalidated_p): Add debugging printfs. 34426 (can_chain_union_be_invalidated_p): Improve check for invalidation 34427 of paths. 34428 (uninit_uses_cannot_happen): Avoid unnecessary if 34429 convert_control_dep_chain_into_preds yielded nothing. 34430 344312018-01-06 Martin Sebor <msebor@redhat.com> 34432 34433 PR tree-optimization/83640 34434 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid 34435 subtracting negative offset from size. 34436 (builtin_access::overlap): Adjust offset bounds of the access to fall 34437 within the size of the object if possible. 34438 344392018-01-06 Richard Sandiford <richard.sandiford@linaro.org> 34440 34441 PR rtl-optimization/83699 34442 * expmed.c (extract_bit_field_1): Restrict the vector usage of 34443 extract_bit_field_as_subreg to cases in which the extracted 34444 value is also a vector. 34445 34446 * lra-constraints.c (process_alt_operands): Test for the equivalence 34447 substitutions when detecting a possible reload cycle. 34448 344492018-01-06 Jakub Jelinek <jakub@redhat.com> 34450 34451 PR debug/83480 34452 * toplev.c (process_options): Don't enable debug_nonbind_markers_p 34453 by default if flag_selective_schedling{,2}. Formatting fixes. 34454 34455 PR rtl-optimization/83682 34456 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE 34457 if it has non-VECTOR_MODE element mode. 34458 (vec_duplicate_p): Likewise. 34459 34460 PR middle-end/83694 34461 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode 34462 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT. 34463 344642018-01-05 Jakub Jelinek <jakub@redhat.com> 34465 34466 PR target/83604 34467 * config/i386/i386-builtin.def 34468 (__builtin_ia32_vgf2p8affineinvqb_v64qi, 34469 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi): 34470 Require also OPTION_MASK_ISA_AVX512F in addition to 34471 OPTION_MASK_ISA_GFNI. 34472 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask, 34473 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require 34474 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition 34475 to OPTION_MASK_ISA_GFNI. 34476 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require 34477 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and 34478 OPTION_MASK_ISA_AVX512BW. 34479 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require 34480 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in 34481 addition to OPTION_MASK_ISA_GFNI. 34482 (__builtin_ia32_vgf2p8affineinvqb_v16qi, 34483 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi): 34484 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition 34485 to OPTION_MASK_ISA_GFNI. 34486 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being 34487 a requirement for all ISAs rather than any of them with a few 34488 exceptions. 34489 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before 34490 processing. 34491 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2 34492 bitmasks to be enabled with 3 exceptions, instead of requiring any 34493 enabled ISA with lots of exceptions. 34494 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>, 34495 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>): 34496 Change avx512bw in isa attribute to avx512f. 34497 * config/i386/sgxintrin.h: Add license boilerplate. 34498 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F 34499 to __AVX512F__ and __AVX512VL to __AVX512VL__. 34500 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128, 34501 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not 34502 defined. 34503 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8, 34504 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable 34505 temporarily sse2 rather than sse if not enabled already. 34506 34507 PR target/83604 34508 * config/i386/sse.md (VI248_VLBW): Rename to ... 34509 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW. 34510 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>, 34511 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz, 34512 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask, 34513 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL 34514 mode iterator instead of VI248_VLBW. 34515 345162018-01-05 Jan Hubicka <hubicka@ucw.cz> 34517 34518 * ipa-fnsummary.c (record_modified_bb_info): Add OP. 34519 (record_modified): Skip clobbers; add debug output. 34520 (param_change_prob): Use sreal frequencies. 34521 345222018-01-05 Richard Sandiford <richard.sandiford@linaro.org> 34523 34524 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't 34525 punt for user-aligned variables. 34526 345272018-01-05 Richard Sandiford <richard.sandiford@linaro.org> 34528 34529 * tree-chrec.c (chrec_contains_symbols): Return true for 34530 POLY_INT_CST. 34531 345322018-01-05 Sudakshina Das <sudi.das@arm.com> 34533 34534 PR target/82439 34535 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications 34536 of (x|y) == x for BICS pattern. 34537 345382018-01-05 Jakub Jelinek <jakub@redhat.com> 34539 34540 PR tree-optimization/83605 34541 * gimple-ssa-strength-reduction.c: Include tree-eh.h. 34542 (find_candidates_dom_walker::before_dom_children): Ignore stmts that 34543 can throw. 34544 345452018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de> 34546 34547 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration. 34548 * config/epiphany/rtems.h: New file. 34549 345502018-01-04 Jakub Jelinek <jakub@redhat.com> 34551 Uros Bizjak <ubizjak@gmail.com> 34552 34553 PR target/83554 34554 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use 34555 QIreg_operand instead of register_operand predicate. 34556 * config/i386/i386.c (ix86_rop_should_change_byte_p, 34557 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in 34558 comments instead of -fmitigate[-_]rop. 34559 345602018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE> 34561 34562 PR bootstrap/81926 34563 * cgraphunit.c (symbol_table::compile): Switch to text_section 34564 before calling assembly_start debug hook. 34565 * run-rtl-passes.c (run_rtl_passes): Likewise. 34566 Include output.h. 34567 345682018-01-04 Richard Sandiford <richard.sandiford@linaro.org> 34569 34570 * tree-vrp.c (extract_range_from_binary_expr_1): Check 34571 range_int_cst_p rather than !symbolic_range_p before calling 34572 extract_range_from_multiplicative_op_1. 34573 345742018-01-04 Jeff Law <law@redhat.com> 34575 34576 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove 34577 redundant test in assertion. 34578 345792018-01-04 Richard Sandiford <richard.sandiford@linaro.org> 34580 34581 * doc/rtl.texi: Document machine_mode wrapper classes. 34582 345832018-01-04 Richard Sandiford <richard.sandiford@linaro.org> 34584 34585 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before 34586 using tree_to_uhwi. 34587 345882018-01-04 Richard Sandiford <richard.sandiford@linaro.org> 34589 34590 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow 34591 the VEC_PERM_EXPR fold to fail. 34592 345932018-01-04 Jakub Jelinek <jakub@redhat.com> 34594 34595 PR debug/83585 34596 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition 34597 to switched_sections. 34598 345992018-01-04 Richard Sandiford <richard.sandiford@linaro.org> 34600 34601 PR target/83680 34602 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted 34603 test for d.testing. 34604 346052018-01-04 Peter Bergner <bergner@vnet.ibm.com> 34606 34607 PR target/83387 34608 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not 34609 allow arguments in FP registers if TARGET_HARD_FLOAT is false. 34610 346112018-01-04 Jakub Jelinek <jakub@redhat.com> 34612 34613 PR debug/83666 34614 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode 34615 is BLKmode and bitpos not zero or mode change is needed. 34616 346172018-01-04 Richard Sandiford <richard.sandiford@linaro.org> 34618 34619 PR target/83675 34620 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require 34621 TARGET_VIS2. 34622 346232018-01-04 Uros Bizjak <ubizjak@gmail.com> 34624 34625 PR target/83628 34626 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT 34627 instead of MULT rtx. Update all corresponding splitters. 34628 (*saddl_se): Ditto. 34629 (*ssub<modesuffix>): Ditto. 34630 (*ssubl_se): Ditto. 34631 (*cmp_sadd_di): Update split patterns. 34632 (*cmp_sadd_si): Ditto. 34633 (*cmp_sadd_sidi): Ditto. 34634 (*cmp_ssub_di): Ditto. 34635 (*cmp_ssub_si): Ditto. 34636 (*cmp_ssub_sidi): Ditto. 34637 * config/alpha/predicates.md (const23_operand): New predicate. 34638 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]: 34639 Look for ASHIFT, not MULT inner operand. 34640 (alpha_split_conditional_move): Update for *sadd<modesuffix> change. 34641 346422018-01-04 Martin Liska <mliska@suse.cz> 34643 34644 PR gcov-profile/83669 34645 * gcov.c (output_intermediate_file): Add version to intermediate 34646 gcov file. 34647 * doc/gcov.texi: Document new field 'version' in intermediate 34648 file format. Fix location of '-k' option of gcov command. 34649 346502018-01-04 Martin Liska <mliska@suse.cz> 34651 34652 PR ipa/82352 34653 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary. 34654 346552018-01-04 Jakub Jelinek <jakub@redhat.com> 34656 34657 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi. 34658 346592018-01-03 Martin Sebor <msebor@redhat.com> 34660 34661 PR tree-optimization/83655 34662 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid 34663 checking calls with invalid arguments. 34664 346652018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 34666 34667 * tree-vect-stmts.c (vect_get_store_rhs): New function. 34668 (vectorizable_mask_load_store): Delete. 34669 (vectorizable_call): Return false for masked loads and stores. 34670 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs 34671 instead of gimple_assign_rhs1. 34672 (vectorizable_load): Handle IFN_MASK_LOAD. 34673 (vect_transform_stmt): Don't set is_store for call_vec_info_type. 34674 346752018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 34676 34677 * tree-vect-stmts.c (vect_build_gather_load_calls): New function, 34678 split out from.., 34679 (vectorizable_mask_load_store): ...here. 34680 (vectorizable_load): ...and here. 34681 346822018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 34683 34684 * tree-vect-stmts.c (vect_build_all_ones_mask) 34685 (vect_build_zero_merge_argument): New functions, split out from... 34686 (vectorizable_load): ...here. 34687 346882018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 34689 34690 * tree-vect-stmts.c (vect_check_store_rhs): New function, 34691 split out from... 34692 (vectorizable_mask_load_store): ...here. 34693 (vectorizable_store): ...and here. 34694 346952018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 34696 34697 * tree-vect-stmts.c (vect_check_load_store_mask): New function, 34698 split out from... 34699 (vectorizable_mask_load_store): ...here. 34700 347012018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 34702 34703 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c 34704 (vect_model_store_cost): Take a vec_load_store_type instead of a 34705 vect_def_type. 34706 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h. 34707 (vect_model_store_cost): Take a vec_load_store_type instead of a 34708 vect_def_type. 34709 (vectorizable_mask_load_store): Update accordingly. 