gcc/ChangeLog

   1 2014-03-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
   2
   3         * doc/generic.texi: Correct typos.
   4
   5 2014-03-24  Tobias Burnus  <burnus@net-b.de>
   6
   7         * doc/invoke.texi (-flto): Expand section about
   8         using static libraries with LTO.
   9
  10 2014-03-24  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
  11
  12         PR rtl-optimization/60501
  13         * optabs.def (addptr3_optab): New optab.
  14         * optabs.c (gen_addptr3_insn, have_addptr3_insn): New function.
  15         * doc/md.texi ("addptrm3"): Document new RTL standard expander.
  16         * expr.h (gen_addptr3_insn, have_addptr3_insn): Add prototypes.
  17
  18         * lra.c (emit_add3_insn): Use the addptr pattern if available.
  19
  20         * config/s390/s390.md ("addptrdi3", "addptrsi3"): New expanders.
  21
  22 2014-03-24  Ulrich Drepper  <drepper@gmail.com>
  23
  24         * config/i386/avx512fintrin.h: Define _mm512_set1_ps and
  25         _mm512_set1_pd.
  26
  27         * config/i386/avxintrin.h (_mm256_undefined_si256): Define.
  28         (_mm256_undefined_ps): Define.
  29         (_mm256_undefined_pd): Define.
  30         * config/i386/emmintrin.h (_mm_undefined_si128): Define.
  31         (_mm_undefined_pd): Define.
  32         * config/i386/xmmintrin.h (_mm_undefined_ps): Define.
  33         * config/i386/avx512fintrin.h (_mm512_undefined_si512): Define.
  34         (_mm512_undefined_ps): Define.
  35         (_mm512_undefined_pd): Define.
  36         Use _mm*_undefined_*.
  37         * config/i386/avx2intrin.h: Use _mm*_undefined_*.
  38
  39 2014-03-24  Alex Velenko  <Alex.Velenko@arm.com>
  40
  41         * config/aarch64/aarch64-simd-builtins.def (lshr): DI mode excluded.
  42         (lshr_simd): DI mode added.
  43         * config/aarch64/aarch64-simd.md (aarch64_lshr_simddi): New pattern.
  44         (aarch64_ushr_simddi): Likewise.
  45         * config/aarch64/aarch64.md (UNSPEC_USHR64): New unspec.
  46         * config/aarch64/arm_neon.h (vshr_n_u64): Intrinsic fixed.
  47         (vshrd_n_u64): Likewise.
  48
  49 2014-03-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
  50
  51         * Makefile.in (s-macro_list): Depend on cc1.
  52
  53 2014-03-23  Teresa Johnson  <tejohnson@google.com>
  54
  55         * ipa-utils.c (ipa_print_order): Use specified dump file.
  56
  57 2014-03-23  Eric Botcazou  <ebotcazou@adacore.com>
  58
  59         PR rtl-optimization/60601
  60         * bb-reorder.c (fix_up_fall_thru_edges): Test EDGE_FALLTHRU everywhere.
  61
  62         * gcc.c (eval_spec_function): Initialize save_growing_value.
  63
  64 2014-03-22  Jakub Jelinek  <jakub@redhat.com>
  65
  66         PR sanitizer/60613
  67         * internal-fn.c (ubsan_expand_si_overflow_addsub_check): For
  68         code == MINUS_EXPR, never swap op0 with op1.
  69
  70         * toplev.c (init_local_tick): Avoid signed integer multiplication
  71         overflow.
  72         * genautomata.c (reserv_sets_hash_value): Fix rotate idiom, avoid
  73         shift by first operand's bitsize.
  74
  75 2014-03-21  Jakub Jelinek  <jakub@redhat.com>
  76
  77         PR target/60610
  78         * config/i386/i386.h (TARGET_64BIT_P): If not TARGET_BI_ARCH,
  79         redefine to 1 or 0.
  80         * config/i386/darwin.h (TARGET_64BIT_P): Redefine to
  81         TARGET_ISA_64BIT_P(x).
  82
  83 2014-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
  84
  85         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Generate a
  86         pattern for vector nor instead of subtract from splat(-1).
  87         (altivec_expand_vec_perm_const_le): Likewise.
  88
  89 2014-03-21  Richard Henderson  <rth@twiddle.net>
  90
  91         PR target/60598
  92         * ifcvt.c (dead_or_predicable): Return FALSE if there are any frame
  93         related insns after epilogue_completed.
  94
  95 2014-03-21  Martin Jambor  <mjambor@suse.cz>
  96
  97         PR ipa/59176
  98         * cgraph.h (symtab_node): New flag body_removed.
  99         * ipa.c (symtab_remove_unreachable_nodes): Set body_removed flag
 100         when removing bodies.
 101         * symtab.c (dump_symtab_base): Dump body_removed flag.
 102         * cgraph.c (verify_edge_corresponds_to_fndecl): Skip nodes which
 103         had their bodies removed.
 104
 105 2014-03-21  Martin Jambor  <mjambor@suse.cz>
 106
 107         PR ipa/60419
 108         * ipa.c (symtab_remove_unreachable_nodes): Clear thunk flag of nodes
 109         in the border.
 110
 111 2014-03-21  Richard Biener  <rguenther@suse.de>
 112
 113         PR tree-optimization/60577
 114         * tree-core.h (struct tree_base): Document nothrow_flag use
 115         in VAR_DECL_NONALIASED.
 116         * tree.h (VAR_DECL_NONALIASED): New.
 117         (may_be_aliased): Adjust.
 118         * coverage.c (build_var): Set VAR_DECL_NONALIASED.
 119
 120 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
 121
 122         * expr.c (expand_expr_real_1): Remove outdated comment.
 123
 124 2014-03-20  Jakub Jelinek  <jakub@redhat.com>
 125
 126         PR middle-end/60597
 127         * ira.c (adjust_cleared_regs): Call copy_rtx on
 128         *reg_equiv[REGNO (loc)].src_p before passing it to
 129         simplify_replace_fn_rtx.
 130
 131         PR target/60568
 132         * config/i386/i386.c (x86_output_mi_thunk): Surround UNSPEC_GOT
 133         into CONST, put pic register as first operand of PLUS.  Use
 134         gen_const_mem for both 32-bit and 64-bit PIC got loads.
 135
 136 2014-03-20  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
 137
 138         * config/aarch64/aarch64.c (MEMORY_MOVE_COST): Delete.
 139
 140 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
 141
 142         * config/sparc/sparc.c (sparc_do_work_around_errata): Implement work
 143         around for store forwarding issue in the FPU on the UT699.
 144         * config/sparc/sparc.md (in_branch_delay): Return false for single FP
 145         loads and operations if -mfix-ut699 is specified.
 146         (divtf3_hq): Tweak attribute.
 147         (sqrttf2_hq): Likewise.
 148
 149 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
 150
 151         * calls.c (store_one_arg): Remove incorrect const qualification on the
 152         type of the temporary.
 153         * cfgexpand.c (expand_return): Likewise.
 154         * expr.c (expand_constructor): Likewise.
 155         (expand_expr_real_1): Likewise.
 156
 157 2014-03-20  Zhenqiang Chen  <zhenqiang.chen@linaro.org>
 158
 159         * config/arm/arm.c (arm_dwarf_register_span): Update the element number
 160         of parts.
 161
 162 2014-03-19  Kaz Kojima  <kkojima@gcc.gnu.org>
 163
 164         PR target/60039
 165         * config/sh/sh.md (udivsi3_i1): Clobber R1 register.
 166
 167 2014-03-19  James Greenhalgh  <james.greenhalgh@arm.com>
 168
 169         * config/arm/aarch-common-protos.h
 170         (alu_cost_table): Fix spelling of "extend".
 171         * config/arm/arm.c (arm_new_rtx_costs): Fix spelling of "extend".
 172
 173 2014-03-19  Richard Biener  <rguenther@suse.de>
 174
 175         PR middle-end/60553
 176         * tree-core.h (tree_type_common): Re-order pointer members
 177         to reduce recursion depth during GC walks.
 178
 179 2014-03-19  Marek Polacek  <polacek@redhat.com>
 180
 181         PR sanitizer/60569
 182         * ubsan.c (ubsan_type_descriptor): Check that DECL_NAME is nonnull
 183         before accessing it.
 184
 185 2014-03-19  Richard Biener  <rguenther@suse.de>
 186
 187         PR lto/59543
 188         * lto-streamer-in.c (input_function): In WPA stage do not drop
 189         debug stmts.
 190
 191 2014-03-19  Jakub Jelinek  <jakub@redhat.com>
 192
 193         PR tree-optimization/60559
 194         * vectorizable_mask_load_store): Replace scalar MASK_LOAD
 195         with build_zero_cst assignment.
 196
 197 2014-03-18  Kai Tietz  <ktietz@redhat.com>
 198
 199         PR rtl-optimization/56356
 200         * sdbout.c (sdbout_parms): Verify that parms'
 201         incoming argument is valid.
 202         (sdbout_reg_parms): Likewise.
 203
 204 2014-03-18  Richard Henderson  <rth@redhat.com>
 205
 206         PR target/60562
 207         * config/i386/i386.md (*float<SWI48x><MODEF>2_i387): Move down to
 208         be shadowed by *float<SWI48><MODEF>2_sse.  Test X87_ENABLE_FLOAT.
 209         (*float<SWI48><MODEF>2_sse): Check X87_ENABLE_FLOAT for alternative 0.
 210
 211 2014-03-18  Basile Starynkevitch  <basile@starynkevitch.net>
 212
 213         * plugin.def: Improve comment for PLUGIN_INCLUDE_FILE.
 214         * doc/plugins.texi (Plugin callbacks): Mention
 215         PLUGIN_INCLUDE_FILE.
 216         Italicize plugin event names in description.  Explain that
 217         PLUGIN_PRAGMAS has no sense for lto1. Explain
 218         PLUGIN_INCLUDE_FILE.
 219         Remind that no GCC functions should be called after
 220         PLUGIN_FINISH.
 221         Explain what pragmas with expansion are.
 222
 223 2014-03-18  Martin Liska  <mliska@suse.cz>
 224
 225         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Added case when
 226         gimple call statement is update.
 227         * gimple-fold.c (gimple_fold_call): Changed order for GIMPLE_ASSIGN and
 228         GIMPLE_CALL, where gsi iterator still points to GIMPLE CALL.
 229
 230 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
 231
 232         PR sanitizer/60557
 233         * ubsan.c (ubsan_instrument_unreachable): Call
 234         initialize_sanitizer_builtins.
 235         (ubsan_pass): Likewise.
 236
 237         PR sanitizer/60535
 238         * ubsan.c (ubsan_type_descriptor, ubsan_create_data): Call
 239         varpool_finalize_decl instead of rest_of_decl_compilation.
 240
 241 2014-03-18  Richard Biener  <rguenther@suse.de>
 242
 243         * df-problems.c (df_rd_confluence_n): Avoid bitmap_copy
 244         by using bitmap_and_compl instead of bitmap_and_compl_into.
 245         (df_rd_transfer_function): Likewise.
 246
 247 2014-03-18  Richard Biener  <rguenther@suse.de>
 248
 249         * doc/lto.texi (fresolution): Fix typo.
 250
 251 2014-03-18  Richard Biener  <rguenther@suse.de>
 252
 253         * doc/invoke.texi (flto): Update for changes in 4.9.
 254
 255 2014-03-18  Richard Biener  <rguenther@suse.de>
 256
 257         * doc/loop.texi: Remove section on the removed lambda framework.
 258         Update loop docs with recent changes in preserving loop structure.
 259
 260 2014-03-18  Richard Biener  <rguenther@suse.de>
 261
 262         * doc/lto.texi (-fresolution): Document.
 263
 264 2014-03-18  Richard Biener  <rguenther@suse.de>
 265
 266         * doc/contrib.texi: Adjust my name.
 267
 268 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
 269
 270         PR ipa/58721
 271         * internal-fn.c: Include diagnostic-core.h.
 272         (expand_BUILTIN_EXPECT): New function.
 273         * gimplify.c (gimplify_call_expr): Use false instead of FALSE.
 274         (gimplify_modify_expr): Gimplify 3 argument __builtin_expect into
 275         IFN_BUILTIN_EXPECT call instead of __builtin_expect builtin call.
 276         * ipa-inline-analysis.c (find_foldable_builtin_expect): Handle
 277         IFN_BUILTIN_EXPECT.
 278         * predict.c (expr_expected_value_1): Handle IFN_BUILTIN_EXPECT.
 279         Revert 3 argument __builtin_expect code.
 280         (strip_predict_hints): Handle IFN_BUILTIN_EXPECT.
 281         * gimple-fold.c (gimple_fold_call): Likewise.
 282         * tree.h (fold_builtin_expect): New prototype.
 283         * builtins.c (build_builtin_expect_predicate): Add predictor
 284         argument, if non-NULL, create 3 argument __builtin_expect.
 285         (fold_builtin_expect): No longer static.  Add ARG2 argument,
 286         pass it through to build_builtin_expect_predicate.
 287         (fold_builtin_2): Adjust caller.
 288         (fold_builtin_3): Handle BUILT_IN_EXPECT.
 289         * internal-fn.def (BUILTIN_EXPECT): New.
 290
 291 2014-03-18  Tobias Burnus  <burnus@net-b.de>
 292
 293         PR ipa/58721
 294         * predict.def (PRED_FORTRAN_OVERFLOW, PRED_FORTRAN_FAIL_ALLOC,
 295         PRED_FORTRAN_FAIL_IO, PRED_FORTRAN_WARN_ONCE, PRED_FORTRAN_SIZE_ZERO,
 296         PRED_FORTRAN_INVALID_BOUND, PRED_FORTRAN_ABSENT_DUMMY): Add.
 297
 298 2014-03-18  Jan Hubicka  <hubicka@ucw.cz>
 299
 300         PR ipa/58721
 301         * predict.c (combine_predictions_for_bb): Fix up formatting.
 302         (expr_expected_value_1, expr_expected_value): Add predictor argument,
 303         fill what it points to if non-NULL.
 304         (tree_predict_by_opcode): Adjust caller, use the predictor.
 305         * predict.def (PRED_COMPARE_AND_SWAP): Add.
 306
 307 2014-03-18  Eric Botcazou  <ebotcazou@adacore.com>
 308
 309         * config/sparc/sparc.c (sparc_do_work_around_errata): Speed up and use
 310         proper constant for the store mode.
 311
 312 2014-03-18  Ilya Enkovich  <ilya.enkovich@intel.com>
 313
 314         * symtab.c (change_decl_assembler_name): Fix transparent alias
 315         chain construction.
 316
 317 2014-03-16  Renlin Li  <Renlin.Li@arm.com>
 318
 319         * config/aarch64/aarch64.c: Correct the comments about the
 320         aarch64 stack layout.
 321
 322 2014-03-18  Thomas Schwinge  <thomas@codesourcery.com>
 323
 324         * omp-low.c (lower_rec_input_clauses) <build_omp_barrier>: Restore
 325         check for GF_OMP_FOR_KIND_FOR.
 326
 327 2013-03-18  Kirill Yukhin  <kirill.yukhin@intel.com>
 328
 329         * config/i386/i386.h (ADDITIONAL_REGISTER_NAMES): Add
 330         ymm and zmm register names.
 331
 332 2014-03-17  Jakub Jelinek  <jakub@redhat.com>
 333
 334         PR target/60516
 335         * config/i386/i386.c (ix86_expand_epilogue): Adjust REG_CFA_ADJUST_CFA
 336         note creation for the 2010-08-31 changes.
 337
 338 2014-03-17  Marek Polacek  <polacek@redhat.com>
 339
 340         PR middle-end/60534
 341         * omp-low.c (omp_max_vf): Treat -fno-tree-loop-optimize the same
 342         as -fno-tree-loop-vectorize.
 343         (expand_omp_simd): Likewise.