34710 (vectorizable_store): Likewise. 34711 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly. 34712 347132018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 34714 34715 * tree-vect-loop.c (vect_transform_loop): Stub out scalar 34716 IFN_MASK_LOAD calls here rather than... 34717 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here. 34718 347192018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 34720 Alan Hayward <alan.hayward@arm.com> 34721 David Sherwood <david.sherwood@arm.com> 34722 34723 * expmed.c (extract_bit_field_1): For vector extracts, 34724 fall back to extract_bit_field_as_subreg if vec_extract 34725 isn't available. 34726 347272018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 34728 Alan Hayward <alan.hayward@arm.com> 34729 David Sherwood <david.sherwood@arm.com> 34730 34731 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether 34732 they are variable or constant sized. 34733 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized 34734 slots for constant-sized data. 34735 347362018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 34737 Alan Hayward <alan.hayward@arm.com> 34738 David Sherwood <david.sherwood@arm.com> 34739 34740 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When 34741 handling COND_EXPRs with boolean comparisons, try to find a better 34742 basis for the mask type than the boolean itself. 34743 347442018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 34745 34746 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default 34747 is calculated and how it can be overridden. 34748 * genmodes.c (max_bitsize_mode_any_mode): New variable. 34749 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE, 34750 if defined. 34751 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE, 34752 if nonzero. 34753 347542018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 34755 Alan Hayward <alan.hayward@arm.com> 34756 David Sherwood <david.sherwood@arm.com> 34757 34758 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate): 34759 Remove the mode argument. 34760 (aarch64_simd_valid_immediate): Remove the mode and inverse 34761 arguments. 34762 * config/aarch64/iterators.md (bitsize): New iterator. 34763 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3) 34764 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate. 34765 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to 34766 aarch64_simd_valid_immediate. 34767 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise. 34768 (aarch64_reg_or_bic_imm): Likewise. 34769 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn 34770 with an insn_type enum and msl with a modifier_type enum. 34771 Replace element_width with a scalar_mode. Change the shift 34772 to unsigned int. Add constructors for scalar_float_mode and 34773 scalar_int_mode elements. 34774 (aarch64_vect_float_const_representable_p): Delete. 34775 (aarch64_can_const_movi_rtx_p) 34776 (aarch64_simd_scalar_immediate_valid_for_move) 34777 (aarch64_simd_make_constant): Update call to 34778 aarch64_simd_valid_immediate. 34779 (aarch64_advsimd_valid_immediate_hs): New function. 34780 (aarch64_advsimd_valid_immediate): Likewise. 34781 (aarch64_simd_valid_immediate): Remove mode and inverse 34782 arguments. Rewrite to use the above. Use const_vec_duplicate_p 34783 to detect duplicated constants and use aarch64_float_const_zero_rtx_p 34784 and aarch64_float_const_representable_p on the result. 34785 (aarch64_output_simd_mov_immediate): Remove mode argument. 34786 Update call to aarch64_simd_valid_immediate and use of 34787 simd_immediate_info. 34788 (aarch64_output_scalar_simd_mov_immediate): Update call 34789 accordingly. 34790 347912018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 34792 Alan Hayward <alan.hayward@arm.com> 34793 David Sherwood <david.sherwood@arm.com> 34794 34795 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION. 34796 (mode_nunits): Likewise CONST_MODE_NUNITS. 34797 * machmode.def (ADJUST_NUNITS): Document. 34798 * genmodes.c (mode_data::need_nunits_adj): New field. 34799 (blank_mode): Update accordingly. 34800 (adj_nunits): New variable. 34801 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ 34802 parameter. 34803 (emit_mode_size_inline): Set need_bytesize_adj for all modes 34804 listed in adj_nunits. 34805 (emit_mode_nunits_inline): Set need_nunits_adj for all modes 34806 listed in adj_nunits. Don't emit case statements for such modes. 34807 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS 34808 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to 34809 nothing if adj_nunits is nonnull. 34810 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl. 34811 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit) 34812 (emit_mode_fbit): Update use of print_maybe_const_decl. 34813 (emit_move_size): Likewise. Treat the array as non-const 34814 if adj_nunits. 34815 (emit_mode_adjustments): Handle adj_nunits. 34816 348172018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 34818 34819 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document. 34820 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro. 34821 (VECTOR_MODES): Use it. 34822 (make_vector_modes): Take the prefix as an argument. 34823 348242018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 34825 Alan Hayward <alan.hayward@arm.com> 34826 David Sherwood <david.sherwood@arm.com> 34827 34828 * mode-classes.def (MODE_VECTOR_BOOL): New mode class. 34829 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true 34830 for MODE_VECTOR_BOOL. 34831 * machmode.def (VECTOR_BOOL_MODE): Document. 34832 * genmodes.c (VECTOR_BOOL_MODE): New macro. 34833 (make_vector_bool_mode): New function. 34834 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle 34835 MODE_VECTOR_BOOL. 34836 * lto-streamer-in.c (lto_input_mode_table): Likewise. 34837 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value): 34838 Likewise. 34839 * stor-layout.c (int_mode_for_mode): Likewise. 34840 * tree.c (build_vector_type_for_mode): Likewise. 34841 * varasm.c (output_constant_pool_2): Likewise. 34842 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and 34843 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx 34844 for MODE_VECTOR_BOOL. 34845 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list 34846 of mode class checks. 34847 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P 34848 instead of a list of mode class checks. 34849 (expand_vector_scalar_condition): Likewise. 34850 (type_for_widest_vector_mode): Handle BImode as an inner mode. 34851 348522018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 34853 Alan Hayward <alan.hayward@arm.com> 34854 David Sherwood <david.sherwood@arm.com> 34855 34856 * machmode.h (mode_size): Change from unsigned short to 34857 poly_uint16_pod. 34858 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short. 34859 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES, 34860 or if measurement_type is not polynomial. 34861 (fixed_size_mode::includes_p): Check for constant-sized modes. 34862 * genmodes.c (emit_mode_size_inline): Make mode_size_inline 34863 return a poly_uint16 rather than an unsigned short. 34864 (emit_mode_size): Change the type of mode_size from unsigned short 34865 to poly_uint16_pod. Use ZERO_COEFFS for the initializer. 34866 (emit_mode_adjustments): Cope with polynomial vector sizes. 34867 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value 34868 for GET_MODE_SIZE. 34869 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value 34870 for GET_MODE_SIZE. 34871 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial. 34872 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise. 34873 * caller-save.c (setup_save_areas): Likewise. 34874 (replace_reg_with_saved_mem): Likewise. 34875 * calls.c (emit_library_call_value_1): Likewise. 34876 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise. 34877 * combine.c (simplify_set, make_extraction, simplify_shift_const_1) 34878 (gen_lowpart_for_combine): Likewise. 34879 * convert.c (convert_to_integer_1): Likewise. 34880 * cse.c (equiv_constant, cse_insn): Likewise. 34881 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise. 34882 (cselib_subst_to_values): Likewise. 34883 * dce.c (word_dce_process_block): Likewise. 34884 * df-problems.c (df_word_lr_mark_ref): Likewise. 34885 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise. 34886 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor) 34887 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor) 34888 (rtl_for_decl_location): Likewise. 34889 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise. 34890 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise. 34891 * expr.c (emit_group_load_1, clear_storage_hints): Likewise. 34892 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise. 34893 (expand_expr_real_1): Likewise. 34894 * function.c (assign_parm_setup_block_p, assign_parm_setup_block) 34895 (pad_below): Likewise. 34896 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise. 34897 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise. 34898 * ira.c (get_subreg_tracking_sizes): Likewise. 34899 * ira-build.c (ira_create_allocno_objects): Likewise. 34900 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise. 34901 (ira_sort_regnos_for_alter_reg): Likewise. 34902 * ira-costs.c (record_operand_costs): Likewise. 34903 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn) 34904 (resolve_simple_move): Likewise. 34905 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise. 34906 (process_addr_reg, simplify_operand_subreg, curr_insn_transform) 34907 (lra_constraints): Likewise. 34908 (CONST_POOL_OK_P): Reject variable-sized modes. 34909 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare) 34910 (add_pseudo_to_slot, lra_spill): Likewise. 34911 * omp-low.c (omp_clause_aligned_alignment): Likewise. 34912 * optabs-query.c (get_best_extraction_insn): Likewise. 34913 * optabs-tree.c (expand_vec_cond_expr_p): Likewise. 34914 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise. 34915 (expand_mult_highpart, valid_multiword_target_p): Likewise. 34916 * recog.c (offsettable_address_addr_space_p): Likewise. 34917 * regcprop.c (maybe_mode_change): Likewise. 34918 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise. 34919 * regrename.c (build_def_use): Likewise. 34920 * regstat.c (dump_reg_info): Likewise. 34921 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload) 34922 (find_reloads, find_reloads_subreg_address): Likewise. 34923 * reload1.c (eliminate_regs_1): Likewise. 