 344
 345 2014-03-15  Eric Botcazou  <ebotcazou@adacore.com>
 346
 347         * config/sparc/sparc-protos.h (tls_call_delay): Delete.
 348         (eligible_for_call_delay): New prototype.
 349         * config/sparc/sparc.c (tls_call_delay): Rename into...
 350         (eligible_for_call_delay): ...this.  Return false if the instruction
 351         cannot be put in the delay slot of a branch.
 352         (eligible_for_restore_insn): Simplify.
 353         (eligible_for_return_delay): Return false if the instruction cannot be
 354         put in the delay slot of a branch and simplify.
 355         (eligible_for_sibcall_delay): Return false if the instruction cannot be
 356         put in the delay slot of a branch.
 357         * config/sparc/sparc.md (fix_ut699): New attribute.
 358         (tls_call_delay): Delete.
 359         (in_call_delay): Reimplement.
 360         (eligible_for_sibcall_delay): Rename into...
 361         (in_sibcall_delay): ...this.
 362         (eligible_for_return_delay): Rename into...
 363         (in_return_delay): ...this.
 364         (in_branch_delay): Reimplement.
 365         (in_uncond_branch_delay): Delete.
 366         (in_annul_branch_delay): Delete.
 367
 368 2014-03-14  Richard Henderson  <rth@redhat.com>
 369
 370         PR target/60525
 371         * config/i386/i386.md (floathi<X87MODEF>2): Delete expander; rename
 372         define_insn from *floathi<X87MODEF>2_i387; allow nonimmediate_operand.
 373         (*floathi<X87MODEF>2_i387_with_temp): Remove.
 374         (floathi splitters): Remove.
 375         (float<SWI48x>xf2): New pattern.
 376         (float<SWI48><MODEF>2): Rename from float<SWI48x><X87MODEF>2.  Drop
 377         code that tried to handle DImode for 32-bit, but which was excluded
 378         by the pattern's condition.  Drop allocation of stack temporary.
 379         (*floatsi<MODEF>2_vector_mixed_with_temp): Remove.
 380         (*float<SWI48><MODEF>2_mixed_with_temp): Remove.
 381         (*float<SWI48><MODEF>2_mixed_interunit): Remove.
 382         (*float<SWI48><MODEF>2_mixed_nointerunit): Remove.
 383         (*floatsi<MODEF>2_vector_sse_with_temp): Remove.
 384         (*float<SWI48><MODEF>2_sse_with_temp): Remove.
 385         (*float<SWI48><MODEF>2_sse_interunit): Remove.
 386         (*float<SWI48><MODEF>2_sse_nointerunit): Remove.
 387         (*float<SWI48x><X87MODEF>2_i387_with_temp): Remove.
 388         (*float<SWI48x><X87MODEF>2_i387): Remove.
 389         (all float _with_temp splitters): Remove.
 390         (*float<SWI48x><MODEF>2_i387): New pattern.
 391         (*float<SWI48><MODEF>2_sse): New pattern.
 392         (float TARGET_USE_VECTOR_CONVERTS splitters): Merge them.
 393         (float TARGET_SSE_PARTIAL_REG_DEPENDENCY splitters): Merge them.
 394
 395 2014-03-14  Jakub Jelinek  <jakub@redhat.com>
 396             Marek Polacek  <polacek@redhat.com>
 397
 398         PR middle-end/60484
 399         * common.opt (dump_base_name_prefixed): New Variable.
 400         * opts.c (finish_options): Don't prepend directory to x_dump_base_name
 401         if x_dump_base_name_prefixed is already set, set it at the end.
 402
 403 2014-03-14  Vladimir Makarov  <vmakarov@redhat.com>
 404
 405         PR rtl-optimization/60508
 406         * lra-constraints.c (get_reload_reg): Add new parameter
 407         in_subreg_p.
 408         (process_addr_reg, simplify_operand_subreg, curr_insn_transform):
 409         Pass the new parameter values.
 410
 411 2014-03-14  Richard Biener  <rguenther@suse.de>
 412
 413         * common.opt: Revert unintented changes from r205065.
 414         * opts.c: Likewise.
 415
 416 2014-03-14  Richard Biener  <rguenther@suse.de>
 417
 418         PR middle-end/60518
 419         * cfghooks.c (split_block): Properly adjust all loops the
 420         block was a latch of.
 421
 422 2014-03-14  Martin Jambor  <mjambor@suse.cz>
 423
 424         PR lto/60461
 425         * ipa-prop.c (ipa_modify_call_arguments): Fix iteration condition
 426         and simplify it.
 427
 428 2014-03-14  Georg-Johann Lay  <avr@gjlay.de>
 429
 430         PR target/59396
 431         * config/avr/avr.c (avr_set_current_function): Pass function name
 432         through default_strip_name_encoding before sanity checking instead
 433         of skipping the first char of the assembler name.
 434
 435 2014-03-13  Richard Henderson  <rth@redhat.com>
 436
 437         PR debug/60438
 438         * config/i386/i386.c (ix86_split_fp_branch): Remove pushed argument.
 439         (ix86_force_to_memory, ix86_free_from_memory): Remove.
 440         * config/i386/i386-protos.h: Likewise.
 441         * config/i386/i386.md (floathi<X87MODEF>2): Use assign_386_stack_local
 442         in the expander instead of a splitter.
 443         (float<SWI48x><X87MODEF>2): Use assign_386_stack_local if there is
 444         any possibility of requiring a memory.
 445         (*floatsi<MODEF>2_vector_mixed): Remove, and the splitters.
 446         (*floatsi<MODEF>2_vector_sse): Remove, and the splitters.
 447         (fp branch splitters): Update for ix86_split_fp_branch.
 448         (*jcc<X87MODEF>_<SWI24>_i387): Remove r/f alternative.
 449         (*jcc<X87MODEF>_<SWI24>_r_i387): Likewise.
 450         (splitter for jcc<X87MODEF>_<SWI24>_i387 r/f): Remove.
 451         (*fop_<MODEF>_2_i387): Remove f/r alternative.
 452         (*fop_<MODEF>_3_i387): Likewise.
 453         (*fop_xf_2_i387, *fop_xf_3_i387): Likewise.
 454         (splitters for the fop_* register patterns): Remove.
 455         (fscalexf4_i387): Rename from *fscalexf4_i387.
 456         (ldexpxf3): Use gen_floatsixf2 and gen_fscalexf4_i387.
 457
 458 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
 459
 460         PR tree-optimization/59779
 461         * tree-dfa.c (get_ref_base_and_extent): Use double_int
 462         type for bitsize and maxsize instead of HOST_WIDE_INT.
 463
 464 2014-03-13  Steven Bosscher  <steven@gcc.gnu.org>
 465
 466         PR rtl-optimization/57320
 467         * function.c (rest_of_handle_thread_prologue_and_epilogue): Cleanup
 468         the CFG after thread_prologue_and_epilogue_insns.
 469
 470 2014-03-13  Vladimir Makarov  <vmakarov@redhat.com>
 471
 472         PR rtl-optimization/57189
 473         * lra-constraints.c (process_alt_operands): Disfavor spilling
 474         vector pseudos.
 475
 476 2014-03-13  Cesar Philippidis  <cesar@codesourcery.com>
 477
 478         * lto-wrapper.c (maybe_unlink_file): Suppress diagnostic
 479         messages.
 480
 481 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
 482
 483         PR tree-optimization/59025
 484         PR middle-end/60418
 485         * tree-ssa-reassoc.c (sort_by_operand_rank): For SSA_NAMEs with the
 486         same rank, sort by bb_rank and gimple_uid of SSA_NAME_DEF_STMT first.
 487
 488 2014-03-13  Georg-Johann Lay  <avr@gjlay.de>
 489
 490         PR target/60486
 491         * config/avr/avr.c (avr_out_plus): Swap cc_plus and cc_minus in
 492         calls of avr_out_plus_1.
 493
 494 2014-03-13  Bin Cheng  <bin.cheng@arm.com>
 495
 496         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Record
 497         BB's single pred and update the father loop's latch info later.
 498
 499 2014-03-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
 500
 501         * config/rs6000/vector.md (VEC_L): Add V1TI mode to vector types.
 502         (VEC_M): Likewise.
 503         (VEC_N): Likewise.
 504         (VEC_R): Likewise.
 505         (VEC_base): Likewise.
 506         (mov<MODE>, VEC_M modes): If we are loading TImode into VSX
 507         registers, we need to swap double words in little endian mode.
 508
 509         * config/rs6000/rs6000-modes.def (V1TImode): Add new vector mode
 510         to be a container mode for 128-bit integer operations added in ISA
 511         2.07.  Unlike TImode and PTImode, the preferred register set is
 512         the Altivec/VMX registers for the 128-bit operations.
 513
 514         * config/rs6000/rs6000-protos.h (rs6000_move_128bit_ok_p): Add
 515         declarations.
 516         (rs6000_split_128bit_ok_p): Likewise.
 517
 518         * config/rs6000/rs6000-builtin.def (BU_P8V_AV_3): Add new support
 519         macros for creating ISA 2.07 normal and overloaded builtin
 520         functions with 3 arguments.
 521         (BU_P8V_OVERLOAD_3): Likewise.
 522         (VPERM_1T): Add support for V1TImode in 128-bit vector operations
 523         for use as overloaded functions.
 524         (VPERM_1TI_UNS): Likewise.
 525         (VSEL_1TI): Likewise.
 526         (VSEL_1TI_UNS): Likewise.
 527         (ST_INTERNAL_1ti): Likewise.
 528         (LD_INTERNAL_1ti): Likewise.
 529         (XXSEL_1TI): Likewise.
 530         (XXSEL_1TI_UNS): Likewise.
 531         (VPERM_1TI): Likewise.
 532         (VPERM_1TI_UNS): Likewise.
 533         (XXPERMDI_1TI): Likewise.
 534         (SET_1TI): Likewise.
 535         (LXVD2X_V1TI): Likewise.
 536         (STXVD2X_V1TI): Likewise.
 537         (VEC_INIT_V1TI): Likewise.
 538         (VEC_SET_V1TI): Likewise.
 539         (VEC_EXT_V1TI): Likewise.
 540         (EQV_V1TI): Likewise.
 541         (NAND_V1TI): Likewise.
 542         (ORC_V1TI): Likewise.
 543         (VADDCUQ): Add support for 128-bit integer arithmetic instructions
 544         added in ISA 2.07.  Add both normal 'altivec' builtins, and the
 545         overloaded builtin.
 546         (VADDUQM): Likewise.
 547         (VSUBCUQ): Likewise.
 548         (VADDEUQM): Likewise.
 549         (VADDECUQ): Likewise.
 550         (VSUBEUQM): Likewise.
 551         (VSUBECUQ): Likewise.
 552
 553         * config/rs6000/rs6000-c.c (__int128_type): New static to hold
 554         __int128_t and __uint128_t types.
 555         (__uint128_type): Likewise.
 556         (altivec_categorize_keyword): Add support for vector __int128_t,
 557         vector __uint128_t, vector __int128, and vector unsigned __int128
 558         as a container type for TImode operations that need to be done in
 559         VSX/Altivec registers.
 560         (rs6000_macro_to_expand): Likewise.
 561         (altivec_overloaded_builtins): Add ISA 2.07 overloaded functions
 562         to support 128-bit integer instructions vaddcuq, vadduqm,
 563         vaddecuq, vaddeuqm, vsubcuq, vsubuqm, vsubecuq, vsubeuqm.
 564         (altivec_resolve_overloaded_builtin): Add support for V1TImode.
 565
 566         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Add support
 567         for V1TImode, and set up preferences to use VSX/Altivec registers.
 568         Setup VSX reload handlers.
 569         (rs6000_debug_reg_global): Likewise.
 570         (rs6000_init_hard_regno_mode_ok): Likewise.
 571         (rs6000_preferred_simd_mode): Likewise.
 572         (vspltis_constant): Do not allow V1TImode as easy altivec constants.
 573         (easy_altivec_constant): Likewise.
 574         (output_vec_const_move): Likewise.
 575         (rs6000_expand_vector_set): Convert V1TImode set and extract to
 576         simple move.
 577         (rs6000_expand_vector_extract): Likewise.
 578         (reg_offset_addressing_ok_p): Setup V1TImode to use VSX reg+reg
 579         addressing.
 580         (rs6000_const_vec): Add support for V1TImode.
 581         (rs6000_emit_le_vsx_load): Swap double words when loading or
 582         storing TImode/V1TImode.
 583         (rs6000_emit_le_vsx_store): Likewise.
 584         (rs6000_emit_le_vsx_move): Likewise.
 585         (rs6000_emit_move): Add support for V1TImode.
 586         (altivec_expand_ld_builtin): Likewise.
 587         (altivec_expand_st_builtin): Likewise.
 588         (altivec_expand_vec_init_builtin): Likewise.
 589         (altivec_expand_builtin): Likewise.
 590         (rs6000_init_builtins): Add support for V1TImode type.  Add
 591         support for ISA 2.07 128-bit integer builtins.  Define type names
 592         for the VSX/Altivec vector types.
 593         (altivec_init_builtins): Add support for overloaded vector
 594         functions with V1TImode type.
 595         (rs6000_preferred_reload_class): Prefer Altivec registers for V1TImode.
 596         (rs6000_move_128bit_ok_p): Move 128-bit move/split validation to
 597         external function.
 598         (rs6000_split_128bit_ok_p): Likewise.
 599         (rs6000_handle_altivec_attribute): Create V1TImode from vector
 600         __int128_t and vector __uint128_t.
 601
 602         * config/rs6000/vsx.md (VSX_L): Add V1TImode to vector iterators
 603         and mode attributes.
 604         (VSX_M): Likewise.
 605         (VSX_M2): Likewise.
 606         (VSm): Likewise.
 607         (VSs): Likewise.
 608         (VSr): Likewise.
 609         (VSv): Likewise.
 610         (VS_scalar): Likewise.
 611         (VS_double): Likewise.
 612         (vsx_set_v1ti): New builtin function to create V1TImode from TImode.
 613
 614         * config/rs6000/rs6000.h (TARGET_VADDUQM): New macro to say whether
 615         we support the ISA 2.07 128-bit integer arithmetic instructions.
 616         (ALTIVEC_OR_VSX_VECTOR_MODE): Add V1TImode.
 617         (enum rs6000_builtin_type_index): Add fields to hold V1TImode
 618         and TImode types for use with the builtin functions.
 619         (V1TI_type_node): Likewise.
 620         (unsigned_V1TI_type_node): Likewise.
 621         (intTI_type_internal_node): Likewise.
 622         (uintTI_type_internal_node): Likewise.
 623
 624         * config/rs6000/altivec.md (UNSPEC_VADDCUQ): New unspecs for ISA 2.07
 625         128-bit builtin functions.
 626         (UNSPEC_VADDEUQM): Likewise.
 627         (UNSPEC_VADDECUQ): Likewise.
 628         (UNSPEC_VSUBCUQ): Likewise.
 629         (UNSPEC_VSUBEUQM): Likewise.
 630         (UNSPEC_VSUBECUQ): Likewise.
 631         (VM): Add V1TImode to vector mode iterators.
 632         (VM2): Likewise.
 633         (VI_unit): Likewise.
 634         (altivec_vadduqm): Add ISA 2.07 128-bit binary builtins.
 635         (altivec_vaddcuq): Likewise.
 636         (altivec_vsubuqm): Likewise.
 637         (altivec_vsubcuq): Likewise.
 638         (altivec_vaddeuqm): Likewise.
 639         (altivec_vaddecuq): Likewise.
 640         (altivec_vsubeuqm): Likewise.
 641         (altivec_vsubecuq): Likewise.
 642
 643         * config/rs6000/rs6000.md (FMOVE128_GPR): Add V1TImode to vector
 644         mode iterators.
 645         (BOOL_128): Likewise.
 646         (BOOL_REGS_OUTPUT): Likewise.
 647         (BOOL_REGS_OP1): Likewise.
 648         (BOOL_REGS_OP2): Likewise.