34924 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise. 34925 * simplify-rtx.c (avoid_constant_pool_reference): Likewise. 34926 (simplify_binary_operation_1, simplify_subreg): Likewise. 34927 * targhooks.c (default_function_arg_padding): Likewise. 34928 (default_hard_regno_nregs, default_class_max_nregs): Likewise. 34929 * tree-cfg.c (verify_gimple_assign_binary): Likewise. 34930 (verify_gimple_assign_ternary): Likewise. 34931 * tree-inline.c (estimate_move_cost): Likewise. 34932 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise. 34933 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise. 34934 (get_address_cost_ainc): Likewise. 34935 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise. 34936 (vect_supportable_dr_alignment): Likewise. 34937 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise. 34938 (vectorizable_reduction): Likewise. 34939 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift) 34940 (vectorizable_operation, vectorizable_load): Likewise. 34941 * tree.c (build_same_sized_truth_vector_type): Likewise. 34942 * valtrack.c (cleanup_auto_inc_dec): Likewise. 34943 * var-tracking.c (emit_note_insn_var_location): Likewise. 34944 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>. 34945 (ADDR_VEC_ALIGN): Likewise. 34946 349472018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 34948 Alan Hayward <alan.hayward@arm.com> 34949 David Sherwood <david.sherwood@arm.com> 34950 34951 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an 34952 unsigned short. 34953 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES, 34954 or if measurement_type is polynomial. 34955 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial. 34956 * combine.c (make_extraction): Likewise. 34957 * dse.c (find_shift_sequence): Likewise. 34958 * dwarf2out.c (mem_loc_descriptor): Likewise. 34959 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise. 34960 (extract_bit_field, extract_low_bits): Likewise. 34961 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise. 34962 (optimize_bitfield_assignment_op, expand_assignment): Likewise. 34963 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise. 34964 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise. 34965 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise. 34966 * reload.c (find_reloads): Likewise. 34967 * reload1.c (alter_reg): Likewise. 34968 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise. 34969 * targhooks.c (default_secondary_memory_needed_mode): Likewise. 34970 * tree-if-conv.c (predicate_mem_writes): Likewise. 34971 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise. 34972 * tree-vect-patterns.c (adjust_bool_pattern): Likewise. 34973 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise. 34974 * valtrack.c (dead_debug_insert_temp): Likewise. 34975 * varasm.c (mergeable_constant_section): Likewise. 34976 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>. 34977 349782018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 34979 Alan Hayward <alan.hayward@arm.com> 34980 David Sherwood <david.sherwood@arm.com> 34981 34982 * expr.c (expand_assignment): Cope with polynomial mode sizes 34983 when assigning to a CONCAT. 34984 349852018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 34986 Alan Hayward <alan.hayward@arm.com> 34987 David Sherwood <david.sherwood@arm.com> 34988 34989 * machmode.h (mode_precision): Change from unsigned short to 34990 poly_uint16_pod. 34991 (mode_to_precision): Return a poly_uint16 rather than an unsigned 34992 short. 34993 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES, 34994 or if measurement_type is not polynomial. 34995 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case 34996 in which the mode is already known to be a scalar_int_mode. 34997 * genmodes.c (emit_mode_precision): Change the type of mode_precision 34998 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the 34999 initializer. 35000 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value 35001 for GET_MODE_PRECISION. 35002 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value 35003 for GET_MODE_PRECISION. 35004 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION 35005 as polynomial. 35006 (try_combine, find_split_point, combine_simplify_rtx): Likewise. 35007 (expand_field_assignment, make_extraction): Likewise. 35008 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise. 35009 (get_last_value): Likewise. 35010 * convert.c (convert_to_integer_1): Likewise. 35011 * cse.c (cse_insn): Likewise. 35012 * expr.c (expand_expr_real_1): Likewise. 35013 * lra-constraints.c (simplify_operand_subreg): Likewise. 35014 * optabs-query.c (can_atomic_load_p): Likewise. 35015 * optabs.c (expand_atomic_load): Likewise. 35016 (expand_atomic_store): Likewise. 35017 * ree.c (combine_reaching_defs): Likewise. 35018 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise. 35019 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise. 35020 * tree.h (type_has_mode_precision_p): Likewise. 35021 * ubsan.c (instrument_si_overflow): Likewise. 35022 350232018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35024 Alan Hayward <alan.hayward@arm.com> 35025 David Sherwood <david.sherwood@arm.com> 35026 35027 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle 35028 polynomial numbers of units. 35029 (SET_TYPE_VECTOR_SUBPARTS): Likewise. 35030 (valid_vector_subparts_p): New function. 35031 (build_vector_type): Remove temporary shim and take the number 35032 of units as a poly_uint64 rather than an int. 35033 (build_opaque_vector_type): Take the number of units as a 35034 poly_uint64 rather than an int. 35035 * tree.c (build_vector_from_ctor): Handle polynomial 35036 TYPE_VECTOR_SUBPARTS. 35037 (type_hash_canon_hash, type_cache_hasher::equal): Likewise. 35038 (uniform_vector_p, vector_type_mode, build_vector): Likewise. 35039 (build_vector_from_val): If the number of units is variable, 35040 use build_vec_duplicate_cst for constant operands and 35041 VEC_DUPLICATE_EXPR otherwise. 35042 (make_vector_type): Remove temporary is_constant (). 35043 (build_vector_type, build_opaque_vector_type): Take the number of 35044 units as a poly_uint64 rather than an int. 35045 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and 35046 VECTOR_CST_NELTS. 35047 * cfgexpand.c (expand_debug_expr): Likewise. 35048 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise. 35049 (store_constructor, expand_expr_real_1): Likewise. 35050 (const_scalar_mask_from_tree): Likewise. 35051 * fold-const-call.c (fold_const_reduction): Likewise. 35052 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise. 35053 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise. 35054 (native_encode_vector, vec_cst_ctor_to_array): Likewise. 35055 (fold_relational_const): Likewise. 35056 (native_interpret_vector): Likewise. Change the size from an 35057 int to an unsigned int. 35058 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial 35059 TYPE_VECTOR_SUBPARTS. 35060 (gimple_fold_indirect_ref, gimple_build_vector): Likewise. 35061 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when 35062 duplicating a non-constant operand into a variable-length vector. 35063 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial 35064 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS. 35065 * ipa-icf.c (sem_variable::equals): Likewise. 35066 * match.pd: Likewise. 35067 * omp-simd-clone.c (simd_clone_subparts): Likewise. 35068 * print-tree.c (print_node): Likewise. 35069 * stor-layout.c (layout_type): Likewise. 35070 * targhooks.c (default_builtin_vectorization_cost): Likewise. 35071 * tree-cfg.c (verify_gimple_comparison): Likewise. 35072 (verify_gimple_assign_binary): Likewise. 35073 (verify_gimple_assign_ternary): Likewise. 35074 (verify_gimple_assign_single): Likewise. 35075 * tree-pretty-print.c (dump_generic_node): Likewise. 35076 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise. 35077 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise. 35078 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise. 35079 (vect_grouped_load_supported, vect_permute_load_chain): Likewise. 35080 (vect_shift_permute_load_chain): Likewise. 35081 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise. 35082 (expand_vector_condition, optimize_vector_constructor): Likewise. 35083 (lower_vec_perm, get_compute_type): Likewise. 35084 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise. 35085 (get_initial_defs_for_reduction, vect_transform_loop): Likewise. 35086 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise. 35087 (vect_recog_mask_conversion_pattern): Likewise. 35088 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise. 35089 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise. 35090 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise. 35091 (get_group_load_store_type, vectorizable_mask_load_store): Likewise. 35092 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment) 35093 (vectorizable_shift, vectorizable_operation, vectorizable_store) 35094 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison) 35095 (supportable_widening_operation): Likewise. 35096 (supportable_narrowing_operation): Likewise. 35097 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts): 35098 Likewise. 35099 * varasm.c (output_constant): Likewise. 35100 351012018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35102 Alan Hayward <alan.hayward@arm.com> 35103 David Sherwood <david.sherwood@arm.com> 35104 35105 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize 35106 so that both the length == 3 and length != 3 cases set up their 35107 own permute vectors. Add comments explaining why we know the 35108 number of elements is constant. 35109 (vect_permute_load_chain): Likewise. 35110 351112018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35112 Alan Hayward <alan.hayward@arm.com> 35113 David Sherwood <david.sherwood@arm.com> 35114 35115 * machmode.h (mode_nunits): Change from unsigned char to 35116 poly_uint16_pod. 35117 (ONLY_FIXED_SIZE_MODES): New macro. 35118 (pod_mode::measurement_type, scalar_int_mode::measurement_type) 35119 (scalar_float_mode::measurement_type, scalar_mode::measurement_type) 35120 (complex_mode::measurement_type, fixed_size_mode::measurement_type): 35121 New typedefs. 35122 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short. 35123 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES, 35124 or if measurement_type is not polynomial. 35125 * genmodes.c (ZERO_COEFFS): New macro. 35126 (emit_mode_nunits_inline): Make mode_nunits_inline return a 35127 poly_uint16. 35128 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod. 35129 Use ZERO_COEFFS when emitting initializers. 35130 * data-streamer.h (bp_pack_poly_value): New function. 35131 (bp_unpack_poly_value): Likewise. 35132 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value 35133 for GET_MODE_NUNITS. 35134 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value 35135 for GET_MODE_NUNITS. 35136 * tree.c (make_vector_type): Remove temporary shim and make 35137 the real function take the number of units as a poly_uint64 35138 rather than an int. 35139 (build_vector_type_for_mode): Handle polynomial nunits. 35140 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise. 35141 * emit-rtl.c (const_vec_series_p_1): Likewise. 35142 (gen_rtx_CONST_VECTOR): Likewise. 35143 * fold-const.c (test_vec_duplicate_folding): Likewise. 35144 * genrecog.c (validate_pattern): Likewise. 35145 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise. 35146 * optabs-tree.c (expand_vec_cond_expr_p): Likewise. 35147 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise. 35148 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise. 35149 (expand_vec_cond_expr, expand_mult_highpart): Likewise. 35150 * rtlanal.c (subreg_get_info): Likewise. 35151 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise. 35152 (vect_grouped_load_supported): Likewise. 35153 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise. 35154 * tree-vect-loop.c (have_whole_vector_shift): Likewise. 35155 * simplify-rtx.c (simplify_unary_operation_1): Likewise. 35156 (simplify_const_unary_operation, simplify_binary_operation_1) 35157 (simplify_const_binary_operation, simplify_ternary_operation) 35158 (test_vector_ops_duplicate, test_vector_ops): Likewise. 35159 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode 35160 instead of CONST_VECTOR_NUNITS. 35161 * varasm.c (output_constant_pool_2): Likewise. 35162 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the 35163 explicit-encoded elements in the XVEC for variable-length vectors. 35164 351652018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35166 35167 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p. 35168 351692018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35170 Alan Hayward <alan.hayward@arm.com> 35171 David Sherwood <david.sherwood@arm.com> 35172 35173 * coretypes.h (fixed_size_mode): Declare. 35174 (fixed_size_mode_pod): New typedef. 35175 * builtins.h (target_builtins::x_apply_args_mode) 35176 (target_builtins::x_apply_result_mode): Change type to 35177 fixed_size_mode_pod. 35178 * builtins.c (apply_args_size, apply_result_size, result_vector) 35179 (expand_builtin_apply_args_1, expand_builtin_apply) 35180 (expand_builtin_return): Update accordingly. 35181 351822018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35183 35184 * cse.c (hash_rtx_cb): Hash only the encoded elements. 35185 * cselib.c (cselib_hash_rtx): Likewise. 35186 * expmed.c (make_tree): Build VECTOR_CSTs directly from the 35187 CONST_VECTOR encoding. 35188 351892018-01-03 Jakub Jelinek <jakub@redhat.com> 35190 Jeff Law <law@redhat.com> 35191 35192 PR target/83641 35193 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For 35194 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop, 35195 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp 35196 and add REG_CFA_ADJUST_CFA notes in that case to both insns. 35197 35198 PR target/83641 35199 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not 35200 explicitly probe *sp in a noreturn function if there were any callee 35201 register saves or frame pointer is needed. 35202 352032018-01-03 Jakub Jelinek <jakub@redhat.com> 35204 35205 PR debug/83621 35206 * cfgexpand.c (expand_debug_expr): Return NULL if mode is 35207 BLKmode for ternary, binary or unary expressions. 35208 35209 PR debug/83645 35210 * var-tracking.c (delete_vta_debug_insn): New inline function. 35211 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just 35212 insns from get_insns () to NULL instead of each bb separately. 35213 Use delete_vta_debug_insn. No longer static. 35214 (vt_debug_insns_local, variable_tracking_main_1): Adjust 35215 delete_vta_debug_insns callers. 35216 * rtl.h (delete_vta_debug_insns): Declare. 35217 * final.c (rest_of_handle_final): Call delete_vta_debug_insns 35218 instead of variable_tracking_main. 35219 352202018-01-03 Martin Sebor <msebor@redhat.com> 35221 35222 PR tree-optimization/83603 35223 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function 35224 arguments past the endof the argument list in functions declared 35225 without a prototype. 35226 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): 35227 Avoid checking when arguments are null. 35228 352292018-01-03 Martin Sebor <msebor@redhat.com> 35230 35231 PR c/83559 35232 * doc/extend.texi (attribute const): Fix a typo. 35233 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid 35234 issuing -Wsuggest-attribute for void functions. 35235 352362018-01-03 Martin Sebor <msebor@redhat.com> 35237 35238 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use 35239 offset_int::from instead of wide_int::to_shwi. 35240 (maybe_diag_overlap): Remove assertion. 35241 Use HOST_WIDE_INT_PRINT_DEC instead of %lli. 35242 * gimple-ssa-sprintf.c (format_directive): Same. 35243 (parse_directive): Same. 35244 (sprintf_dom_walker::compute_format_length): Same. 35245 (try_substitute_return_value): Same. 35246 352472018-01-03 Jeff Law <law@redhat.com> 35248 35249 PR middle-end/83654 35250 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a 35251 non-constant residual for zero at runtime and avoid probing in 35252 that case. Reorganize code for trailing problem to mirror handling 35253 of the residual. 35254 352552018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org> 35256 35257 PR tree-optimization/83501 35258 * tree-ssa-strlen.c (get_string_cst): New. 35259 (handle_char_store): Call get_string_cst. 35260 352612018-01-03 Martin Liska <mliska@suse.cz> 35262 35263 PR tree-optimization/83593 35264 * tree-ssa-strlen.c: Include tree-cfg.h. 35265 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh. 35266 (strlen_dom_walker): Add new member variable m_cleanup_cfg. 35267 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg 35268 to false. 35269 (strlen_dom_walker::before_dom_children): Call 35270 gimple_purge_dead_eh_edges. Dump tranformation with details 35271 dump flags. 35272 (strlen_dom_walker::before_dom_children): Update call by adding 35273 new argument cleanup_eh. 35274 (pass_strlen::execute): Return TODO_cleanup_cfg if needed. 35275 352762018-01-03 Martin Liska <mliska@suse.cz> 35277 35278 PR ipa/83549 35279 * cif-code.def (VARIADIC_THUNK): New enum value. 35280 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic 35281 thunks. 35282 352832018-01-03 Jan Beulich <jbeulich@suse.com> 35284 35285 * sse.md (mov<mode>_internal): Tighten condition for when to use 35286 vmovdqu<ssescalarsize> for TI and OI modes. 35287 352882018-01-03 Jakub Jelinek <jakub@redhat.com> 35289 35290 Update copyright years. 35291 352922018-01-03 Martin Liska <mliska@suse.cz> 35293 35294 PR ipa/83594 35295 * ipa-visibility.c (function_and_variable_visibility): Skip 35296 functions with noipa attribure. 35297 352982018-01-03 Jakub Jelinek <jakub@redhat.com> 35299 35300 * gcc.c (process_command): Update copyright notice dates. 35301 * gcov-dump.c (print_version): Ditto. 35302 * gcov.c (print_version): Ditto. 35303 * gcov-tool.c (print_version): Ditto. 35304 * gengtype.c (create_file): Ditto. 35305 * doc/cpp.texi: Bump @copying's copyright year. 35306 * doc/cppinternals.texi: Ditto. 35307 * doc/gcc.texi: Ditto. 35308 * doc/gccint.texi: Ditto. 35309 * doc/gcov.texi: Ditto. 35310 * doc/install.texi: Ditto. 35311 * doc/invoke.texi: Ditto. 35312 353132018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35314 35315 * vector-builder.h (vector_builder::m_full_nelts): Change from 35316 unsigned int to poly_uint64. 35317 (vector_builder::full_nelts): Update prototype accordingly. 35318 (vector_builder::new_vector): Likewise. 35319 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts. 35320 (vector_builder::operator ==): Likewise. 35321 (vector_builder::finalize): Likewise. 35322 * int-vector-builder.h (int_vector_builder::int_vector_builder): 35323 Take the number of elements as a poly_uint64 rather than an 35324 unsigned int. 35325 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change 35326 from unsigned int to poly_uint64. 35327 (vec_perm_indices::vec_perm_indices): Update prototype accordingly. 35328 (vec_perm_indices::new_vector): Likewise. 35329 (vec_perm_indices::length): Likewise. 35330 (vec_perm_indices::nelts_per_input): Likewise. 35331 (vec_perm_indices::input_nelts): Likewise. 35332 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the 35333 number of elements per input as a poly_uint64 rather than an 35334 unsigned int. Use the original encoding for variable-length 35335 vectors, rather than clamping each individual element. 35336 For the second and subsequent elements in each pattern, 35337 clamp the step and base before clamping their sum. 35338 (vec_perm_indices::series_p): Handle polynomial element counts. 35339 (vec_perm_indices::all_in_range_p): Likewise. 35340 (vec_perm_indices_to_tree): Likewise. 35341 (vec_perm_indices_to_rtx): Likewise. 35342 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise. 35343 * tree-vector-builder.c (tree_vector_builder::new_unary_operation) 35344 (tree_vector_builder::new_binary_operation): Handle polynomial 35345 element counts. Return false if we need to know the number 35346 of elements at compile time. 35347 * fold-const.c (fold_vec_perm): Punt if the number of elements 35348 isn't known at compile time. 35349 353502018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35351 35352 * vec-perm-indices.h (vec_perm_builder): Change element type 35353 from HOST_WIDE_INT to poly_int64. 35354 (vec_perm_indices::element_type): Update accordingly. 35355 (vec_perm_indices::clamp): Handle polynomial element_types. 35356 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise. 35357 (vec_perm_indices::all_in_range_p): Likewise. 35358 (tree_to_vec_perm_builder): Check for poly_int64 trees rather 35359 than shwi trees. 35360 * vector-builder.h (vector_builder::stepped_sequence_p): Handle 35361 polynomial vec_perm_indices element types. 35362 * int-vector-builder.h (int_vector_builder::equal_p): Likewise. 35363 * fold-const.c (fold_vec_perm): Likewise. 35364 * optabs.c (shift_amt_for_vec_perm_mask): Likewise. 35365 * tree-vect-generic.c (lower_vec_perm): Likewise. 35366 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise. 35367 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm 35368 element type to HOST_WIDE_INT. 35369 353702018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35371 Alan Hayward <alan.hayward@arm.com> 35372 David Sherwood <david.sherwood@arm.com> 35373 35374 * alias.c (addr_side_effect_eval): Take the size as a poly_int64 35375 rather than an int. Use plus_constant. 35376 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints. 35377 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT. 35378 353792018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35380 Alan Hayward <alan.hayward@arm.com> 35381 David Sherwood <david.sherwood@arm.com> 35382 35383 * calls.c (emit_call_1, expand_call): Change struct_value_size from 35384 a HOST_WIDE_INT to a poly_int64. 