 649         (BOOL_REGS_UNARY): Likewise.
 650         (BOOL_REGS_AND_CR0): Likewise.
 651
 652         * config/rs6000/altivec.h (vec_vaddcuq): Add support for ISA 2.07
 653         128-bit integer builtin support.
 654         (vec_vadduqm): Likewise.
 655         (vec_vaddecuq): Likewise.
 656         (vec_vaddeuqm): Likewise.
 657         (vec_vsubecuq): Likewise.
 658         (vec_vsubeuqm): Likewise.
 659         (vec_vsubcuq): Likewise.
 660         (vec_vsubuqm): Likewise.
 661
 662         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
 663         Document vec_vaddcuq, vec_vadduqm, vec_vaddecuq, vec_vaddeuqm,
 664         vec_subecuq, vec_subeuqm, vec_vsubcuq, vec_vsubeqm builtins adding
 665         128-bit integer add/subtract to ISA 2.07.
 666
 667 2014-03-12  Joern Rennecke  <joern.rennecke@embecosm.com>
 668
 669         * config/arc/arc.c (arc_predicate_delay_insns):
 670         Fix third argument passed to conditionalize_nonjump.
 671
 672 2014-03-12  Yufeng Zhang  <yufeng.zhang@arm.com>
 673
 674         * config/aarch64/aarch64-builtins.c
 675         (aarch64_builtin_vectorized_function): Add BUILT_IN_LFLOORF,
 676         BUILT_IN_LLFLOOR, BUILT_IN_LCEILF and BUILT_IN_LLCEIL.
 677         * config/aarch64/arm_neon.h (vcvtaq_u64_f64): Call __builtin_llfloor
 678         instead of __builtin_lfloor.
 679         (vcvtnq_u64_f64): Call __builtin_llceil instead of __builtin_lceil.
 680
 681 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
 682
 683         * tree-ssa-ifcombine.c (forwarder_block_to): New function.
 684         (tree_ssa_ifcombine_bb_1): New function.
 685         (tree_ssa_ifcombine_bb): Use it.  Handle also cases where else_bb
 686         is an empty forwarder block to then_bb or vice versa and then_bb
 687         and else_bb are effectively swapped.
 688
 689 2014-03-12  Christian Bruel  <christian.bruel@st.com>
 690
 691         PR target/60264
 692         * config/arm/arm.c (arm_emit_vfp_multi_reg_pop): Emit a
 693         REG_CFA_DEF_CFA note.
 694         (arm_expand_epilogue_apcs_frame): call arm_add_cfa_adjust_cfa_note.
 695         (arm_unwind_emit): Allow REG_CFA_DEF_CFA.
 696
 697 2014-03-12  Thomas Preud'homme  <thomas.preudhomme@arm.com>
 698
 699         PR tree-optimization/60454
 700         * tree-ssa-math-opts.c (find_bswap_1): Fix bswap detection.
 701
 702 2014-03-12  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 703
 704         * config.gcc (aarch64*-*-*): Use ISA flags from aarch64-arches.def.
 705         Do not define target_cpu_default2 to generic.
 706         * config/aarch64/aarch64.h (TARGET_CPU_DEFAULT): Use generic cpu.
 707         * config/aarch64/aarch64.c (aarch64_override_options): Update comment.
 708         * config/aarch64/aarch64-arches.def (armv8-a): Use generic cpu.
 709
 710 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
 711             Marc Glisse  <marc.glisse@inria.fr>
 712
 713         PR tree-optimization/60502
 714         * tree-ssa-reassoc.c (eliminate_not_pairs): Use build_all_ones_cst
 715         instead of build_low_bits_mask.
 716
 717 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
 718
 719         PR middle-end/60482
 720         * tree-vrp.c (register_edge_assert_for_1): Don't add assert
 721         if there are multiple uses, but op doesn't live on E edge.
 722         * tree-cfg.c (assert_unreachable_fallthru_edge_p): Also ignore
 723         clobber stmts before __builtin_unreachable.
 724
 725 2014-03-11  Richard Sandiford  <rdsandiford@googlemail.com>
 726
 727         * builtins.c (expand_builtin_setjmp_receiver): Use and clobber
 728         hard_frame_pointer_rtx.
 729         * cse.c (cse_insn): Remove volatile check.
 730         * cselib.c (cselib_process_insn): Likewise.
 731         * dse.c (scan_insn): Likewise.
 732
 733 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
 734
 735         * config/arc/arc.c (conditionalize_nonjump): New function,
 736         broken out of ...
 737         (arc_ifcvt): ... this.
 738         (arc_predicate_delay_insns): Use it.
 739
 740 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
 741
 742         * config/arc/predicates.md (extend_operand): During/after reload,
 743         allow const_int_operand.
 744         * config/arc/arc.md (mulsidi3_700): Use extend_operand predicate.
 745         (umulsidi3_700): Likewise.  Change operand 2 constraint back to "cL".
 746         (mulsi3_highpart): Change operand 2 constraint alternatives 2 and 3
 747         to "i".
 748         (umulsi3_highpart_i): Likewise.
 749
 750 2014-03-11  Richard Biener  <rguenther@suse.de>
 751
 752         * tree-ssa-structalias.c (get_constraint_for_ptr_offset):
 753         Add asserts to guard possible wrong-code bugs.
 754
 755 2014-03-11  Richard Biener  <rguenther@suse.de>
 756
 757         PR tree-optimization/60429
 758         PR tree-optimization/60485
 759         * tree-ssa-structalias.c (set_union_with_increment): Properly
 760         take into account all fields that overlap the shifted vars.
 761         (do_sd_constraint): Likewise.
 762         (do_ds_constraint): Likewise.
 763         (get_constraint_for_ptr_offset): Likewise.
 764
 765 2014-03-11  Chung-Lin Tang  <cltang@codesourcery.com>
 766
 767         * config/nios2/nios2.c (machine_function): Add fp_save_offset field.
 768         (nios2_compute_frame_layout):
 769         Add calculation of cfun->machine->fp_save_offset.
 770         (nios2_expand_prologue): Correct setting of frame pointer register
 771         in prologue.
 772         (nios2_expand_epilogue): Update recovery of stack pointer from
 773         frame pointer accordingly.
 774         (nios2_initial_elimination_offset): Update calculation of offset
 775         for eliminating to HARD_FRAME_POINTER_REGNUM.
 776
 777 2014-03-10  Jakub Jelinek  <jakub@redhat.com>
 778
 779         PR ipa/60457
 780         * ipa.c (symtab_remove_unreachable_nodes): Don't call
 781         cgraph_get_create_node on VAR_DECLs.
 782
 783 2014-03-10  Richard Biener  <rguenther@suse.de>
 784
 785         PR middle-end/60474
 786         * tree.c (signed_or_unsigned_type_for): Handle OFFSET_TYPEs.
 787
 788 2014-03-08  Douglas B Rupp  <rupp@gnat.com>
 789
 790         * config/vms/vms.opt (vms_float_format): New variable.
 791
 792 2014-03-08  Tobias Burnus  <burnus@net-b.de>
 793
 794         * doc/invoke.texi (-fcilkplus): Update implementation status.
 795
 796 2014-03-08  Paulo Matos  <paulo@matos-sorge.com>
 797             Richard Biener  <rguenther@suse.de>
 798
 799         * lto-wrapper.c (merge_and_complain): Ensure -fshort-double is used
 800         consistently accross all TUs.
 801         (run_gcc): Enable -fshort-double automatically at link at link-time
 802         and disallow override.
 803
 804 2014-03-08  Richard Sandiford  <rdsandiford@googlemail.com>
 805
 806         PR target/58271
 807         * config/mips/mips.c (mips_option_override): Promote -mpaired-single
 808         warning to an error.  Disable TARGET_PAIRED_SINGLE and TARGET_MIPS3D
 809         if they can't be used.
 810
 811 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
 812
 813         * configure.ac (HAVE_AS_IX86_TLSLDMPLT): Improve test
 814         for Solaris 11/x86 ld.
 815         * configure: Regenerate.
 816
 817 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
 818
 819         * configure.ac (TLS_SECTION_ASM_FLAG): Save as tls_section_flag.
 820         (LIB_TLS_SPEC): Save as ld_tls_libs.
 821         (HAVE_AS_IX86_TLSLDMPLT): Define as 1/0.
 822         (HAVE_AS_IX86_TLSLDM): New test.
 823         * configure, config.in: Regenerate.
 824         * config/i386/i386.c (legitimize_tls_address): Fall back to
 825         TLS_MODEL_GLOBAL_DYNAMIC on 32-bit Solaris/x86 if tool chain
 826         cannot support TLS_MODEL_LOCAL_DYNAMIC.
 827         * config/i386/i386.md (*tls_local_dynamic_base_32_gnu): Use if
 828         instead of #ifdef in HAVE_AS_IX86_TLSLDMPLT test.
 829
 830 2014-03-07  Paulo Matos  <paulo@matos-sorge.com>
 831
 832         * common.opt (fira-loop-pressure): Mark as optimization.
 833
 834 2014-03-07  Thomas Schwinge  <thomas@codesourcery.com>
 835
 836         * langhooks.c (lhd_omp_mappable_type): The error_mark_node is not
 837         an OpenMP mappable type.
 838
 839 2014-03-06  Matthias Klose  <doko@ubuntu.com>
 840
 841         * Makefile.in (s-mlib): Only pass MULTIARCH_DIRNAME if
 842         MULTILIB_OSDIRNAMES is not defined.
 843
 844 2014-03-06  Jakub Jelinek  <jakub@redhat.com>
 845             Meador Inge  <meadori@codesourcery.com>
 846
 847         PR target/58595
 848         * config/arm/arm.c (arm_tls_symbol_p): Remove.
 849         (arm_legitimize_address): Call legitimize_tls_address for any
 850         arm_tls_referenced_p expression, handle constant addend.  Call it
 851         before testing for !TARGET_ARM.
 852         (thumb_legitimize_address): Don't handle arm_tls_symbol_p here.
 853
 854 2014-03-06  Richard Biener  <rguenther@suse.de>
 855
 856         PR middle-end/60445
 857         PR lto/60424
 858         PR lto/60427
 859         Revert
 860         2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
 861
 862         * tree-streamer.c (record_common_node): Assert we don't record
 863         nodes with type double.
 864         (preload_common_node): Skip type double, complex double and double
 865         pointer since it is now frontend dependent due to fshort-double option.
 866
 867 2014-03-06  Richard Biener  <rguenther@suse.de>
 868
 869         * gcc.c (PLUGIN_COND): Always enable unless -fno-use-linker-plugin
 870         or -fno-lto is specified and the linker has full plugin support.
 871         * collect2.c (lto_mode): Default to LTO_MODE_WHOPR if LTO is enabled.
 872         (main): Remove -flto processing, adjust lto_mode using use_plugin late.
 873         * lto-wrapper.c (merge_and_complain): Merge compile-time
 874         optimization levels.
 875         (run_gcc): And pass it through to the link options.
 876
 877 2014-03-06  Alexandre Oliva <aoliva@redhat.com>
 878
 879         PR debug/60381
 880         Revert:
 881         2014-02-28  Alexandre Oliva <aoliva@redhat.com>
 882         PR debug/59992
 883         * cselib.c (remove_useless_values): Skip to avoid quadratic
 884         behavior if the condition moved from...
 885         (cselib_process_insn): ... here holds.
 886
 887 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
 888
 889         PR plugins/59335
 890         * Makefile.in (PLUGIN_HEADERS): Add tree-phinodes.h, stor-layout.h,
 891         ssa-iterators.h, $(RESOURCE_H) and tree-cfgcleanup.h.
 892
 893         PR plugins/59335
 894         * config/i386/t-i386 (OPTIONS_H_EXTRA): Add stringop.def.
 895         (TM_H): Add x86-tune.def.
 896
 897 2014-03-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 898
 899         * config/aarch64/aarch64.c (generic_tunings):
 900         Use cortexa57_extra_costs.
 901
 902 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
 903
 904         PR lto/60404
 905         * cfgexpand.c (expand_used_vars): Do not assume all SSA_NAMEs
 906         of PARM/RESULT_DECLs must be coalesced with optimize && in_lto_p.
 907         * tree-ssa-coalesce.c (coalesce_ssa_name): Use MUST_COALESCE_COST - 1
 908         cost for in_lto_p.
 909
 910 2014-03-04  Heiher  <r@hev.cc>
 911
 912         * config/mips/mips-cpus.def (loongson3a): Mark as a MIPS64r2 processor.
 913         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Adjust accordingly.
 914
 915 2014-03-04  Uros Bizjak  <ubizjak@gmail.com>
 916
 917         * config/i386/predicates.md (const2356_operand): Change to ...
 918         (const2367_operand): ... this.
 919         * config/i386/sse.md (avx512pf_scatterpf<mode>sf): Use
 920         const2367_operand.
 921         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
 922         (*avx512pf_scatterpf<mode>sf): Ditto.
 923         (avx512pf_scatterpf<mode>df): Ditto.
 924         (*avx512pf_scatterpf<mode>df_mask): Ditto.
 925         (*avx512pf_scatterpf<mode>df): Ditto.
 926         * config/i386/i386.c (ix86_expand_builtin): Update
 927         incorrect hint operand error message.
 928
 929 2014-03-04  Richard Biener  <rguenther@suse.de>
 930
 931         * lto-section-in.c (lto_get_section_data): Fix const cast.
 932
 933 2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
 934
 935         * tree-streamer.c (record_common_node): Assert we don't record
 936         nodes with type double.
 937         (preload_common_node): Skip type double, complex double and double
 938         pointer since it is now frontend dependent due to fshort-double option.
 939
 940 2014-03-04  Richard Biener  <rguenther@suse.de>
 941
 942         PR lto/60405
 943         * lto-streamer-in.c (lto_read_body): Remove LTO bytecode version check.
 944         (lto_input_toplevel_asms): Likewise.
 945         * lto-section-in.c (lto_get_section_data): Instead do it here
 946         for every section.
 947
 948 2014-03-04  Richard Biener  <rguenther@suse.de>
 949
 950         PR tree-optimization/60382
 951         * tree-vect-loop.c (vect_is_simple_reduction_1): Do not consider
 952         dead PHIs a reduction.
 953
 954 2014-03-03  Uros Bizjak  <ubizjak@gmail.com>
 955
 956         * config/i386/xmmintrin.h (enum _mm_hint) <_MM_HINT_ET0>: Correct
 957         hint value.
 958         (_mm_prefetch): Move out of GCC target("sse") pragma.
 959         * config/i386/prfchwintrin.h (_m_prefetchw): Move out of
 960         GCC target("prfchw") pragma.
 961         * config/i386/i386.md (prefetch): Emit prefetchwt1 only
 962         for locality <= 2.
 963         * config/i386/i386.c (ix86_option_override_internal): Enable
 964         -mprfchw with -mprefetchwt1.
 965
 966 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
 967
 968         * config/arc/arc.md (casesi_load) <length attribute alternative 0>:
 969         Mark as varying.
 970
 971 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
 972
 973         * opts.h (CL_PCH_IGNORE): Define.
 974         * targhooks.c (option_affects_pch_p):
 975         Return false for options that have CL_PCH_IGNORE set.
 976         * opt-functions.awk: Process PchIgnore.
 977         * doc/options.texi: Document PchIgnore.
 978
 979         * config/arc/arc.opt (misize): Add PchIgnore property.
 980
 981 2014-03-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
 982
 983         * config/rs6000/rs6000.c (rs6000_preferred_reload_class): Disallow
 984         reload of PLUS rtx's outside of GENERAL_REGS or BASE_REGS; relax
 985         constraint on constants to permit them being loaded into
 986         GENERAL_REGS or BASE_REGS.
 987
 988 2014-03-03  Nick Clifton  <nickc@redhat.com>
 989
 990         * config/rl78/rl78-real.md (cbranchsi4_real_signed): Add
 991         anti-cacnonical alternatives.
 992         (negandhi3_real): New pattern.
 993         * config/rl78/rl78-virt.md (negandhi3_virt): New pattern.
 994
 995 2014-03-03  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
 996
 997         * config/avr/avr-mcus.def: Remove atxmega16x1.
 998         * config/avr/avr-tables.opt: Regenerate.
 999         * config/avr/t-multilib: Regenerate.
1000         * doc/avr-mmcu.texi: Regenerate.
1001
1002 2014-03-03  Tobias Grosser  <tobias@grosser.es>
1003             Mircea Namolaru  <mircea.namolaru@inria.fr>
1004
1005         PR tree-optimization/58028
1006         * graphite-clast-to-gimple.c (set_cloog_options): Don't remove
1007         scalar dimensions.
1008
1009 2014-03-03  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
1010
1011         * config/arm/neon.md (*movmisalign<mode>): Legitimize addresses
1012         not handled by recognizers.
1013
1014 2014-03-03  Jakub Jelinek  <jakub@redhat.com>
1015
1016         PR middle-end/60175
1017         * function.c (expand_function_end): Don't emit
1018         clobber_return_register sequence if clobber_after is a BARRIER.
1019         * cfgexpand.c (construct_exit_block): Append instructions before
1020         return_label to prev_bb.
1021
1022 2014-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1023
1024         * config/rs6000/constraints.md: Document reserved use of "wc".
1025
1026 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
1027
1028         PR ipa/60150
1029         * ipa.c (function_and_variable_visibility): When dissolving comdat
1030         group, also set all symbols to local.
1031
1032 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
1033
1034         PR ipa/60306
1035
1036         Revert:
1037         2013-12-14   Jan Hubicka  <jh@suse.cz>
1038         PR middle-end/58477
1039         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Skip clobbers.
1040
1041 2014-03-02  Jon Beniston  <jon@beniston.com>
1042
1043         PR bootstrap/48230
1044         PR bootstrap/50927
1045         PR bootstrap/52466
1046         PR target/46898
1047         * config/lm32/lm32.c (lm32_legitimate_constant_p): Remove, as incorrect.
1048           (TARGET_LEGITIMATE_CONSTANT_P): Undefine, as not needed.
1049         * config/lm32/lm32.md (movsi_insn): Add 32-bit immediate support.
1050         (simple_return, *simple_return): New patterns
1051         * config/lm32/predicates.md (movsi_rhs_operand): Remove as obsolete.
1052         * configure.ac (force_sjlj_exceptions): Force sjlj exceptions for lm32.
1053
1054 2014-03-01  Paolo Carlini  <paolo.carlini@oracle.com>
1055
1056         * dwarf2out.c (gen_subprogram_die): Tidy.
1057
1058 2014-03-01  Oleg Endo  <olegendo@gcc.gnu.org>
1059
1060         PR target/60071
1061         * config/sh/sh.md (*mov_t_msb_neg): Split into ...
1062         (*mov_t_msb_neg_negc): ... this new insn.
1063
1064 2014-02-28  Jason Merrill  <jason@redhat.com>
1065
1066         PR c++/58678
1067         * ipa-devirt.c (ipa_devirt): Don't choose an implicitly-declared
1068         function.
1069
1070 2014-02-28  Paolo Carlini  <paolo.carlini@oracle.com>
1071
1072         PR c++/60314
1073         * dwarf2out.c (decltype_auto_die): New static.
1074         (gen_subprogram_die): Handle 'decltype(auto)' like 'auto'.
1075         (gen_type_die_with_usage): Handle 'decltype(auto)'.
1076         (is_cxx_auto): Likewise.
1077
1078 2014-02-28  Ian Bolton  <ian.bolton@arm.com>
1079
1080         * config/aarch64/aarch64.h: Define __ARM_NEON by default if
1081         we are not using general regs only.
1082
1083 2014-02-28  Richard Biener  <rguenther@suse.de>
1084
1085         PR target/60280
1086         * tree-cfgcleanup.c (tree_forwarder_block_p): Restrict
1087         previous fix and only allow to remove trivial pre-headers
1088         and latches.  Also honor LOOPS_MAY_HAVE_MULTIPLE_LATCHES.
1089         (remove_forwarder_block): Properly update the latch of a loop.
1090
1091 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
1092
1093         PR debug/59992
1094         * cselib.c (cselib_hasher::equal): Special-case VALUE lookup.
1095         (cselib_preserved_hash_table): New.
1096         (preserve_constants_and_equivs): Move preserved vals to it.
1097         (cselib_find_slot): Look it up first.
1098         (cselib_init): Initialize it.
1099         (cselib_finish): Release it.
1100         (dump_cselib_table): Dump it.
1101
1102 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
1103
1104         PR debug/59992
1105         * cselib.c (remove_useless_values): Skip to avoid quadratic
1106         behavior if the condition moved from...
1107         (cselib_process_insn): ... here holds.
1108
1109 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
1110
1111         PR debug/57232
1112         * var-tracking.c (vt_initialize): Apply the same condition to
1113         preserve the CFA base value.
1114
1115 2014-02-28  Joey Ye  <joey.ye@arm.com>
1116
1117         PR target/PR60169
1118         * config/arm/arm.c (thumb_far_jump_used_p): Don't change
1119         if reload in progress or completed.
1120
1121 2014-02-28  Tobias Burnus  <burnus@net-b.de>
1122
1123         PR middle-end/60147
1124         * tree-pretty-print.c (dump_generic_node, print_declaration): Handle
1125         NAMELIST_DECL.
1126
1127 2014-02-27  H.J. Lu  <hongjiu.lu@intel.com>
1128
1129         * doc/tm.texi.in (Condition Code Status): Update documention for
1130         relative locations of cc0-setter and cc0-user.
1131
1132 2014-02-27  Jeff Law  <law@redhat.com>
1133
1134         PR rtl-optimization/52714
1135         * combine.c (try_combine): When splitting an unrecognized PARALLEL
1136         into two independent simple sets, if I3 is a jump, ensure the
1137         pattern we place into I3 is a (set (pc) ...).
1138
1139 2014-02-27  Mikael Pettersson  <mikpe@it.uu.se>
1140             Jeff Law  <law@redhat.com>
1141
1142         PR rtl-optimization/49847
1143         * cse.c (fold_rtx) Handle case where cc0 setter and cc0 user
1144         are in different blocks.
1145         * doc/tm.texi (Condition Code Status): Update documention for
1146         relative locations of cc0-setter and cc0-user.
1147
1148 2014-02-27  Vladimir Makarov  <vmakarov@redhat.com>
1149
1150         PR target/59222
1151         * lra.c (lra_emit_add): Check SUBREG too.
1152
1153 2014-02-27  Andreas Schwab  <schwab@suse.de>
1154
1155         * config/m68k/m68k.c (m68k_option_override): Disable
1156         -flive-range-shrinkage for classic m68k.
1157         (m68k_override_options_after_change): Likewise.
1158
1159 2014-02-27  Marek Polacek  <polacek@redhat.com>
1160
1161         PR middle-end/59223
1162         * tree-ssa-uninit.c (gate_warn_uninitialized): Run the pass even for
1163         -Wmaybe-uninitialized.
1164
1165 2014-02-27  Alan Modra  <amodra@gmail.com>
1166
1167         PR target/57936
1168         * reload1.c (emit_input_reload_insns): When reload_override_in,
1169         set old to rl->in_reg when rl->in_reg is a subreg.
1170
1171 2014-02-26  Richard Biener  <rguenther@suse.de>
1172
1173         PR bootstrap/60343
1174         * lra-assigns.c (spill_for): Avoid mixed-sign comparison.
1175
1176 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
1177
1178         * common/config/i386/predicates.md (const1256_operand): Remove.
1179         (const2356_operand): New.
1180         (const_1_to_2_operand): Remove.
1181         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
1182         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
1183         (*avx512pf_gatherpf<mode>sf): Ditto.
1184         (avx512pf_gatherpf<mode>df): Ditto.
1185         (*avx512pf_gatherpf<mode>df_mask): Ditto.
1186         (*avx512pf_gatherpf<mode>df): Ditto.
1187         (avx512pf_scatterpf<mode>sf): Ditto.
1188         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1189         (*avx512pf_scatterpf<mode>sf): Ditto.
1190         (avx512pf_scatterpf<mode>df): Ditto.
1191         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1192         (*avx512pf_scatterpf<mode>df): Ditto.
1193         * common/config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET0.
1194
1195 2014-02-26  Ilya Tocar  <ilya.tocar@intel.com>
1196
1197         * config/i386/avx512fintrin.h (_mm512_testn_epi32_mask),
1198         (_mm512_mask_testn_epi32_mask), (_mm512_testn_epi64_mask),
1199         (_mm512_mask_testn_epi64_mask): Move to ...
1200         * config/i386/avx512cdintrin.h: Here.
1201         * config/i386/i386.c (bdesc_args): Change MASK_ISA for testnm.
1202         * config/i386/sse.md (avx512f_vmscalef<mode><round_name>): Remove %.
1203         (avx512f_scalef<mode><mask_name><round_name>): Ditto.
1204         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Change conditon to
1205         TARGET_AVX512F from TARGET_AVX512CD.
1206
1207 2014-02-26  Richard Biener  <rguenther@suse.de>
1208
1209         PR ipa/60327
1210         * ipa.c (walk_polymorphic_call_targets): Properly guard
1211         call to inline_update_overall_summary.
1212
1213 2014-02-26  Bin Cheng  <bin.cheng@arm.com>
1214
1215         PR target/60280
1216         * tree-cfgcleanup.c (tree_forwarder_block_p): Protect loop preheaders
1217         and latches only if requested.  Fix latch if it is removed.
1218         * tree-ssa-dom.c (tree_ssa_dominator_optimize): Set
1219         LOOPS_HAVE_PREHEADERS.
1220
1221 2014-02-25  Andrew Pinski  <apinski@cavium.com>
1222
1223         * builtins.c (expand_builtin_thread_pointer): Create a new target
1224         when the target is NULL.
1225
1226 2014-02-25  Vladimir Makarov  <vmakarov@redhat.com>
1227
1228         PR rtl-optimization/60317
1229         * params.def (PARAM_LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
1230         * params.h (LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
1231         * lra-assigns.c: Include params.h.
1232         (spill_for): Use LRA_MAX_CONSIDERED_RELOAD_PSEUDOS as guard for
1233         other reload pseudos considerations.
1234
1235 2014-02-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1236
1237         * config/rs6000/vector.md (*vector_unordered<mode>): Change split
1238         to use canonical form for nor<mode>3.
1239
1240 2014-02-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1241
1242         PR target/55426
1243         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Allow 128 to 64-bit
1244         conversions.
1245
1246 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
1247
1248         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PREFETCHWT1_SET),
1249         (OPTION_MASK_ISA_PREFETCHWT1_UNSET): New.
1250         (ix86_handle_option): Handle OPT_mprefetchwt1.
1251         * config/i386/cpuid.h (bit_PREFETCHWT1): New.
1252         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1253         PREFETCHWT1 CPUID.
1254         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1255         OPTION_MASK_ISA_PREFETCHWT1.
1256         * config/i386/i386.c (ix86_target_string): Handle mprefetchwt1.
1257         (PTA_PREFETCHWT1): New.
1258         (ix86_option_override_internal): Handle PTA_PREFETCHWT1.
1259         (ix86_valid_target_attribute_inner_p): Handle OPT_mprefetchwt1.
1260         * config/i386/i386.h (TARGET_PREFETCHWT1, TARGET_PREFETCHWT1_P): New.
1261         * config/i386/i386.md (prefetch): Check TARGET_PREFETCHWT1
1262         (*prefetch_avx512pf_<mode>_: Change into ...
1263         (*prefetch_prefetchwt1_<mode>: This.
1264         * config/i386/i386.opt (mprefetchwt1): New.
1265         * config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET1.
1266         (_mm_prefetch): Handle intent to write.
1267         * doc/invoke.texi (mprefetchwt1), (mno-prefetchwt1): Doccument.
1268
1269 2014-02-25  Richard Biener  <rguenther@suse.de>
1270
1271         PR middle-end/60291
1272         * emit-rtl.c (mem_attrs_htab): Remove.
1273         (mem_attrs_htab_hash): Likewise.
1274         (mem_attrs_htab_eq): Likewise.
1275         (set_mem_attrs): Always allocate new mem-attrs when something changed.
1276         (init_emit_once): Do not allocate mem_attrs_htab.
1277
1278 2014-02-25  Richard Biener  <rguenther@suse.de>
1279
1280         PR lto/60319
1281         * lto-opts.c (lto_write_options): Output non-explicit conservative
1282         -fwrapv, -fno-trapv and -fno-strict-overflow.
1283         * lto-wrapper.c (merge_and_complain): Handle merging those options.
1284         (run_gcc): And pass them through.
1285
1286 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
1287
1288         * sel-sched.c (calculate_new_fences): New parameter ptime.
1289         Calculate it as a maximum over all fence cycles.
1290         (sel_sched_region_2): Adjust the call to calculate_new_fences.
1291         Print the final schedule timing when sched_verbose.
1292
1293 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
1294
1295         PR rtl-optimization/60292
1296         * sel-sched.c (fill_vec_av_set): Do not reset target availability
1297         bit fot the fence instruction.
1298
1299 2014-02-24  Alangi Derick  <alangiderick@gmail.com>
1300
1301         * calls.h: Fix typo in comment.
1302
1303 2014-02-24  John David Anglin  <danglin@gcc.gnu.org>
1304
1305         * config/pa/pa.c (pa_output_move_double): Don't valididate when
1306         adjusting offsetable addresses.
1307
1308 2014-02-24  Guozhi Wei  <carrot@google.com>
1309
1310         * sparseset.h (sparseset_pop): Fix the wrong index.
1311
1312 2014-02-24  Walter Lee  <walt@tilera.com>
1313
1314         * config.gcc (tilepro-*-*): Change to tilepro*-*-*.
1315         (tilegx-*-linux*): Change to tilegx*-*-linux*; Support tilegxbe
1316         triplet.
1317         * common/config/tilegx/tilegx-common.c
1318         (TARGET_DEFAULT_TARGET_FLAGS): Define.
1319         * config/tilegx/linux.h (ASM_SPEC): Add endian_spec.
1320         (LINK_SPEC): Ditto.
1321         * config/tilegx/sync.md (atomic_test_and_set): Handle big endian.
1322         * config/tilegx/tilegx.c (tilegx_return_in_msb): New.
1323         (tilegx_gimplify_va_arg_expr): Handle big endian.
1324         (tilegx_expand_unaligned_load): Ditto.
1325         (tilegx_expand_unaligned_store): Ditto.
1326         (TARGET_RETURN_IN_MSB): New.
1327         * config/tilegx/tilegx.h (TARGET_DEFAULT): New.
1328         (TARGET_ENDIAN_DEFAULT): New.
1329         (TARGET_BIG_ENDIAN): Handle big endian.
1330         (BYTES_BIG_ENDIAN): Ditto.
1331         (WORDS_BIG_ENDIAN): Ditto.
1332         (FLOAT_WORDS_BIG_ENDIAN): Ditto.
1333         (ENDIAN_SPEC): New.
1334         (EXTRA_SPECS): New.
1335         * config/tilegx/tilegx.md (extv): Handle big endian.
1336         (extzv): Ditto.
1337         (insn_st<n>): Ditto.
1338         (insn_st<n>_add<bitsuffix>): Ditto.
1339         (insn_stnt<n>): Ditto.
1340         (insn_stnt<n>_add<bitsuffix>):Ditto.
1341         (vec_interleave_highv8qi): Handle big endian.
1342         (vec_interleave_highv8qi_be): New.
1343         (vec_interleave_highv8qi_le): New.
1344         (insn_v1int_h): Handle big endian.
1345         (vec_interleave_lowv8qi): Handle big endian.
1346         (vec_interleave_lowv8qi_be): New.