35385 353862018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35387 Alan Hayward <alan.hayward@arm.com> 35388 David Sherwood <david.sherwood@arm.com> 35389 35390 * calls.c (load_register_parameters): Cope with polynomial 35391 mode sizes. Require a constant size for BLKmode parameters 35392 that aren't described by a PARALLEL. If BLOCK_REG_PADDING 35393 forces a parameter to be padded at the lsb end in order to 35394 fill a complete number of words, require the parameter size 35395 to be ordered wrt UNITS_PER_WORD. 35396 353972018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35398 Alan Hayward <alan.hayward@arm.com> 35399 David Sherwood <david.sherwood@arm.com> 35400 35401 * reload1.c (spill_stack_slot_width): Change element type 35402 from unsigned int to poly_uint64_pod. 35403 (alter_reg): Treat mode sizes as polynomial. 35404 354052018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35406 Alan Hayward <alan.hayward@arm.com> 35407 David Sherwood <david.sherwood@arm.com> 35408 35409 * reload.c (complex_word_subreg_p): New function. 35410 (reload_inner_reg_of_subreg, push_reload): Use it. 35411 354122018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35413 Alan Hayward <alan.hayward@arm.com> 35414 David Sherwood <david.sherwood@arm.com> 35415 35416 * lra-constraints.c (process_alt_operands): Reject matched 35417 operands whose sizes aren't ordered. 35418 (match_reload): Refer to this check here. 35419 354202018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35421 Alan Hayward <alan.hayward@arm.com> 35422 David Sherwood <david.sherwood@arm.com> 35423 35424 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert 35425 that the mode size is in the set {1, 2, 4, 8, 16}. 35426 354272018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35428 Alan Hayward <alan.hayward@arm.com> 35429 David Sherwood <david.sherwood@arm.com> 35430 35431 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial. 35432 Use plus_constant instead of gen_rtx_PLUS. 35433 354342018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35435 Alan Hayward <alan.hayward@arm.com> 35436 David Sherwood <david.sherwood@arm.com> 35437 35438 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare. 35439 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to... 35440 * config/cr16/cr16.c (cr16_push_rounding): ...this new function. 35441 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare. 35442 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to... 35443 * config/h8300/h8300.c (h8300_push_rounding): ...this new function. 35444 * config/i386/i386-protos.h (ix86_push_rounding): Declare. 35445 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to... 35446 * config/i386/i386.c (ix86_push_rounding): ...this new function. 35447 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return 35448 a poly_int64. 35449 * config/m32c/m32c.c (m32c_push_rounding): Likewise. 35450 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare. 35451 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to... 35452 * config/m68k/m68k.c (m68k_push_rounding): ...this new function. 35453 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare. 35454 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to... 35455 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function. 35456 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare. 35457 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to... 35458 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new 35459 function. 35460 * expr.c (emit_move_resolve_push): Treat the input and result 35461 of PUSH_ROUNDING as a poly_int64. 35462 (emit_move_complex_push, emit_single_push_insn_1): Likewise. 35463 (emit_push_insn): Likewise. 35464 * lra-eliminations.c (mark_not_eliminable): Likewise. 35465 * recog.c (push_operand): Likewise. 35466 * reload1.c (elimination_effects): Likewise. 35467 * rtlanal.c (nonzero_bits1): Likewise. 35468 * calls.c (store_one_arg): Likewise. Require the padding to be 35469 known at compile time. 35470 354712018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35472 Alan Hayward <alan.hayward@arm.com> 35473 David Sherwood <david.sherwood@arm.com> 35474 35475 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial. 35476 Use plus_constant instead of gen_rtx_PLUS. 35477 354782018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35479 Alan Hayward <alan.hayward@arm.com> 35480 David Sherwood <david.sherwood@arm.com> 35481 35482 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64 35483 rather than an int. 35484 354852018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35486 Alan Hayward <alan.hayward@arm.com> 35487 David Sherwood <david.sherwood@arm.com> 35488 35489 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64 35490 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs 35491 via stack temporaries. Treat the mode size as polynomial too. 35492 354932018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35494 Alan Hayward <alan.hayward@arm.com> 35495 David Sherwood <david.sherwood@arm.com> 35496 35497 * expr.c (expand_expr_real_2): When handling conversions involving 35498 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than 35499 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE 35500 as a poly_uint64 too. 35501 355022018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35503 Alan Hayward <alan.hayward@arm.com> 35504 David Sherwood <david.sherwood@arm.com> 35505 35506 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes. 35507 355082018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35509 Alan Hayward <alan.hayward@arm.com> 35510 David Sherwood <david.sherwood@arm.com> 35511 35512 * combine.c (can_change_dest_mode): Handle polynomial 35513 REGMODE_NATURAL_SIZE. 35514 * expmed.c (store_bit_field_1): Likewise. 35515 * expr.c (store_constructor): Likewise. 35516 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes 35517 and polynomial REGMODE_NATURAL_SIZE. 35518 (gen_lowpart_common): Likewise. 35519 * reginfo.c (record_subregs_of_mode): Likewise. 35520 * rtlanal.c (read_modify_subreg_p): Likewise. 35521 355222018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35523 Alan Hayward <alan.hayward@arm.com> 35524 David Sherwood <david.sherwood@arm.com> 35525 35526 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial 35527 numbers of elements. 35528 355292018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35530 Alan Hayward <alan.hayward@arm.com> 35531 David Sherwood <david.sherwood@arm.com> 35532 35533 * match.pd: Cope with polynomial numbers of vector elements. 35534 355352018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35536 Alan Hayward <alan.hayward@arm.com> 35537 David Sherwood <david.sherwood@arm.com> 35538 35539 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets 35540 in a POINTER_PLUS_EXPR. 35541 355422018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35543 Alan Hayward <alan.hayward@arm.com> 35544 David Sherwood <david.sherwood@arm.com> 35545 35546 * omp-simd-clone.c (simd_clone_subparts): New function. 35547 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS. 35548 (ipa_simd_modify_function_body): Likewise. 35549 355502018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35551 Alan Hayward <alan.hayward@arm.com> 35552 David Sherwood <david.sherwood@arm.com> 35553 35554 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function. 35555 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS. 35556 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise. 35557 (expand_vector_condition, vector_element): Likewise. 35558 (subparts_gt): New function. 35559 (get_compute_type): Use subparts_gt. 35560 (count_type_subparts): Delete. 35561 (expand_vector_operations_1): Use subparts_gt instead of 35562 count_type_subparts. 35563 355642018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35565 Alan Hayward <alan.hayward@arm.com> 35566 David Sherwood <david.sherwood@arm.com> 35567 35568 * tree-vect-data-refs.c (vect_no_alias_p): Replace with... 35569 (vect_compile_time_alias): ...this new function. Do the calculation 35570 on poly_ints rather than trees. 35571 (vect_prune_runtime_alias_test_list): Update call accordingly. 35572 355732018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35574 Alan Hayward <alan.hayward@arm.com> 35575 David Sherwood <david.sherwood@arm.com> 35576 35577 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial 35578 numbers of units. 35579 (vect_schedule_slp_instance): Likewise. 35580 355812018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35582 Alan Hayward <alan.hayward@arm.com> 35583 David Sherwood <david.sherwood@arm.com> 35584 35585 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject 35586 constant and extern definitions for variable-length vectors. 35587 (vect_get_constant_vectors): Note that the number of units 35588 is known to be constant. 35589 355902018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35591 Alan Hayward <alan.hayward@arm.com> 35592 David Sherwood <david.sherwood@arm.com> 35593 35594 * tree-vect-stmts.c (vectorizable_conversion): Treat the number 35595 of units as polynomial. Choose between WIDE and NARROW based 35596 on multiple_p. 35597 355982018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35599 Alan Hayward <alan.hayward@arm.com> 35600 David Sherwood <david.sherwood@arm.com> 35601 35602 * tree-vect-stmts.c (simd_clone_subparts): New function. 35603 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS. 35604 356052018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35606 Alan Hayward <alan.hayward@arm.com> 35607 David Sherwood <david.sherwood@arm.com> 35608 35609 * tree-vect-stmts.c (vectorizable_call): Treat the number of 35610 vectors as polynomial. Use build_index_vector for 35611 IFN_GOMP_SIMD_LANE. 35612 356132018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35614 Alan Hayward <alan.hayward@arm.com> 35615 David Sherwood <david.sherwood@arm.com> 35616 35617 * tree-vect-stmts.c (get_load_store_type): Treat the number of 35618 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP 35619 for variable-length vectors. 35620 (vectorizable_mask_load_store): Treat the number of units as 35621 polynomial, asserting that it is constant if the condition has 35622 already been enforced. 35623 (vectorizable_store, vectorizable_load): Likewise. 35624 356252018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35626 Alan Hayward <alan.hayward@arm.com> 35627 David Sherwood <david.sherwood@arm.com> 35628 35629 * tree-vect-loop.c (vectorizable_live_operation): Treat the number 35630 of units as polynomial. Punt if we can't tell at compile time 35631 which vector contains the final result. 