1347         (vec_interleave_lowv8qi_le): New.
1348         (insn_v1int_l): Handle big endian.
1349         (vec_interleave_highv4hi): Handle big endian.
1350         (vec_interleave_highv4hi_be): New.
1351         (vec_interleave_highv4hi_le): New.
1352         (insn_v2int_h): Handle big endian.
1353         (vec_interleave_lowv4hi): Handle big endian.
1354         (vec_interleave_lowv4hi_be): New.
1355         (vec_interleave_lowv4hi_le): New.
1356         (insn_v2int_l): Handle big endian.
1357         (vec_interleave_highv2si): Handle big endian.
1358         (vec_interleave_highv2si_be): New.
1359         (vec_interleave_highv2si_le): New.
1360         (insn_v4int_h): Handle big endian.
1361         (vec_interleave_lowv2si): Handle big endian.
1362         (vec_interleave_lowv2si_be): New.
1363         (vec_interleave_lowv2si_le): New.
1364         (insn_v4int_l): Handle big endian.
1365         * config/tilegx/tilegx.opt (mbig-endian): New option.
1366         (mlittle-endian): New option.
1367         * doc/install.texi: Document tilegxbe-linux.
1368         * doc/invoke.texi: Document -mbig-endian and -mlittle-endian.
1369
1370 2014-02-24  Martin Jambor  <mjambor@suse.cz>
1371
1372         PR ipa/60266
1373         * ipa-cp.c (propagate_constants_accross_call): Bail out early if
1374         there are no parameter descriptors.
1375
1376 2014-02-24  Andrey Belevantsev  <abel@ispras.ru>
1377
1378         PR rtl-optimization/60268
1379         * sched-rgn.c (haifa_find_rgns): Move the nr_regions_initial variable
1380         initialization to ...
1381         (sched_rgn_init): ... here.
1382         (schedule_region): Check for SCHED_PRESSURE_NONE earlier.
1383
1384 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1385
1386         * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg
1387         names.
1388
1389 2014-02-23  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
1390
1391         * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
1392         definition.
1393
1394 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1395
1396         * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk,
1397         define TARGET_ASM_OUTPUT_MI_THUNK and TARGET_ASM_CAN_OUTPUT_MI_THUNK.
1398
1399 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1400
1401         * config/microblaze/predicates.md: Add cmp_op predicate.
1402         * config/microblaze/microblaze.md: Add branch_compare instruction
1403         which uses cmp_op predicate and emits cmp insn before branch.
1404         * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
1405         to microblaze_expand_conditional_branch and consolidate logic.
1406         (microblaze_expand_conditional_branch): emit branch_compare
1407         insn instead of handling cmp op separate from branch insn.
1408
1409 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1410
1411         * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
1412         to permit subregs.
1413
1414 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1415
1416         * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
1417         define_insn with define_expand and new define_insn
1418         *altivec_lve<VI_char>x_internal.
1419         (altivec_stve<VI_char>x): Replace define_insn with define_expand
1420         and new define_insn *altivec_stve<VI_char>x_internal.
1421         * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
1422         prototype.
1423         * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
1424         lve*x built-ins.
1425         (altivec_expand_stvex_be): New function.
1426
1427 2014-02-22  Joern Rennecke  <joern.rennecke@embecosm.com>
1428
1429         * config/avr/avr.c (avr_can_eliminate): Allow elimination from
1430         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM if !frame_pointer_needed.
1431         * config/avr/avr.c (ELIMINABLE_REGS): Add elimination from
1432         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM.
1433
1434 2014-02-21  Vladimir Makarov  <vmakarov@redhat.com>
1435
1436         PR target/60298
1437         * lra-constraints.c (inherit_reload_reg): Use lra_emit_move
1438         instead of emit_move_insn.
1439
1440 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1441
1442         * config/rs6000/altivec.md (altivec_vsumsws): Replace second
1443         vspltw with vsldoi.
1444         (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
1445         gen_altivec_vsumsws.
1446
1447 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1448
1449         * config/rs6000/altivec.md (altivec_lvxl): Rename as
1450         *altivec_lvxl_<mode>_internal and use VM2 iterator instead of V4SI.
1451         (altivec_lvxl_<mode>): New define_expand incorporating
1452         -maltivec=be semantics where needed.
1453         (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
1454         (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
1455         semantics where needed.
1456         (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
1457         (altivec_stvx_<mode>): New define_expand incorporating
1458         -maltivec=be semantics where needed.
1459         (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
1460         VM2 iterator instead of V4SI.
1461         (altivec_stvxl_<mode>): New define_expand incorporating
1462         -maltivec=be semantics where needed.
1463         * config/rs6000/rs6000-builtin.def: Add new built-in definitions
1464         LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
1465         LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI, STVX_V2DF,
1466         STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI, STVXL_V2DF,
1467         STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI, STVXL_V16QI.
1468         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
1469         ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
1470         similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
1471         ALTIVEC_BUILTIN_STVXL.
1472         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New prototype.
1473         (altivec_expand_stvx_be): Likewise.
1474         * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
1475         (altivec_expand_lvx_be): Likewise.
1476         (altivec_expand_stvx_be): Likewise.
1477         (altivec_expand_builtin): Add cases for
1478         ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
1479         ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
1480         (altivec_init_builtins): Add definitions for
1481         __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
1482         __builtin_altivec_stvx_<mode>, and __builtin_altivec_stvxl_<mode>.
1483
1484 2014-02-21  Catherine Moore  <clm@codesourcery.com>
1485
1486         * doc/invoke.texi (mvirt, mno-virt): Document.
1487         * config/mips/mips.opt (mvirt): New option.
1488         * config/mips/mips.h (ASM_SPEC): Pass mvirt to the assembler.
1489
1490 2014-02-21  Richard Biener  <rguenther@suse.de>
1491
1492         PR tree-optimization/60276
1493         * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
1494         (STMT_VINFO_MIN_NEG_DIST): New macro.
1495         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
1496         STMT_VINFO_MIN_NEG_DIST.
1497         * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
1498         made for negative dependence distances still hold.
1499
1500 2014-02-21  Richard Biener  <rguenther@suse.de>
1501
1502         PR middle-end/60291
1503         * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
1504         DECL_INITIAL for globals not in the current function context.
1505
1506 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
1507
1508         PR tree-optimization/56490
1509         * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
1510         * tree-ssa-uninit.c: Include params.h.
1511         (compute_control_dep_chain): Add num_calls argument, return false
1512         if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
1513         num_calls to recursive call.
1514         (find_predicates): Change dep_chain into normal array,
1515         cur_chain into auto_vec<edge, MAX_CHAIN_LEN + 1>, add num_calls
1516         variable and adjust compute_control_dep_chain caller.
1517         (find_def_preds): Likewise.
1518
1519 2014-02-21  Thomas Schwinge  <thomas@codesourcery.com>
1520
1521         * gimple-pretty-print.c (dump_gimple_omp_for) [flags & TDF_RAW]
1522         <case GF_OMP_FOR_KIND_CILKSIMD>: Add missing break statement.
1523
1524 2014-02-21  Nick Clifton  <nickc@redhat.com>
1525
1526         * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
1527         (pushhi1): Likewise.
1528         (popqi1): Add mode to pre_dec.
1529         (pophi1): Likewise.
1530
1531 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
1532
1533         * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
1534         mode for mask of V8SFmode permutation.
1535
1536 2014-02-20  Richard Henderson  <rth@redhat.com>
1537
1538         PR c++/60272
1539         * builtins.c (expand_builtin_atomic_compare_exchange): Always make
1540         a new pseudo for OLDVAL.
1541
1542 2014-02-20  Jakub Jelinek  <jakub@redhat.com>
1543
1544         PR target/57896
1545         * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
1546         gen_reg_rtx if d->testing_p.
1547         (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
1548         if d->testing_p and we will certainly return true.
1549         (expand_vec_perm_even_odd_1): Likewise.  Don't call gen_reg_rtx
1550         if d->testing_p.
1551
1552 2014-02-20  Uros Bizjak  <ubizjak@gmail.com>
1553
1554         * emit-rtl.c (gen_reg_rtx): Assert that
1555         crtl->emit.regno_pointer_align_length is non-zero.
1556
1557 2014-02-20  Richard Henderson  <rth@redhat.com>
1558
1559         PR c++/60272
1560         * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
1561         on failure the store back into EXPECT.
1562
1563 2014-02-20  Chung-Lin Tang  <cltang@codesourcery.com>
1564             Sandra Loosemore  <sandra@codesourcery.com>
1565
1566         * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
1567         * config/nios2/nios2.c (nios2_function_profiler): Add
1568         -fPIC (flag_pic == 2) support.
1569         (nios2_handle_custom_fpu_cfg): Fix warning parameter.
1570         (nios2_large_offset_p): New function.
1571         (nios2_unspec_reloc_p): Move up position, update to use
1572         nios2_large_offset_p.
1573         (nios2_unspec_address): Remove function.
1574         (nios2_unspec_offset): New function.
1575         (nios2_large_got_address): New function.
1576         (nios2_got_address): Add large offset support.
1577         (nios2_legitimize_tls_address): Update usage of removed and new
1578         functions.
1579         (nios2_symbol_binds_local_p): New function.
1580         (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
1581         (nios2_legitimize_address): Update to use nios2_large_offset_p.
1582         (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
1583         (nios2_print_operand): Merge H/L processing, add hiadj/lo
1584         processing for (const (unspec ...)).
1585         (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
1586
1587 2014-02-20  Richard Biener  <rguenther@suse.de>
1588
1589         * tree-cfg.c (replace_uses_by): Mark altered BBs before
1590         doing the substitution.
1591         (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
1592
1593 2014-02-20  Martin Jambor  <mjambor@suse.cz>
1594
1595         PR ipa/55260
1596         * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
1597         info when checking whether lattices are bottom.
1598
1599 2014-02-20  Richard Biener  <rguenther@suse.de>
1600
1601         PR middle-end/60221
1602         * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
1603         regions at -O0.
1604
1605 2014-02-20  Jan Hubicka  <hubicka@ucw.cz>
1606
1607         PR ipa/58555
1608         * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
1609         parameter specifying the scaling.
1610         (inline_call): Update.
1611         (want_inline_recursively): Guard division by zero.
1612         (recursive_inlining): Update.
1613         * ipa-inline.h (clone_inlined_nodes): Update.
1614
1615 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
1616
1617         PR target/60204
1618         * config/i386/i386.c (classify_argument): Pass structures of size
1619         64 bytes or less in register.
1620
1621 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
1622             Kirill Yukhin  <kirill.yukhin@intel.com>
1623
1624         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
1625         (_mm_rcp28_round_ss): Ditto.
1626         (_mm_rsqrt28_round_sd): Ditto.
1627         (_mm_rsqrt28_round_ss): Ditto.
1628         * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
1629         (_mm_rcp14_round_ss): Ditto.
1630         (_mm_rsqrt14_round_sd): Ditto.
1631         (_mm_rsqrt14_round_ss): Ditto.
1632         * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
1633         the first input operand, get rid of match_dup.
1634         (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
1635         attribute to sse.
1636         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
1637         Ditto.
1638         (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
1639         operand as the first input operand, set type attribute.
1640         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
1641         Set type attribute.
1642         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
1643         operand as the first input operand, set type attribute.
1644
1645 2014-02-19  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1646
1647         * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
1648         bit of zero.
1649
1650 2014-02-19  H.J. Lu  <hongjiu.lu@intel.com>
1651
1652         PR target/60207
1653         * config/i386/i386.c (construct_container): Remove TFmode check
1654         for X86_64_INTEGER_CLASS.
1655
1656 2014-02-19  Uros Bizjak  <ubizjak@gmail.com>
1657
1658         PR target/59794
1659         * config/i386/i386.c (type_natural_mode): Warn for ABI changes
1660         only when -Wpsabi is enabled.
1661
1662 2014-02-19  Michael Hudson-Doyle  <michael.hudson@linaro.org>
1663
1664          PR target/59799
1665         * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
1666         passing arrays in registers are the same as for structs, so remove the
1667         special case for them.
1668
1669 2014-02-19  Eric Botcazou  <ebotcazou@adacore.com>
1670
1671         * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
1672         destination type, extract only the valid bits if the source type is not
1673         integral and has a different mode.
1674
1675 2014-02-19  Richard Biener  <rguenther@suse.de>
1676
1677         PR ipa/60243
1678         * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
1679         for all calls.
1680
1681 2014-02-19  Richard Biener  <rguenther@suse.de>
1682
1683         PR ipa/60243
1684         * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
1685         (ipa_modify_call_arguments): Emit an argument load explicitely and
1686         preserve virtual SSA form there and for the replacement call.
1687         Do not update SSA form nor free dominance info.
1688
1689 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1690
1691         * ipa.c (function_and_variable_visibility): Also clear WEAK
1692         flag when disolving COMDAT_GROUP.
1693
1694 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1695
1696         * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
1697         * ipa-prop.c (ipa_set_jf_known_type): Return early when
1698         not devirtualizing.
1699         (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
1700         do more sanity checks.
1701         (detect_type_change): Return true when giving up early.
1702         (compute_complex_assign_jump_func): Fix type parameter of
1703         ipa_set_ancestor_jf.
1704         (compute_complex_ancestor_jump_func): Likewise.
1705         (update_jump_functions_after_inlining): Fix updating of
1706         ancestor function.
1707         * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
1708
1709 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1710
1711         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
1712         inline clones when edge disappears.
1713
1714 2014-02-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
1715
1716         PR target/60203
1717         * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
1718         Split 64-bit moves into 2 patterns.  Do not allow the use of
1719         direct move for TDmode in little endian, since the decimal value
1720         has little endian bytes within a word, but the 64-bit pieces are
1721         ordered in a big endian fashion, and normal subreg's of TDmode are
1722         not allowed.
1723         (mov<mode>_64bit_dm): Likewise.
1724         (movtd_64bit_nodm): Likewise.
1725
1726 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
1727
1728         PR tree-optimization/60174
1729         * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
1730         statement of an SSA_NAME that occurs in an abnormal PHI node.
1731
1732 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
1733
1734         PR sanitizer/60142
1735         * final.c (SEEN_BB): Remove.
1736         (SEEN_NOTE, SEEN_EMITTED): Renumber.
1737         (final_scan_insn): Don't force_source_line on second
1738         NOTE_INSN_BASIC_BLOCK.
1739
1740 2014-02-18  Uros Bizjak  <ubizjak@gmail.com>
1741
1742         PR target/60205
1743         * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
1744         * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
1745         (type_natural_mode): Warn ABI change when %zmm register is not
1746         available for AVX512F vector value passing.
1747
1748 2014-02-18  Kai Tietz  <ktietz@redhat.com>
1749
1750         PR target/60193
1751         * config/i386/i386.c (ix86_expand_prologue): Use value in
1752         rax register as displacement when restoring %r10 or %rax.
1753         Fix wrong offset when restoring both registers.
1754
1755 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
1756
1757         * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
1758         assertion with conditional return.
1759
1760 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
1761             Uros Bizjak  <ubizjak@gmail.com>
1762
1763         PR driver/60233
1764         * config/i386/driver-i386.c (host_detect_local_cpu): If
1765         YMM state is not saved by the OS, also clear has_f16c.  Move
1766         CPUID 0x80000001 handling before YMM state saving checking.
1767
1768 2014-02-18  Andrey Belevantsev  <abel@ispras.ru>
1769
1770         PR rtl-optimization/58960
1771         * haifa-sched.c (alloc_global_sched_pressure_data): New,
1772         factored out from ...
1773         (sched_init): ... here.
1774         (free_global_sched_pressure_data): New, factored out from ...
1775         (sched_finish): ... here.
1776         * sched-int.h (free_global_sched_pressure_data): Declare.
1777         * sched-rgn.c (nr_regions_initial): New static global.
1778         (haifa_find_rgns): Initialize it.
1779         (schedule_region): Disable sched-pressure for the newly
1780         generated regions.