35632 356332018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35634 Alan Hayward <alan.hayward@arm.com> 35635 David Sherwood <david.sherwood@arm.com> 35636 35637 * tree-vect-loop.c (vectorizable_induction): Treat the number 35638 of units as polynomial. Punt on SLP inductions. Use an integer 35639 VEC_SERIES_EXPR for variable-length integer reductions. Use a 35640 cast of such a series for variable-length floating-point 35641 reductions. 35642 356432018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35644 Alan Hayward <alan.hayward@arm.com> 35645 David Sherwood <david.sherwood@arm.com> 35646 35647 * tree.h (build_index_vector): Declare. 35648 * tree.c (build_index_vector): New function. 35649 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number 35650 of units as polynomial, forcibly converting it to a constant if 35651 vectorizable_reduction has already enforced the condition. 35652 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector 35653 to create a {1,2,3,...} vector. 35654 (vectorizable_reduction): Treat the number of units as polynomial. 35655 Choose vectype_in based on the largest scalar element size rather 35656 than the smallest number of units. Enforce the restrictions 35657 relied on above. 35658 356592018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35660 Alan Hayward <alan.hayward@arm.com> 35661 David Sherwood <david.sherwood@arm.com> 35662 35663 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the 35664 number of units as polynomial. 35665 356662018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35667 Alan Hayward <alan.hayward@arm.com> 35668 David Sherwood <david.sherwood@arm.com> 35669 35670 * target.h (vector_sizes, auto_vector_sizes): New typedefs. 35671 * target.def (autovectorize_vector_sizes): Return the vector sizes 35672 by pointer, using vector_sizes rather than a bitmask. 35673 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly. 35674 * targhooks.c (default_autovectorize_vector_sizes): Likewise. 35675 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes): 35676 Likewise. 35677 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise. 35678 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise. 35679 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise. 35680 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise. 35681 * omp-general.c (omp_max_vf): Likewise. 35682 * omp-low.c (omp_clause_aligned_alignment): Likewise. 35683 * optabs-query.c (can_vec_mask_load_store_p): Likewise. 35684 * tree-vect-loop.c (vect_analyze_loop): Likewise. 35685 * tree-vect-slp.c (vect_slp_bb): Likewise. 35686 * doc/tm.texi: Regenerate. 35687 * tree-vectorizer.h (current_vector_size): Change from an unsigned int 35688 to a poly_uint64. 35689 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take 35690 the vector size as a poly_uint64 rather than an unsigned int. 35691 (current_vector_size): Change from an unsigned int to a poly_uint64. 35692 (get_vectype_for_scalar_type): Update accordingly. 35693 * tree.h (build_truth_vector_type): Take the size and number of 35694 units as a poly_uint64 rather than an unsigned int. 35695 (build_vector_type): Add a temporary overload that takes 35696 the number of units as a poly_uint64 rather than an unsigned int. 35697 * tree.c (make_vector_type): Likewise. 35698 (build_truth_vector_type): Take the number of units as a poly_uint64 35699 rather than an unsigned int. 35700 357012018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35702 Alan Hayward <alan.hayward@arm.com> 35703 David Sherwood <david.sherwood@arm.com> 35704 35705 * target.def (get_mask_mode): Take the number of units and length 35706 as poly_uint64s rather than unsigned ints. 35707 * targhooks.h (default_get_mask_mode): Update accordingly. 35708 * targhooks.c (default_get_mask_mode): Likewise. 35709 * config/i386/i386.c (ix86_get_mask_mode): Likewise. 35710 * doc/tm.texi: Regenerate. 35711 357122018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35713 Alan Hayward <alan.hayward@arm.com> 35714 David Sherwood <david.sherwood@arm.com> 35715 35716 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int. 35717 * omp-general.c (omp_max_vf): Likewise. 35718 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf. 35719 (expand_omp_simd): Handle polynomial safelen. 35720 * omp-low.c (omplow_simd_context): Add a default constructor. 35721 (omplow_simd_context::max_vf): Change from int to poly_uint64. 35722 (lower_rec_simd_input_clauses): Update accordingly. 35723 (lower_rec_input_clauses): Likewise. 35724 357252018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35726 Alan Hayward <alan.hayward@arm.com> 35727 David Sherwood <david.sherwood@arm.com> 35728 35729 * tree-vectorizer.h (vect_nunits_for_cost): New function. 35730 * tree-vect-loop.c (vect_model_reduction_cost): Use it. 35731 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise. 35732 (vect_analyze_slp_cost): Likewise. 35733 * tree-vect-stmts.c (vect_model_store_cost): Likewise. 35734 (vect_model_load_cost): Likewise. 35735 357362018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35737 Alan Hayward <alan.hayward@arm.com> 35738 David Sherwood <david.sherwood@arm.com> 35739 35740 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1) 35741 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits 35742 from an unsigned int * to a poly_uint64_pod *. 35743 (calculate_unrolling_factor): New function. 35744 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits. 35745 357462018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35747 Alan Hayward <alan.hayward@arm.com> 35748 David Sherwood <david.sherwood@arm.com> 35749 35750 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change 35751 from an unsigned int to a poly_uint64. 35752 (_loop_vec_info::slp_unrolling_factor): Likewise. 35753 (_loop_vec_info::vectorization_factor): Change from an int 35754 to a poly_uint64. 35755 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX. 35756 (vect_get_num_vectors): New function. 35757 (vect_update_max_nunits, vect_vf_for_cost): Likewise. 35758 (vect_get_num_copies): Use vect_get_num_vectors. 35759 (vect_analyze_data_ref_dependences): Change max_vf from an int * 35760 to an unsigned int *. 35761 (vect_analyze_data_refs): Change min_vf from an int * to a 35762 poly_uint64 *. 35763 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather 35764 than an unsigned HOST_WIDE_INT. 35765 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr) 35766 (vect_analyze_data_ref_dependence): Change max_vf from an int * 35767 to an unsigned int *. 35768 (vect_analyze_data_ref_dependences): Likewise. 35769 (vect_compute_data_ref_alignment): Handle polynomial vf. 35770 (vect_enhance_data_refs_alignment): Likewise. 35771 (vect_prune_runtime_alias_test_list): Likewise. 35772 (vect_shift_permute_load_chain): Likewise. 35773 (vect_supportable_dr_alignment): Likewise. 35774 (dependence_distance_ge_vf): Take the vectorization factor as a 35775 poly_uint64 rather than an unsigned HOST_WIDE_INT. 35776 (vect_analyze_data_refs): Change min_vf from an int * to a 35777 poly_uint64 *. 35778 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take 35779 vfm1 as a poly_uint64 rather than an int. Make the same change 35780 for the returned bound_scalar. 35781 (vect_gen_vector_loop_niters): Handle polynomial vf. 35782 (vect_do_peeling): Likewise. Update call to 35783 vect_gen_scalar_loop_niters and handle polynomial bound_scalars. 35784 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must 35785 be constant. 35786 * tree-vect-loop.c (vect_determine_vectorization_factor) 35787 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf. 35788 (vect_get_known_peeling_cost): Likewise. 35789 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise. 35790 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise. 35791 (vect_transform_loop): Likewise. Use the lowest possible VF when 35792 updating the upper bounds of the loop. 35793 (vect_min_worthwhile_factor): Make static. Return an unsigned int 35794 rather than an int. 35795 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with 35796 polynomial unroll factors. 35797 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise. 35798 (vect_make_slp_decision): Likewise. 35799 (vect_supported_load_permutation_p): Likewise, and polynomial 35800 vf too. 35801 (vect_analyze_slp_cost): Handle polynomial vf. 35802 (vect_slp_analyze_node_operations): Likewise. 35803 (vect_slp_analyze_bb_1): Likewise. 35804 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather 35805 than an unsigned HOST_WIDE_INT. 35806 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store) 35807 (vectorizable_load): Handle polynomial vf. 35808 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to 35809 a poly_uint64. 35810 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly. 35811 358122018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35813 Alan Hayward <alan.hayward@arm.com> 35814 David Sherwood <david.sherwood@arm.com> 35815 35816 * match.pd: Handle bit operations involving three constants 35817 and try to fold one pair. 35818 358192018-01-03 Richard Sandiford <richard.sandiford@linaro.org> 35820 35821 * tree-vect-loop-manip.c: Include gimple-fold.h. 35822 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and 35823 niters_maybe_zero parameters. Handle other cases besides a step of 1. 35824 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter. 35825 Add a path that uses a step of VF instead of 1, but disable it 35826 for now. 35827 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var 35828 and niters_no_overflow parameters. Update calls to 35829 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters. 35830 Create a new SSA name if the latter choses to use a ste other 35831 than zero, and return it via niters_vector_mult_vf_var. 35832 * tree-vect-loop.c (vect_transform_loop): Update calls to 35833 vect_do_peeling, vect_gen_vector_loop_niters and 35834 slpeel_make_loop_iterate_ntimes. 35835 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling) 35836 (vect_gen_vector_loop_niters): Update declarations after above changes. 35837 358382018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com> 35839 35840 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE 35841 128-bit round to integer instructions. 35842 (ceil<mode>2): Likewise. 35843 (btrunc<mode>2): Likewise. 35844 (round<mode>2): Likewise. 35845 358462018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com> 35847 35848 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of 35849 unaligned VSX load/store on P8/P9. 35850 (expand_block_clear): Allow the use of unaligned VSX 35851 load/store on P8/P9. 35852 358532018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com> 35854 35855 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store): 35856 New function. 35857 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a 35858 swap associated with both a load and a store. 