1781
1782 2014-02-17  Richard Biener  <rguenther@suse.de>
1783
1784         * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
1785         release SSA defs of pattern stmts.
1786
1787 2014-02-17  Richard Biener  <rguenther@suse.de>
1788
1789         * tree-inline.c (expand_call_inline): Release the virtual
1790         operand defined by the call we are about to inline.
1791
1792 2014-02-17  Richard Biener  <rguenther@suse.de>
1793
1794         * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
1795
1796 2014-02-17  Kirill Yukhin  <kirill.yukhin@intel.com>
1797             Ilya Tocar  <ilya.tocar@intel.com>
1798
1799         * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
1800         arguments order in builtin.
1801         (_mm512_permutexvar_epi64): Ditto.
1802         (_mm512_mask_permutexvar_epi64): Ditto
1803         (_mm512_maskz_permutexvar_epi32): Ditto
1804         (_mm512_permutexvar_epi32): Ditto
1805         (_mm512_mask_permutexvar_epi32): Ditto
1806
1807 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1808
1809         * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
1810         (p8_vmrgow): Likewise.
1811
1812 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1813
1814         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
1815         endian targets.
1816
1817 2014-02-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
1818
1819         PR target/60203
1820         * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
1821         (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
1822         into 64-bit and 32-bit moves.  On 64-bit moves, add support for
1823         using direct move instructions on ISA 2.07.  Also adjust
1824         instruction length for 64-bit.
1825         (mov<mode>_64bit, TFmode/TDmode): Likewise.
1826         (mov<mode>_32bit, TFmode/TDmode): Likewise.
1827
1828 2014-02-15  Alan Modra  <amodra@gmail.com>
1829
1830         PR target/58675
1831         PR target/57935
1832         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
1833         find_replacement on parts of insn rtl that might be reloaded.
1834
1835 2014-02-15  Richard Biener  <rguenther@suse.de>
1836
1837         PR tree-optimization/60183
1838         * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
1839         (tree_ssa_phiprop): Calculate and free post-dominators.
1840
1841 2014-02-14  Jeff Law  <law@redhat.com>
1842
1843         PR rtl-optimization/60131
1844         * ree.c (get_extended_src_reg): New function.
1845         (combine_reaching_defs): Use it rather than assuming location of REG.
1846         (find_and_remove_re): Verify first operand of extension is
1847         a REG before adding the insns to the copy list.
1848
1849 2014-02-14  Roland McGrath  <mcgrathr@google.com>
1850
1851         * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
1852         * configure: Regenerated.
1853         * config.in: Regenerated.
1854         * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
1855         instead of ASM_SHORT.
1856
1857 2014-02-14  Vladimir Makarov  <vmakarov@redhat.com>
1858             Richard Earnshaw  <rearnsha@arm.com>
1859
1860         PR rtl-optimization/59535
1861         * lra-constraints.c (process_alt_operands): Encourage alternative
1862         when unassigned pseudo class is superset of the alternative class.
1863         (inherit_reload_reg): Don't inherit when optimizing for code size.
1864         * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
1865         returning CORE_REGS for anything but Thumb1 and BASE_REGS for
1866         modes not less than 4 for Thumb1.
1867
1868 2014-02-14  Kyle McMartin  <kyle@redhat.com>
1869
1870         PR pch/60010
1871         * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
1872
1873 2014-02-14  Richard Biener  <rguenther@suse.de>
1874
1875         * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
1876         (get_frame_arg): Drop the assert with langhook types_compatible_p.
1877         Do not strip INDIRECT_REFs.
1878
1879 2014-02-14  Richard Biener  <rguenther@suse.de>
1880
1881         PR lto/60179
1882         * lto-streamer-out.c (DFS_write_tree_body): Do not follow
1883         DECL_FUNCTION_SPECIFIC_TARGET.
1884         (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
1885         * tree-streamer-out.c (pack_ts_target_option): Remove.
1886         (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
1887         (write_ts_function_decl_tree_pointers): Do not stream
1888         DECL_FUNCTION_SPECIFIC_TARGET.
1889         * tree-streamer-in.c (unpack_ts_target_option): Remove.
1890         (unpack_value_fields): Do not stream TS_TARGET_OPTION.
1891         (lto_input_ts_function_decl_tree_pointers): Do not stream
1892         DECL_FUNCTION_SPECIFIC_TARGET.
1893
1894 2014-02-14  Jakub Jelinek  <jakub@redhat.com>
1895
1896         * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
1897         (get_initial_def_for_induction, vectorizable_induction): Ignore
1898         debug stmts when looking for exit_phi.
1899         (vectorizable_live_operation): Fix up condition.
1900
1901 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1902
1903         * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
1904         nreverse() because it changes the content of original tree list.
1905
1906 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1907
1908         * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
1909         * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
1910
1911 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1912
1913         * config/nds32/nds32.c (nds32_naked_function_p): Follow the
1914         GNU coding standards.
1915
1916 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
1917
1918         PR debug/60152
1919         * dwarf2out.c (gen_subprogram_die): Don't call
1920         add_calling_convention_attribute if subr_die is old_die.
1921
1922 2014-02-13  Sharad Singhai  <singhai@google.com>
1923
1924         * doc/optinfo.texi: Fix order of nodes.
1925
1926 2014-02-13  Uros Bizjak  <ubizjak@gmail.com>
1927
1928         * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
1929         operands[2], not operands[3].
1930
1931 2014-02-13  Richard Biener  <rguenther@suse.de>
1932
1933         PR bootstrap/59878
1934         * doc/install.texi (ISL): Update recommended version to 0.12.2,
1935         mention the possibility of an in-tree build.
1936         (CLooG): Update recommended version to 0.18.1, mention the
1937         possibility of an in-tree build and clarify that the ISL
1938         bundled with CLooG does not work.
1939
1940 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
1941
1942         PR target/43546
1943         * expr.c (compress_float_constant): If x is a hard register,
1944         extend into a pseudo and then move to x.
1945
1946 2014-02-13  Dominik Vogt  <vogt@linux.vnet.ibm.com>
1947
1948         * config/s390/s390.c (s390_asm_output_function_label): Fix crash
1949         caused by bad second argument to warning_at() with -mhotpatch and
1950         nested functions (e.g. with gfortran).
1951
1952 2014-02-13  Richard Sandiford  <rdsandiford@googlemail.com>
1953
1954         * opts.c (option_name): Remove "enabled by default" rider.
1955
1956 2014-02-12  John David Anglin  <danglin@gcc.gnu.org>
1957
1958         * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
1959
1960 2014-02-12  H.J. Lu  <hongjiu.lu@intel.com>
1961             Uros Bizjak  <ubizjak@gmail.com>
1962
1963         PR target/60151
1964         * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
1965         * configure: Regenerated.
1966
1967 2014-02-12  Richard Biener  <rguenther@suse.de>
1968
1969         * vec.c (vec_prefix::calculate_allocation): Move as
1970         inline variant to vec.h.
1971         (vec_prefix::calculate_allocation_1): New out-of-line version.
1972         * vec.h (vec_prefix::calculate_allocation_1): Declare.
1973         (vec_prefix::m_has_auto_buf): Rename to ...
1974         (vec_prefix::m_using_auto_storage): ... this.
1975         (vec_prefix::calculate_allocation): Inline the easy cases
1976         and dispatch to calculate_allocation_1 which doesn't need the
1977         prefix address.
1978         (va_heap::reserve): Use gcc_checking_assert.
1979         (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
1980         m_using_auto_storage.
1981         (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
1982         member and adjust.
1983         (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
1984         (vec<T, va_heap, vl_ptr>::release): Avoid casting.
1985         (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
1986
1987 2014-02-12  Richard Biener  <rguenther@suse.de>
1988
1989         * gcse.c (compute_transp): break from loop over canon_modify_mem_list
1990         when we found a dependence.
1991
1992 2014-02-12  Thomas Schwinge  <thomas@codesourcery.com>
1993
1994         * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
1995         common code...
1996         (maybe_fold_stmt): ... into this new function.
1997         * omp-low.c (lower_omp): Update comment.
1998
1999         * omp-low.c (lower_omp_target): Add clobber for sizes array, after
2000         last use.
2001
2002         * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
2003         dereference.
2004
2005 2014-02-12  James Greenhalgh  <james.greenhalgh@arm.com>
2006
2007         * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
2008         identifiers in comments.
2009         (cortexa53_extra_costs): Likewise.
2010         * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
2011         (cortexa7_extra_costs): Likewise.
2012         (cortexa12_extra_costs): Likewise.
2013         (cortexa15_extra_costs): Likewise.
2014         (v7m_extra_costs): Likewise.
2015
2016 2014-02-12  Richard Biener  <rguenther@suse.de>
2017
2018         PR middle-end/60092
2019         * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
2020         of posix_memalign being successful.
2021         (lower_stmt): Restrict lowering of posix_memalign to when
2022         -ftree-bit-ccp is enabled.
2023
2024 2014-02-12  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
2025
2026         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
2027         arg_loc.
2028         * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
2029
2030 2014-02-12  Eric Botcazou  <ebotcazou@adacore.com>
2031
2032         PR rtl-optimization/60116
2033         * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
2034         other_insn once the combination has been validated.
2035
2036 2014-02-11  Jan Hubicka  <hubicka@ucw.cz>
2037
2038         PR lto/59468
2039         * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
2040         and wrapper.
2041         * ipa-devirt.c: Include demangle.h
2042         (odr_violation_reported): New static variable.
2043         (add_type_duplicate): Update odr_violations.
2044         (maybe_record_node): Add completep parameter; update it.
2045         (record_target_from_binfo): Add COMPLETEP parameter;
2046         update it as needed.
2047         (possible_polymorphic_call_targets_1): Likewise.
2048         (struct polymorphic_call_target_d): Add nonconstruction_targets;
2049         rename FINAL to COMPLETE.
2050         (record_targets_from_bases): Sanity check we found the binfo;
2051         fix COMPLETEP updating.
2052         (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
2053         parameter, fix computing of COMPLETEP.
2054         (dump_possible_polymorphic_call_targets): Imrove readability of dump;
2055         at LTO time do demangling.
2056         (ipa_devirt): Use nonconstruction_targets; Improve dumps.
2057         * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
2058         parameter.
2059         (gimple_get_virt_method_for_binfo): Likewise.
2060         * gimple-fold.h (gimple_get_virt_method_for_binfo,
2061         gimple_get_virt_method_for_vtable): Update prototypes.
2062
2063 2014-02-11  Vladimir Makarov  <vmakarov@redhat.com>
2064
2065         PR target/49008
2066         * genautomata.c (add_presence_absence): Fix typo with
2067         {final_}presence_list.
2068
2069 2014-02-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
2070
2071         PR target/60137
2072         * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
2073         for VSX/Altivec vectors that land in GPR registers.
2074
2075 2014-02-11  Richard Henderson  <rth@redhat.com>
2076             Jakub Jelinek  <jakub@redhat.com>
2077
2078         PR debug/59776
2079         * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
2080         around drhs if type conversion to lacc->type is not useless.
2081
2082 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2083
2084         * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
2085         tuning struct.
2086         (cortex-a57.cortex-a53): Likewise.
2087         * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
2088
2089 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2090
2091         * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
2092         arm_restrict_it.
2093
2094 2014-02-11  Renlin Li  <Renlin.Li@arm.com>
2095
2096         * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
2097         add_options_for_arm_vfp3.
2098
2099 2014-02-11  Jeff Law  <law@redhat.com>
2100
2101         PR middle-end/54041
2102         * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
2103         object with an undesirable mode.
2104
2105 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2106
2107         PR libgomp/60107
2108         * config/i386/sol2-9.h: New file.
2109         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
2110         *-*-solaris2.9*): Use it.
2111
2112 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
2113
2114         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
2115         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
2116
2117 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
2118
2119         * config/microblaze/microblaze.c: Extend mcpu version format
2120
2121 2014-02-10  David Holsgrove <david.holsgrove@xilinx.com>
2122
2123         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
2124
2125 2014-02-10  Richard Henderson  <rth@redhat.com>
2126
2127         PR target/59927
2128         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
2129         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
2130         ms-abi vs -mno-accumulate-outgoing-args.
2131         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
2132         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
2133         respect to ms-abi.
2134
2135 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2136
2137         PR middle-end/60080
2138         * cfgexpand.c (expand_asm_operands): Attach source location to
2139         ASM_INPUT rtx objects.
2140         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
2141
2142 2014-02-10  Nick Clifton  <nickc@redhat.com>
2143
2144         * config/mn10300/mn10300.c (popcount): New function.
2145         (mn10300_expand_prologue): Include saved registers in stack usage
2146         count.
2147
2148 2014-02-10  Jeff Law  <law@redhat.com>
2149
2150         PR middle-end/52306
2151         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
2152         when changing the SET_DEST of a prior insn to avoid an input reload.
2153
2154 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
2155
2156         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
2157         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
2158         -mcall-openbsd, or -mcall-linux.
2159         (CC1_ENDIAN_BIG_SPEC): Remove.
2160         (CC1_ENDIAN_LITTLE_SPEC): Remove.
2161         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
2162         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
2163         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
2164         and %cc1_endian_default.
2165         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
2166
2167 2014-02-10  Richard Biener  <rguenther@suse.de>
2168
2169         PR tree-optimization/60115
2170         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
2171         MEM_REF handling.  Properly verify that the accesses are not
2172         out of the objects bound.
2173
2174 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2175
2176         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
2177         coretex to cortex.
2178
2179 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
2180
2181         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
2182         proper constants and fix formatting.
2183         (possible_polymorphic_call_targets): Fix formatting.
2184
2185 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
2186             Ilya Tocar  <ilya.tocar@intel.com>
2187
2188         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
2189         (_mm512_loadu_epi32): Renamed into...
2190         (_mm512_loadu_si512): This.
2191         (_mm512_storeu_epi32): Renamed into...
2192         (_mm512_storeu_si512): This.
2193         (_mm512_maskz_ceil_ps): Removed.
2194         (_mm512_maskz_ceil_pd): Ditto.
2195         (_mm512_maskz_floor_ps): Ditto.
2196         (_mm512_maskz_floor_pd): Ditto.
2197         (_mm512_floor_round_ps): Ditto.
2198         (_mm512_floor_round_pd): Ditto.
2199         (_mm512_ceil_round_ps): Ditto.
2200         (_mm512_ceil_round_pd): Ditto.
2201         (_mm512_mask_floor_round_ps): Ditto.
2202         (_mm512_mask_floor_round_pd): Ditto.
2203         (_mm512_mask_ceil_round_ps): Ditto.
2204         (_mm512_mask_ceil_round_pd): Ditto.
2205         (_mm512_maskz_floor_round_ps): Ditto.
2206         (_mm512_maskz_floor_round_pd): Ditto.
2207         (_mm512_maskz_ceil_round_ps): Ditto.
2208         (_mm512_maskz_ceil_round_pd): Ditto.
2209         (_mm512_expand_pd): Ditto.
2210         (_mm512_expand_ps): Ditto.
2211         * config/i386/i386.c (ix86_builtins): Remove
2212         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
2213         (bdesc_args): Ditto.
2214         * config/i386/predicates.md (const1256_operand): New.
2215         (const_1_to_2_operand): Ditto.
2216         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
2217         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
2218         (*avx512pf_gatherpf<mode>sf): Ditto.
2219         (avx512pf_gatherpf<mode>df): Ditto.
2220         (*avx512pf_gatherpf<mode>df_mask): Ditto.
2221         (*avx512pf_gatherpf<mode>df): Ditto.
2222         (avx512pf_scatterpf<mode>sf): Ditto.
2223         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2224         (*avx512pf_scatterpf<mode>sf): Ditto.
2225         (avx512pf_scatterpf<mode>df): Ditto.
2226         (*avx512pf_scatterpf<mode>df_mask): Ditto.
2227         (*avx512pf_scatterpf<mode>df): Ditto.