35859 358602018-01-02 Andrew Waterman <andrew@sifive.com> 35861 35862 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New. 35863 * config/riscv/riscv.md (clear_cache): Use it. 35864 358652018-01-02 Artyom Skrobov <tyomitch@gmail.com> 35866 35867 * web.c: Remove out-of-date comment. 35868 358692018-01-02 Richard Sandiford <richard.sandiford@linaro.org> 35870 35871 * expr.c (fixup_args_size_notes): Check that any existing 35872 REG_ARGS_SIZE notes are correct, and don't try to re-add them. 35873 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to... 35874 (emit_single_push_insn): ...here. 35875 358762018-01-02 Richard Sandiford <richard.sandiford@linaro.org> 35877 35878 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt. 35879 (const_vector_encoded_nelts): New function. 35880 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS. 35881 (const_vector_int_elt, const_vector_elt): Declare. 35882 * emit-rtl.c (const_vector_int_elt_1): New function. 35883 (const_vector_elt): Likewise. 35884 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address 35885 of CONST_VECTOR_ELT. 35886 358872018-01-02 Richard Sandiford <richard.sandiford@linaro.org> 35888 35889 * expr.c: Include rtx-vector-builder.h. 35890 (const_vector_mask_from_tree): Use rtx_vector_builder and operate 35891 directly on the tree encoding. 35892 (const_vector_from_tree): Likewise. 35893 * optabs.c: Include rtx-vector-builder.h. 35894 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating 35895 sequence of "u" values. 35896 * vec-perm-indices.c: Include rtx-vector-builder.h. 35897 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate 35898 directly on the vec_perm_indices encoding. 35899 359002018-01-02 Richard Sandiford <richard.sandiford@linaro.org> 35901 35902 * doc/rtl.texi (const_vector): Describe new encoding scheme. 35903 * Makefile.in (OBJS): Add rtx-vector-builder.o. 35904 * rtx-vector-builder.h: New file. 35905 * rtx-vector-builder.c: Likewise. 35906 * rtl.h (rtx_def::u2): Add a const_vector field. 35907 (CONST_VECTOR_NPATTERNS): New macro. 35908 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise. 35909 (CONST_VECTOR_DUPLICATE_P): Likewise. 35910 (CONST_VECTOR_STEPPED_P): Likewise. 35911 (CONST_VECTOR_ENCODED_ELT): Likewise. 35912 (const_vec_duplicate_p): Check for a duplicated vector encoding. 35913 (unwrap_const_vec_duplicate): Likewise. 35914 (const_vec_series_p): Check for a non-duplicated vector encoding. 35915 Say that the function only returns true for integer vectors. 35916 * emit-rtl.c: Include rtx-vector-builder.h. 35917 (gen_const_vec_duplicate_1): Delete. 35918 (gen_const_vector): Call gen_const_vec_duplicate instead of 35919 gen_const_vec_duplicate_1. 35920 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding. 35921 (gen_const_vec_duplicate): Use rtx_vector_builder. 35922 (gen_const_vec_series): Likewise. 35923 (gen_rtx_CONST_VECTOR): Likewise. 35924 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h. 35925 (swap_const_vector_halves): Take an rtx pointer rather than rtx. 35926 Build a new vector rather than modifying a CONST_VECTOR in-place. 35927 (handle_special_swappables): Update call accordingly. 35928 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h. 35929 (swap_const_vector_halves): Take an rtx pointer rather than rtx. 35930 Build a new vector rather than modifying a CONST_VECTOR in-place. 35931 (handle_special_swappables): Update call accordingly. 35932 359332018-01-02 Richard Sandiford <richard.sandiford@linaro.org> 35934 35935 * simplify-rtx.c (simplify_const_binary_operation): Use 35936 CONST_VECTOR_ELT instead of XVECEXP. 35937 359382018-01-02 Richard Sandiford <richard.sandiford@linaro.org> 35939 35940 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of 35941 the selector elements to be different from the data elements 35942 if the selector is a VECTOR_CST. 35943 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of 35944 ssizetype for the selector. 35945 359462018-01-02 Richard Sandiford <richard.sandiford@linaro.org> 35947 35948 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p 35949 before testing each element individually. 35950 * tree-vect-generic.c (lower_vec_perm): Likewise. 35951 359522018-01-02 Richard Sandiford <richard.sandiford@linaro.org> 35953 35954 * selftest.h (selftest::vec_perm_indices_c_tests): Declare. 35955 * selftest-run-tests.c (selftest::run_tests): Call it. 35956 * vector-builder.h (vector_builder::operator ==): New function. 35957 (vector_builder::operator !=): Likewise. 35958 * vec-perm-indices.h (vec_perm_indices::series_p): Declare. 35959 (vec_perm_indices::all_from_input_p): New function. 35960 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise. 35961 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise. 35962 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder 35963 instead of reading the VECTOR_CST directly. Detect whether both 35964 vector inputs are the same before constructing the vec_perm_indices, 35965 and update the number of inputs argument accordingly. Use the 35966 utility functions added above. Only construct sel2 if we need to. 35967 359682018-01-02 Richard Sandiford <richard.sandiford@linaro.org> 35969 35970 * optabs.c (expand_vec_perm_var): Use an explicit encoding for 35971 the broadcast of the low byte. 35972 (expand_mult_highpart): Use an explicit encoding for the permutes. 35973 * optabs-query.c (can_mult_highpart_p): Likewise. 35974 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise. 35975 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise. 35976 (vectorizable_bswap): Likewise. 35977 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an 35978 explicit encoding for the power-of-2 permutes. 35979 (vect_permute_store_chain): Likewise. 35980 (vect_grouped_load_supported): Likewise. 35981 (vect_permute_load_chain): Likewise. 35982 359832018-01-02 Richard Sandiford <richard.sandiford@linaro.org> 35984 35985 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare. 35986 * vec-perm-indices.c (vec_perm_indices_to_tree): New function. 35987 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it. 35988 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise. 35989 * tree-vect-stmts.c (vectorizable_bswap): Likewise. 35990 (vect_gen_perm_mask_any): Likewise. 35991 359922018-01-02 Richard Sandiford <richard.sandiford@linaro.org> 35993 35994 * int-vector-builder.h: New file. 35995 * vec-perm-indices.h: Include int-vector-builder.h. 35996 (vec_perm_indices): Redefine as an int_vector_builder. 35997 (auto_vec_perm_indices): Delete. 35998 (vec_perm_builder): Redefine as a stand-alone class. 35999 (vec_perm_indices::vec_perm_indices): New function. 36000 (vec_perm_indices::clamp): Likewise. 36001 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h. 36002 (vec_perm_indices::new_vector): New function. 36003 (vec_perm_indices::new_expanded_vector): Update for new 36004 vec_perm_indices class. 36005 (vec_perm_indices::rotate_inputs): New function. 36006 (vec_perm_indices::all_in_range_p): Operate directly on the 36007 encoded form, without computing elided elements. 36008 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST 36009 encoding. Update for new vec_perm_indices class. 36010 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for 36011 the given vec_perm_builder. 36012 (expand_vec_perm_var): Update vec_perm_builder constructor. 36013 (expand_mult_highpart): Use vec_perm_builder instead of 36014 auto_vec_perm_indices. 36015 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and 36016 vec_perm_indices instead of auto_vec_perm_indices. Use a single 36017 or double series encoding as appropriate. 36018 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and 36019 vec_perm_indices instead of auto_vec_perm_indices. 36020 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise. 36021 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise. 36022 (vect_permute_store_chain): Likewise. 36023 (vect_grouped_load_supported): Likewise. 36024 (vect_permute_load_chain): Likewise. 36025 (vect_shift_permute_load_chain): Likewise. 36026 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise. 36027 (vect_transform_slp_perm_load): Likewise. 36028 (vect_schedule_slp_instance): Likewise. 36029 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise. 36030 (vectorizable_mask_load_store): Likewise. 36031 (vectorizable_bswap): Likewise. 36032 (vectorizable_store): Likewise. 36033 (vectorizable_load): Likewise. 36034 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and 36035 vec_perm_indices instead of auto_vec_perm_indices. Use 36036 tree_to_vec_perm_builder to read the vector from a tree. 36037 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a 36038 vec_perm_builder instead of a vec_perm_indices. 36039 (have_whole_vector_shift): Use vec_perm_builder and 36040 vec_perm_indices instead of auto_vec_perm_indices. Leave the 36041 truncation to calc_vec_perm_mask_for_shift. 36042 (vect_create_epilog_for_reduction): Likewise. 36043 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change 36044 from auto_vec_perm_indices to vec_perm_indices. 36045 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm 36046 instead of changing individual elements. 36047 (aarch64_vectorize_vec_perm_const): Use new_vector to install 36048 the vector in d.perm. 36049 * config/arm/arm.c (expand_vec_perm_d::perm): Change 36050 from auto_vec_perm_indices to vec_perm_indices. 36051 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm 36052 instead of changing individual elements. 36053 (arm_vectorize_vec_perm_const): Use new_vector to install 36054 the vector in d.perm. 36055 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even): 36056 Update vec_perm_builder constructor. 36057 (rs6000_expand_interleave): Likewise. 36058 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise. 36059 (rs6000_expand_interleave): Likewise. 36060 360612018-01-02 Richard Sandiford <richard.sandiford@linaro.org> 36062 36063 * optabs-query.c (can_vec_perm_var_p): Check whether lowering 36064 to qimode could truncate the indices. 36065 * optabs.c (expand_vec_perm_var): Likewise. 36066 360672018-01-02 Richard Sandiford <richard.sandiford@linaro.org> 36068 36069 * Makefile.in (OBJS): Add vec-perm-indices.o. 36070 * vec-perm-indices.h: New file. 36071 * vec-perm-indices.c: Likewise. 36072 * target.h (vec_perm_indices): Replace with a forward class 36073 declaration. 36074 (auto_vec_perm_indices): Move to vec-perm-indices.h. 36075 * optabs.h: Include vec-perm-indices.h. 36076 (expand_vec_perm): Delete. 36077 (selector_fits_mode_p, expand_vec_perm_var): Declare. 36078 (expand_vec_perm_const): Declare. 36079 * target.def (vec_perm_const_ok): Replace with... 36080 (vec_perm_const): ...this new hook. 36081 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with... 36082 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook. 36083 * doc/tm.texi: Regenerate. 36084 * optabs.def (vec_perm_const): Delete. 36085 * doc/md.texi (vec_perm_const): Likewise. 