2228         (avx512f_expand<mode>): Removed.
2229         (<shift_insn><mode>3<mask_name>): Change predicate type.
2230
2231 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
2232
2233         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
2234         not at the end of datarefs vector use ordered_remove to avoid
2235         reordering datarefs vector.
2236
2237         PR c/59984
2238         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
2239         mark local addressable non-static vars as GOVD_PRIVATE
2240         instead of GOVD_LOCAL.
2241         * omp-low.c (lower_omp_for): Move gimple_bind_vars
2242         and BLOCK_VARS of gimple_bind_block to new_stmt rather
2243         than copying them.
2244
2245         PR middle-end/60092
2246         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
2247         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
2248         assume_aligned or alloc_align attributes.
2249         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
2250         arguments.  Handle also assume_aligned and alloc_align attributes.
2251         (evaluate_stmt): Adjust bit_value_assume_aligned caller.  Handle
2252         calls to functions with assume_aligned or alloc_align attributes.
2253         * doc/extend.texi: Document assume_aligned and alloc_align attributes.
2254
2255 2014-02-08  Terry Guo  <terry.guo@arm.com>
2256
2257         * doc/invoke.texi: Document ARM -march=armv7e-m.
2258
2259 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
2260
2261         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
2262         flag on __cilkrts_rethrow builtin.
2263
2264         PR ipa/60026
2265         * ipa-cp.c (determine_versionability): Fail at -O0
2266         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
2267         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
2268
2269         Revert:
2270         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
2271
2272         PR ipa/60026
2273         * tree-inline.c (copy_forbidden): Fail for
2274         __attribute__((optimize (0))) functions.
2275
2276 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
2277
2278         * varpool.c: Include pointer-set.h.
2279         (varpool_remove_unreferenced_decls): Variables in other partitions
2280         will not be output; be however careful to not lose information
2281         about partitioning.
2282
2283 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
2284
2285         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
2286         lookup in the vtable constructor.
2287
2288 2014-02-07  Jeff Law  <law@redhat.com>
2289
2290         PR target/40977
2291         * config/m68k/m68k.md (ashldi_extsi): Turn into a
2292         define_insn_and_split.
2293
2294         * ipa-inline.c (inline_small_functions): Fix typos.
2295
2296 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2297
2298         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
2299         (s390_can_use_return_insn): Declare.
2300         * config/s390/s390.h (EPILOGUE_USES): Define.
2301         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
2302         instructions.
2303         (s390_chunkify_start): Handle return JUMP_LABELs.
2304         (s390_early_mach): Emit a main_pool instruction on the entry edge.
2305         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
2306         (s390_can_use_return_insn): New functions.
2307         (s390_fix_long_loop_prediction): Handle conditional returns.
2308         (TARGET_SET_UP_BY_PROLOGUE): Define.
2309         * config/s390/s390.md (ANY_RETURN): New code iterator.
2310         (*creturn, *csimple_return, return, simple_return): New patterns.
2311
2312 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2313
2314         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
2315         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
2316         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
2317         REG_CFA_RESTORE list when deciding not to restore a register.
2318
2319 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2320
2321         * config/s390/s390.c: Include tree-pass.h and context.h.
2322         (s390_early_mach): New function, split out from...
2323         (s390_emit_prologue): ...here.
2324         (pass_data_s390_early_mach): New pass structure.
2325         (pass_s390_early_mach): New class.
2326         (s390_option_override): Create and register early_mach pass.
2327         Move to end of file.
2328
2329 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2330
2331         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
2332         to match for the exit block.
2333
2334 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2335
2336         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
2337         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
2338         Reject misaligned operands.
2339
2340 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2341
2342         * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
2343
2344 2014-02-07  Richard Biener  <rguenther@suse.de>
2345
2346         PR middle-end/60092
2347         * gimple-low.c (lower_builtin_posix_memalign): New function.
2348         (lower_stmt): Call it to lower posix_memalign in a way
2349         to make alignment info accessible.
2350
2351 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
2352
2353         PR c++/60082
2354         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
2355         __builtin_setjmp_receiver.
2356
2357 2014-02-07  Richard Biener  <rguenther@suse.de>
2358
2359         PR middle-end/60092
2360         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
2361         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
2362         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2363         Handle BUILT_IN_POSIX_MEMALIGN.
2364         (find_func_clobbers): Likewise.
2365         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
2366         (call_may_clobber_ref_p_1): Likewise.
2367
2368 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2369
2370         PR ipa/59918
2371         * ipa-devirt.c (record_target_from_binfo): Remove overactive
2372         sanity check.
2373
2374 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2375
2376         PR ipa/59469
2377         * lto-cgraph.c (lto_output_node): Use
2378         symtab_get_symbol_partitioning_class.
2379         (lto_output_varpool_node): likewise.
2380         (symtab_get_symbol_partitioning_class): Move here from
2381         lto/lto-partition.c
2382         * cgraph.h (symbol_partitioning_class): Likewise.
2383         (symtab_get_symbol_partitioning_class): Declare.
2384
2385 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2386
2387         * ggc.h (ggc_internal_cleared_alloc): New macro.
2388         * vec.h (vec_safe_copy): Handle memory stats.
2389         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
2390         * target-globals.c (save_target_globals): Likewise.
2391
2392 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2393
2394         PR target/60077
2395         * expr.c (emit_move_resolve_push): Export; be bit more selective
2396         on when to clear alias set.
2397         * expr.h (emit_move_resolve_push): Declare.
2398         * function.h (struct function): Add tail_call_marked.
2399         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
2400         * config/i386/i386-protos.h (ix86_expand_push): Remove.
2401         * config/i386/i386.md (TImode move expander): De not call
2402         ix86_expand_push.
2403         (FP push expanders): Preserve memory attributes.
2404         * config/i386/sse.md (push<mode>1): Remove.
2405         * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
2406         (ix86_expand_push): Remove.
2407         * config/i386/mmx.md (push<mode>1): Remove.
2408
2409 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2410
2411         PR rtl-optimization/60030
2412         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
2413         lopart with paradoxical subreg before shifting it up by hprec.
2414
2415 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2416
2417         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
2418         Remove extra newline at end of file.
2419         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
2420         (arm_issue_rate): Handle cortexa57.
2421         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
2422         (cortex-a57.cortex-a53): Likewise.
2423
2424 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2425
2426         PR target/59575
2427         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
2428         don't record in REG_FRAME_RELATED_EXPR registers not set in that
2429         bitmask.
2430         (arm_expand_prologue): Adjust all callers.
2431         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
2432         info, registers also at the lowest numbered registers side.  Use
2433         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
2434         XEXP.
2435
2436         PR debug/59992
2437         * var-tracking.c (adjust_mems): Before adding a SET to
2438         amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
2439
2440 2014-02-06  Alan Modra  <amodra@gmail.com>
2441
2442         PR target/60032
2443         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
2444         change SDmode to DDmode when lra_in_progress.
2445
2446 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2447
2448         PR middle-end/59150
2449         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
2450         free_data_ref on the dr first, and before goto again also set dr
2451         to the next dr.  For simd_lane_access, free old datarefs[i] before
2452         overwriting it.  For get_vectype_for_scalar_type failure, don't
2453         free_data_ref if simd_lane_access.
2454
2455         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
2456
2457         PR target/60062
2458         * tree.h (opts_for_fn): New inline function.
2459         (opt_for_fn): Define.
2460         * config/i386/i386.c (ix86_function_regparm): Use
2461         opt_for_fn (decl, optimize) instead of optimize.
2462
2463 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
2464
2465         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
2466         for SYMBOL_REF in large memory model.
2467
2468 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2469
2470         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
2471         and crypto support.
2472         (cortex-a57): Likewise.
2473         (cortex-a57.cortex-a53): Likewise.
2474
2475 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
2476             Kugan Vivekanandarajah  <kuganv@linaro.org>
2477
2478         * config/arm/arm.c (arm_vector_alignment_reachable): Check
2479         unaligned_access.
2480         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
2481
2482 2014-02-06  Richard Biener  <rguenther@suse.de>
2483
2484         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
2485         set_loop_copy and initialize_original_copy_tables.
2486
2487 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
2488
2489         * config/aarch64/aarch64-simd.md
2490         (aarch64_ashr_simddi): Change QI to SI.
2491
2492 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
2493             Jakub Jelinek  <jakub@redhat.com>
2494
2495         PR middle-end/60013
2496         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
2497         of the dataflow.
2498
2499 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2500
2501         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
2502         CODE_FOR_altivec_vpku[hw]um to
2503         CODE_FOR_altivec_vpku[hw]um_direct.
2504         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
2505         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
2506         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
2507         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
2508
2509 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2510
2511         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
2512         generation for -maltivec=be.
2513         (altivec_vsumsws): Simplify redundant test.
2514
2515 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2516
2517         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
2518         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
2519         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
2520         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
2521         gen_altivec_vpkuwum.
2522         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
2523         BYTES_BIG_ENDIAN.
2524         (altivec_vpks<VI_char>ss): Likewise.
2525         (altivec_vpks<VI_char>us): Likewise.
2526         (altivec_vpku<VI_char>us): Likewise.
2527         (altivec_vpku<VI_char>um): Likewise.
2528         (altivec_vpku<VI_char>um_direct): New (copy of
2529         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
2530         internal use).
2531         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
2532         target is little endian and -maltivec=be is not specified.
2533         (*altivec_vupkhs<VU_char>_direct): New (copy of
2534         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
2535         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
2536         target is little endian and -maltivec=be is not specified.
2537         (*altivec_vupkls<VU_char>_direct): New (copy of
2538         altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
2539         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
2540         little endian and -maltivec=be is not specified.
2541         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
2542         little endian and -maltivec=be is not specified.
2543
2544 2014-02-05  Richard Henderson  <rth@redhat.com>
2545
2546         PR debug/52727
2547         * combine-stack-adj.c: Revert r206943.
2548         * sched-int.h (struct deps_desc): Add last_args_size.
2549         * sched-deps.c (init_deps): Initialize it.
2550         (sched_analyze_insn): Add OUTPUT dependencies between insns that
2551         contain REG_ARGS_SIZE notes.
2552
2553 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
2554
2555         * lto-cgraph.c (asm_nodes_output): Make global.
2556         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
2557         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
2558         (driver_handle_option): Handle OPT_fwpa.
2559
2560 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
2561
2562         PR ipa/59947
2563         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
2564         a comment typo and formatting issue.  If odr_hash hasn't been
2565         created, return vNULL and set *completep to false.
2566
2567         PR middle-end/57499
2568         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
2569         bb with no successors.
2570
2571 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
2572
2573         PR target/59718
2574         * doc/invoke.texi (-march): Clarify documentation for ARM.
2575         (-mtune): Likewise.
2576         (-mcpu): Likewise.
2577
2578 2014-02-05  Richard Biener  <rguenther@suse.de>
2579
2580         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
2581         when not vectorizing because of too many alias checks.
2582         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
2583         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
2584
2585 2014-02-05  Nick Clifton  <nickc@redhat.com>
2586
2587         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
2588         accept extended registers in any mode when compiling for the MN10300.
2589
2590 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
2591
2592         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
2593         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
2594         sanitization attributes.
2595         (can_inline_edge_p): Likewise.
2596         (sanitize_attrs_match_for_inline_p): New function.
2597
2598 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2599
2600         * ipa-prop.c (detect_type_change): Shor circuit testing of
2601         type changes on THIS pointer.
2602
2603 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
2604
2605         PR target/59777
2606         * config/pa/pa.c (legitimize_tls_address): Return original address
2607         if not passed a SYMBOL_REF rtx.
2608         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
2609         addresses.
2610         (pa_emit_move_sequence): Simplify TLS source operands.
2611         (pa_legitimate_constant_p): Reject all TLS constants.
2612         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
2613         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
2614
2615 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2616
2617         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
2618         groups when we know they are controlled by LTO.
2619         * varasm.c (default_binds_local_p_1): If object is in other partition,
2620         it will be resolved locally.
2621
2622 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2623
2624         * config/host-linux.c (linux_gt_pch_use_address): Don't
2625         use SSIZE_MAX because it is not always defined.
2626
2627 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
2628
2629         PR bootstrap/59913
2630         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
2631         threshold for pseudo splitting.
2632         (update_ebb_live_info): Process call argument hard registers and
2633         hard registers from insn definition too.
2634         (max_small_class_regs_num): New constant.
2635         (inherit_in_ebb): Update live hard regs through EBBs.  Update
2636         reloads_num only for small register classes.  Don't split for
2637         outputs of jumps.
2638
2639 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
2640
2641         PR ipa/60058
2642         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
2643         is non-null.
2644
2645 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2646
2647         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
2648         visibility is safe.
2649
2650 2014-02-04  Marek Polacek  <polacek@redhat.com>
2651
2652         * gdbinit.in (pel): Define.
2653
2654 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2655
2656         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
2657         behavior.
2658
2659 2014-02-04  Richard Biener  <rguenther@suse.de>
2660
2661         PR lto/59723
2662         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
2663         in function context local.
2664         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
2665         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
2666         similar to LTO_imported_decl_ref.
2667
2668 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
2669
2670         PR tree-optimization/60002
2671         * cgraphclones.c (build_function_decl_skip_args): Clear
2672         DECL_LANG_SPECIFIC.
2673
2674         PR tree-optimization/60023
2675         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
2676         false to gsi_replace.
2677         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
2678         has been in some EH region and vec_stmt could throw, add
2679         vec_stmt into the same EH region.
2680         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
2681         has no lhs, ignore it.
2682         * internal-fn.c (expand_MASK_LOAD): Likewise.
2683
2684         PR ipa/60026
2685         * tree-inline.c (copy_forbidden): Fail for
2686         __attribute__((optimize (0))) functions.
2687
2688         PR other/58712
2689         * omp-low.c (simd_clone_struct_copy): If from->inbranch
2690         is set, copy one less argument.
2691         (expand_simd_clones): Don't subtract clone_info->inbranch
2692         from simd_clone_struct_alloc argument.
2693
2694         PR rtl-optimization/57915
2695         * recog.c (simplify_while_replacing): If all unary/binary/relational
2696         operation arguments are constant, attempt to simplify those.
2697
2698         PR middle-end/59261
2699         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
2700         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
2701
2702 2014-02-04  Richard Biener  <rguenther@suse.de>
2703
2704         PR tree-optimization/60012
2705         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
2706         TBAA disambiguation to all DDRs.
2707
2708 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2709
2710         PR target/59788
2711         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
2712         (LINK_SPEC): Use it for -shared, -shared-libgcc.
2713
2714 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2715
2716         PR ipa/59882
2717         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
2718
2719 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2720
2721         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
2722         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
2723
2724 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2725
2726         PR ipa/59831
2727         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
2728         to figure out targets of polymorphic calls with known decl.
2729         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2730         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
2731         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
2732         (get_polymorphic_call_info): ... here.
2733         (get_polymorphic_call_info_from_invariant): New function.
2734
2735 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2736
2737         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
2738         lookup via vtable pointer; check for type consistency
2739         and turn inconsitent facts into UNREACHABLE.
2740         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2741         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
2742         type inconsistent querries; return UNREACHABLE instead.
2743
2744 2014-02-03  Richard Henderson  <rth@twiddle.net>
2745
2746         PR tree-opt/59924
2747         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
2748         already processed this node.
2749         (normalize_one_pred_1): Pass along mark_set.
2750         (normalize_one_pred): Create and destroy a pointer_set_t.
2751         (normalize_one_pred_chain): Likewise.
2752
2753 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
2754
2755         PR gcov-profile/58602
2756         * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
2757
2758 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2759
2760         PR ipa/59831
2761         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
2762         -fno-devirtualize; try to devirtualize by the knowledge of
2763         virtual table pointer given by aggregate propagation.
2764         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2765         (ipa_print_node_jump_functions): Dump also offset that
2766         is relevant for polymorphic calls.
2767         (determine_known_aggregate_parts): Add arg_type parameter; use it
2768         instead of determining the type from pointer type.