36086 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST. 36087 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than 36088 expand_vec_perm for constant permutation vectors. Assert that 36089 the mode of variable permutation vectors is the integer equivalent 36090 of the mode that is being permuted. 36091 * optabs-query.h (selector_fits_mode_p): Declare. 36092 * optabs-query.c: Include vec-perm-indices.h. 36093 (selector_fits_mode_p): New function. 36094 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const 36095 is defined, instead of checking whether the vec_perm_const_optab 36096 exists. Use targetm.vectorize.vec_perm_const instead of 36097 targetm.vectorize.vec_perm_const_ok. Check whether the indices 36098 fit in the vector mode before using a variable permute. 36099 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a 36100 vec_perm_indices instead of an rtx. 36101 (expand_vec_perm): Replace with... 36102 (expand_vec_perm_const): ...this new function. Take the selector 36103 as a vec_perm_indices rather than an rtx. Also take the mode of 36104 the selector. Update call to shift_amt_for_vec_perm_mask. 36105 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab. 36106 Use vec_perm_indices::new_expanded_vector to expand the original 36107 selector into bytes. Check whether the indices fit in the vector 36108 mode before using a variable permute. 36109 (expand_vec_perm_var): Make global. 36110 (expand_mult_highpart): Use expand_vec_perm_const. 36111 * fold-const.c: Includes vec-perm-indices.h. 36112 * tree-ssa-forwprop.c: Likewise. 36113 * tree-vect-data-refs.c: Likewise. 36114 * tree-vect-generic.c: Likewise. 36115 * tree-vect-loop.c: Likewise. 36116 * tree-vect-slp.c: Likewise. 36117 * tree-vect-stmts.c: Likewise. 36118 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const): 36119 Delete. 36120 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete. 36121 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const) 36122 (aarch64_vectorize_vec_perm_const_ok): Fuse into... 36123 (aarch64_vectorize_vec_perm_const): ...this new function. 36124 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete. 36125 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine. 36126 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete. 36127 * config/arm/vec-common.md (vec_perm_const<mode>): Delete. 36128 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete. 36129 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine. 36130 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge 36131 into... 36132 (arm_vectorize_vec_perm_const): ...this new function. Explicitly 36133 check for NEON modes. 36134 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete. 36135 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete. 36136 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment. 36137 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge 36138 into... 36139 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate 36140 the old VEC_PERM_CONST conditions. 36141 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete. 36142 * config/ia64/vect.md (vec_perm_const<mode>): Delete. 36143 * config/ia64/ia64.c (ia64_expand_vec_perm_const) 36144 (ia64_vectorize_vec_perm_const_ok): Merge into... 36145 (ia64_vectorize_vec_perm_const): ...this new function. 36146 * config/mips/loongson.md (vec_perm_const<mode>): Delete. 36147 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete. 36148 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete. 36149 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete. 36150 * config/mips/mips.c (mips_expand_vec_perm_const) 36151 (mips_vectorize_vec_perm_const_ok): Merge into... 36152 (mips_vectorize_vec_perm_const): ...this new function. 36153 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete. 36154 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete. 36155 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete. 36156 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete. 36157 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const) 36158 (rs6000_expand_vec_perm_const): Delete. 36159 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): 36160 Delete. 36161 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine. 36162 (altivec_expand_vec_perm_const_le): Take each operand individually. 36163 Operate on constant selectors rather than rtxes. 36164 (altivec_expand_vec_perm_const): Likewise. Update call to 36165 altivec_expand_vec_perm_const_le. 36166 (rs6000_expand_vec_perm_const): Delete. 36167 (rs6000_vectorize_vec_perm_const_ok): Delete. 36168 (rs6000_vectorize_vec_perm_const): New function. 36169 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of 36170 an element count and rtx array. 36171 (rs6000_expand_extract_even): Update call accordingly. 36172 (rs6000_expand_interleave): Likewise. 36173 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete. 36174 * config/rs6000/paired.md (vec_perm_constv2sf): Delete. 36175 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete. 36176 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const) 36177 (rs6000_expand_vec_perm_const): Delete. 36178 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete. 36179 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine. 36180 (altivec_expand_vec_perm_const_le): Take each operand individually. 36181 Operate on constant selectors rather than rtxes. 36182 (altivec_expand_vec_perm_const): Likewise. Update call to 36183 altivec_expand_vec_perm_const_le. 36184 (rs6000_expand_vec_perm_const): Delete. 36185 (rs6000_vectorize_vec_perm_const_ok): Delete. 36186 (rs6000_vectorize_vec_perm_const): New function. Remove stray 36187 reference to the SPE evmerge intructions. 36188 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of 36189 an element count and rtx array. 36190 (rs6000_expand_extract_even): Update call accordingly. 36191 (rs6000_expand_interleave): Likewise. 36192 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of... 36193 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this 36194 new function. 36195 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine. 36196 361972018-01-02 Richard Sandiford <richard.sandiford@linaro.org> 36198 36199 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer 36200 vector mode and that that mode matches the mode of the data 36201 being permuted. 36202 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors 36203 out into expand_vec_perm_var. Do all CONST_VECTOR handling here, 36204 directly using expand_vec_perm_1 when forcing selectors into 36205 registers. 36206 (expand_vec_perm_var): New function, split out from expand_vec_perm. 36207 362082018-01-02 Richard Sandiford <richard.sandiford@linaro.org> 36209 36210 * optabs-query.h (can_vec_perm_p): Delete. 36211 (can_vec_perm_var_p, can_vec_perm_const_p): Declare. 36212 * optabs-query.c (can_vec_perm_p): Split into... 36213 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions. 36214 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a 36215 particular selector is valid. 36216 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise. 36217 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise. 36218 (vect_grouped_load_supported): Likewise. 36219 (vect_shift_permute_load_chain): Likewise. 36220 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise. 36221 (vect_transform_slp_perm_load): Likewise. 36222 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise. 36223 (vectorizable_bswap): Likewise. 36224 (vect_gen_perm_mask_checked): Likewise. 36225 * fold-const.c (fold_ternary_loc): Likewise. Don't take 36226 implementations of variable permutation vectors into account 36227 when deciding which selector to use. 36228 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether 36229 vec_perm_const_optab is supported; instead use can_vec_perm_const_p 36230 with a false third argument. 36231 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p 36232 to test whether the constant selector is valid and can_vec_perm_var_p 36233 to test whether a variable selector is valid. 36234 362352018-01-02 Richard Sandiford <richard.sandiford@linaro.org> 36236 36237 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *. 36238 * optabs-query.c (can_vec_perm_p): Likewise. 36239 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices & 36240 instead of vec_perm_indices. 36241 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise, 36242 (vect_gen_perm_mask_checked): Likewise, 36243 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise, 36244 (vect_gen_perm_mask_checked): Likewise, 36245 362462018-01-02 Richard Sandiford <richard.sandiford@linaro.org> 36247 36248 * optabs-query.h (qimode_for_vec_perm): Declare. 36249 * optabs-query.c (can_vec_perm_p): Split out qimode search to... 36250 (qimode_for_vec_perm): ...this new function. 36251 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm. 36252 362532018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com> 36254 36255 * rtlanal.c (canonicalize_condition): Return 0 if final rtx 36256 does not have a conditional at the top. 36257 362582018-01-02 Richard Biener <rguenther@suse.de> 36259 36260 * ipa-inline.c (big_speedup_p): Fix expression. 36261 362622018-01-02 Jan Hubicka <hubicka@ucw.cz> 36263 36264 PR target/81616 36265 * config/i386/x86-tune-costs.h: Increase cost of integer load costs 36266 for generic 4->6. 36267 362682018-01-02 Jan Hubicka <hubicka@ucw.cz> 36269 36270 PR target/81616 36271 Generic tuning. 36272 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17, 36273 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14 36274 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase 36275 cond_taken_branch_cost 3->4. 36276 362772018-01-01 Jakub Jelinek <jakub@redhat.com> 36278 36279 PR tree-optimization/83581 36280 * tree-loop-distribution.c (pass_loop_distribution::execute): Return 36281 TODO_cleanup_cfg if any changes have been made. 36282 36283 PR middle-end/83608 36284 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of 36285 convert_modes if target mode has the right side, but different mode 36286 class. 36287 36288 PR middle-end/83609 36289 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg 36290 last argument when extracting from CONCAT. If either from_real or 36291 from_imag is NULL, use expansion through memory. If result is not 36292 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg 36293 the parts directly to inner mode, if even that fails, use expansion 36294 through memory. 36295 36296 PR middle-end/83623 36297 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT, 36298 check for bswap in mode rather than HImode and use that in expand_unop 36299 too. 36300 36301Copyright (C) 2018 Free Software Foundation, Inc. 36302 36303Copying and distribution of this file, with or without modification, 36304are permitted in any medium without royalty provided the copyright 36305notice and this notice are preserved. 36306