2769         (ipa_compute_jump_functions_for_edge): Update call of
2770         determine_known_aggregate_parts.
2771         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
2772         (gimple_get_virt_method_for_binfo): ... here; simplify using
2773         vtable_pointer_value_to_vtable.
2774         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
2775         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
2776         to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
2777         (vtable_pointer_value_to_vtable): Break out from ...; handle also
2778         POINTER_PLUS_EXPR.
2779         (vtable_pointer_value_to_binfo): ... here.
2780         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
2781
2782 2014-02-03  Teresa Johnson  <tejohnson@google.com>
2783
2784         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
2785         redef of outer loop index variable.
2786
2787 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
2788
2789         PR c++/53017
2790         PR c++/59211
2791         * doc/extend.texi (Function Attributes): Typo.
2792
2793 2014-02-03  Cong Hou  <congh@google.com>
2794
2795         PR tree-optimization/60000
2796         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
2797         if the vectorized statement is a store.  A store statement can only
2798         appear at the end of pattern statements.
2799
2800 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
2801
2802         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
2803         (ix86_option_override_internal): Default long double to 64-bit for
2804         32-bit Bionic and to 128-bit for 64-bit Bionic.
2805
2806         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
2807         TARGET_LONG_DOUBLE_128 is true.
2808         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2809
2810         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
2811         (mlong-double-64): Negate -mlong-double-128.
2812         (mlong-double-128): New option.
2813
2814         * config/i386/i386-c.c (ix86_target_macros): Define
2815         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
2816
2817         * doc/invoke.texi: Document -mlong-double-128.
2818
2819 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
2820
2821         PR rtl-optimization/60024
2822         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
2823
2824 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
2825
2826         * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
2827
2828 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
2829
2830         PR rtl-optimization/57662
2831         * sel-sched.c (code_motion_path_driver): Do not mark already not
2832         existing blocks in the visiting bitmap.
2833
2834 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
2835
2836         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
2837         on the insn being emitted.
2838
2839 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
2840             Will Deacon  <will.deacon@arm.com>
2841
2842         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
2843
2844 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2845
2846         * config/arm/arm-tables.opt: Regenerate.
2847
2848 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2849
2850         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
2851         for vector types other than V16QImode.
2852         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
2853         define_expand, and call altivec_expand_vec_perm_le when producing
2854         code with little endian element order.
2855         (*altivec_vperm_<mode>_internal): New insn having previous
2856         behavior of altivec_vperm_<mode>.
2857         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
2858         altivec_expand_vec_perm_le when producing code with little endian
2859         element order.
2860         (*altivec_vperm_<mode>_uns_internal): New insn having previous
2861         behavior of altivec_vperm_<mode>_uns.
2862
2863 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2864
2865         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
2866         (altivec_vsumsws): Add handling for -maltivec=be with a little
2867         endian target.
2868         (altivec_vsumsws_direct): New.
2869         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
2870         gen_altivec_vsumsws.
2871
2872 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
2873
2874         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
2875         vtable_pointer_value_to_binfo): New functions.
2876         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
2877         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
2878
2879 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
2880
2881         * config/nios2/nios2.md (load_got_register): Initialize GOT
2882         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
2883         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
2884
2885 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
2886
2887         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
2888         preserverd by passthrough, do not propagate the type.
2889
2890 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2891
2892         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
2893         (mips_atomic_assign_expand_fenv): New function.
2894         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
2895
2896 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2897
2898         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
2899         (__builtin_mips_set_fcsr): Likewise.
2900         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
2901         MIPS_USI_FTYPE_VOID.
2902         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
2903         (mips16_expand_set_fcsr): Likewise.
2904         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
2905         (mips16_set_fcsr_stub): Likewise.
2906         (mips16_get_fcsr_one_only_stub): New class.
2907         (mips16_set_fcsr_one_only_stub): Likewise.
2908         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
2909         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
2910         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
2911         (hard_float): New availability predicate.
2912         (mips_builtins): Add get_fcsr and set_fcsr.
2913         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
2914         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
2915         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
2916         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
2917         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
2918         patterns.
2919
2920 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2921
2922         * config/mips/mips.c (mips_one_only_stub): New class.
2923         (mips_need_mips16_rdhwr_p): Replace with...
2924         (mips16_rdhwr_stub): ...this new variable.
2925         (mips16_stub_call_address): New function.
2926         (mips16_rdhwr_one_only_stub): New class.
2927         (mips_expand_thread_pointer): Use mips16_stub_call_address.
2928         (mips_output_mips16_rdhwr): Delete.
2929         (mips_finish_stub): New function.
2930         (mips_code_end): Use it to handle rdhwr stubs.
2931
2932 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
2933
2934         PR target/60017
2935         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
2936         when calculating size of integer atomic types.
2937
2938 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
2939
2940         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
2941
2942 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
2943
2944         PR tree-optimization/60003
2945         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
2946         * profile.c (branch_prob): Use gimple_call_builtin_p
2947         to check for BUILT_IN_SETJMP_RECEIVER.
2948         * tree-inline.c (copy_bb): Call notice_special_calls.
2949
2950 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
2951
2952         PR bootstrap/59985
2953         * lra-constraints.c (process_alt_operands): Update reload_sum only
2954         on the first pass.
2955
2956 2014-01-31  Richard Henderson  <rth@redhat.com>
2957
2958         PR middle-end/60004
2959         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
2960         until after else_eh is processed.
2961
2962 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
2963
2964         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
2965         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
2966         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
2967         in smmintrin.h, remove them.
2968         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
2969         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
2970         * config/i386/i386.md (ROUND_SAE): Fix value.
2971         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
2972         (const48_operand): New.
2973         * config/i386/subst.md (round), (round_expand): Use
2974         const_4_or_8_to_11_operand.
2975         (round_saeonly), (round_saeonly_expand): Use const48_operand.
2976
2977 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
2978
2979         * config/i386/constraints.md (Yk): Swap meaning with k.
2980         * config/i386/i386.md (movhi_internal): Change Yk to k.
2981         (movqi_internal): Ditto.
2982         (*k<logic><mode>): Ditto.
2983         (*andhi_1): Ditto.
2984         (*andqi_1): Ditto.
2985         (kandn<mode>): Ditto.
2986         (*<code>hi_1): Ditto.
2987         (*<code>qi_1): Ditto.
2988         (kxnor<mode>): Ditto.
2989         (kortestzhi): Ditto.
2990         (kortestchi): Ditto.
2991         (kunpckhi): Ditto.
2992         (*one_cmplhi2_1): Ditto.
2993         (*one_cmplqi2_1): Ditto.
2994         * config/i386/sse.md (): Change k to Yk.
2995         (avx512f_load<mode>_mask): Ditto.
2996         (avx512f_blendm<mode>): Ditto.
2997         (avx512f_store<mode>_mask): Ditto.
2998         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
2999         (avx512f_storedqu<mode>_mask): Ditto.
3000         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
3001         Ditto.
3002         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
3003         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
3004         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
3005         (avx512f_maskcmp<mode>3): Ditto.
3006         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
3007         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
3008         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
3009         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
3010         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
3011         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
3012         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
3013         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
3014         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
3015         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
3016         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
3017         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
3018         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
3019         (vec_extract_lo_<mode>_maskm): Ditto.
3020         (vec_extract_hi_<mode>_maskm): Ditto.
3021         (avx512f_vternlog<mode>_mask): Ditto.
3022         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
3023         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
3024         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
3025         (avx512f_<code>v8div16qi2_mask): Ditto.
3026         (avx512f_<code>v8div16qi2_mask_store): Ditto.
3027         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
3028         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
3029         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
3030         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
3031         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
3032         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3033         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
3034         (*avx512pf_scatterpf<mode>df_mask): Ditto.
3035         (avx512cd_maskb_vec_dupv8di): Ditto.
3036         (avx512cd_maskw_vec_dupv16si): Ditto.
3037         (avx512f_vpermi2var<mode>3_maskz): Ditto.
3038         (avx512f_vpermi2var<mode>3_mask): Ditto.
3039         (avx512f_vpermi2var<mode>3_mask): Ditto.
3040         (avx512f_vpermt2var<mode>3_maskz): Ditto.
3041         (*avx512f_gathersi<mode>): Ditto.
3042         (*avx512f_gathersi<mode>_2): Ditto.
3043         (*avx512f_gatherdi<mode>): Ditto.
3044         (*avx512f_gatherdi<mode>_2): Ditto.
3045         (*avx512f_scattersi<mode>): Ditto.
3046         (*avx512f_scatterdi<mode>): Ditto.
3047         (avx512f_compress<mode>_mask): Ditto.
3048         (avx512f_compressstore<mode>_mask): Ditto.
3049         (avx512f_expand<mode>_mask): Ditto.
3050         * config/i386/subst.md (mask): Change k to Yk.
3051         (mask_scalar_merge): Ditto.
3052         (sd): Ditto.
3053
3054 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
3055
3056         * doc/extend.texi (Vector Extensions): Document ?: in C++.
3057
3058 2014-01-31  Richard Biener  <rguenther@suse.de>
3059
3060         PR middle-end/59990
3061         * builtins.c (fold_builtin_memory_op): Make sure to not
3062         use a floating-point mode or a boolean or enumeral type for
3063         the copy operation.
3064
3065 2014-01-30  DJ Delorie  <dj@redhat.com>
3066
3067         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
3068         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
3069         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
3070         whenever main() has an epilogue.
3071
3072 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3073
3074         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
3075         unused variable "field".
3076         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
3077         (vsx_mergeh_<mode>): Likewise.
3078         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
3079         (altivec_vmrghh): Likewise.
3080         (altivec_vmrghw): Likewise.
3081         (altivec_vmrglb): Likewise.
3082         (altivec_vmrglh): Likewise.
3083         (altivec_vmrglw): Likewise.
3084         (altivec_vspltb): Add missing uses.
3085         (altivec_vsplth): Likewise.
3086         (altivec_vspltw): Likewise.
3087         (altivec_vspltsf): Likewise.
3088
3089 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
3090
3091         PR target/59923
3092         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
3093         frame related instructions.
3094
3095 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
3096
3097         PR rtl-optimization/59959
3098         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
3099         any reload of register whose subreg is invalid.
3100
3101 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
3102
3103         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
3104         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
3105         Add missing return type - void.
3106
3107 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3108
3109         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
3110         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
3111         remove element index adjustment for endian (now handled in vsx.md
3112         and altivec.md).
3113         (altivec_expand_vec_perm_const): Use
3114         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
3115         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
3116         (vsx_xxspltw_<mode>): Adjust element index for little endian.
3117         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
3118         define_expand and a new define_insn *altivec_vspltb_internal;
3119         adjust for -maltivec=be on a little endian target.
3120         (altivec_vspltb_direct): New.
3121         (altivec_vsplth): Divide into a define_expand and a new
3122         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
3123         little endian target.
3124         (altivec_vsplth_direct): New.
3125         (altivec_vspltw): Divide into a define_expand and a new
3126         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
3127         little endian target.
3128         (altivec_vspltw_direct): New.
3129         (altivec_vspltsf): Divide into a define_expand and a new
3130         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
3131         a little endian target.
3132
3133 2014-01-30  Richard Biener  <rguenther@suse.de>
3134
3135         PR tree-optimization/59993
3136         * tree-ssa-forwprop.c (associate_pointerplus): Check we
3137         can propagate form the earlier stmt and avoid the transform
3138         when the intermediate result is needed.
3139
3140 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
3141
3142         * README.Portability: Fix typo.
3143
3144 2014-01-30  David Holsgrove <david.holsgrove@xilinx.com>
3145
3146         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
3147         comparison_operator with ordered_comparison_operator.
3148
3149 2014-01-30  Nick Clifton  <nickc@redhat.com>
3150
3151         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
3152         Rename to mn10300_store_multiple_regs.
3153         * config/mn10300/mn10300.c: Likewise.
3154         * config/mn10300/mn10300.md (store_movm): Fix typo: call
3155         store_multiple_regs.
3156         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
3157         Call mn10300_store_multiple_regs.
3158
3159 2014-01-30  Nick Clifton  <nickc@redhat.com>
3160             DJ Delorie  <dj@redhat.com>
3161
3162         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
3163         %fp 2 to keep registers after it properly word-aligned.
3164         (rl78_alloc_physical_registers_umul): Handle the case where both
3165         input operands are the same.
3166
3167 2014-01-30  Richard Biener  <rguenther@suse.de>
3168
3169         PR tree-optimization/59903
3170         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
3171         check properly.
3172
3173 2014-01-30  Jason Merrill  <jason@redhat.com>
3174
3175         PR c++/59633
3176         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
3177
3178         PR c++/59645
3179         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
3180
3181 2014-01-30  Richard Biener  <rguenther@suse.de>
3182
3183         PR tree-optimization/59951
3184         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
3185
3186 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
3187
3188         PR target/59784
3189         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
3190         SFmode to DFmode case.
3191
3192 2014-01-29  DJ Delorie  <dj@redhat.com>
3193
3194         * config/msp430/msp430.opt (-minrt): New.
3195         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
3196         if -minrt given.
3197         (ENDFILE_SPEC): Likewise.
3198
3199 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
3200
3201         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
3202         (estimate_function_body_sizes): Use it.
3203
3204 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
3205
3206         PR c++/58561
3207         * dwarf2out.c (is_cxx_auto): New.
3208         (is_base_type): Use it.
3209         (gen_type_die_with_usage): Likewise.
3210
3211 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3212
3213         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
3214         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
3215         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
3216         -maltivec=be with LE targets.
3217         (vsx_mergeh_<mode>): Likewise.
3218         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
3219         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
3220         (altivec_vmrghb): Replace with define_expand and new
3221         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
3222         (altivec_vmrghb_direct): New define_insn.
3223         (altivec_vmrghh): Replace with define_expand and new
3224         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
3225         (altivec_vmrghh_direct): New define_insn.
3226         (altivec_vmrghw): Replace with define_expand and new
3227         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
3228         (altivec_vmrghw_direct): New define_insn.
3229         (*altivec_vmrghsf): Adjust for endianness.
3230         (altivec_vmrglb): Replace with define_expand and new
3231         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
3232         (altivec_vmrglb_direct): New define_insn.
3233         (altivec_vmrglh): Replace with define_expand and new
3234         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
3235         (altivec_vmrglh_direct): New define_insn.
3236         (altivec_vmrglw): Replace with define_expand and new
3237         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
3238         (altivec_vmrglw_direct): New define_insn.
3239         (*altivec_vmrglsf): Adjust for endianness.
3240         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
3241         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
3242         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
3243         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
3244         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
3245         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
3246         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
3247         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
3248
3249 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
3250
3251         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
3252         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
3253         whitespace.
3254
3255 2014-01-29  Richard Biener  <rguenther@suse.de>
3256
3257         PR tree-optimization/58742
3258         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
3259         associate_pointerplus_align.
3260         (associate_pointerplus_diff): New function.
3261         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
3262         and associate_pointerplus_diff.
3263
3264 2014-01-29  Richard Biener  <rguenther@suse.de>
3265
3266         * lto-streamer.h (LTO_major_version): Bump to 3.
3267         (LTO_minor_version): Reset to 0.
3268
3269 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
3270
3271         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
3272         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
3273         (arm_file_start): Generate correct asm header for armv7ve.
3274         * config/arm/bpabi.h: Add multilib support for armv7ve.
3275         * config/arm/driver-arm.c: Change the architectures of cortex-a7
3276         and cortex-a15 to armv7ve.
3277         * config/arm/t-aprofile: Add multilib support for armv7ve.
3278         * doc/invoke.texi: Document -march=armv7ve.
3279
3280 2014-01-29  Richard Biener  <rguenther@suse.de>
3281
3282         PR tree-optimization/58742
3283         * tree-ssa-forwprop.c (associate_plusminus): Return true
3284         if we changed sth, defer EH cleanup to ...
3285         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
3286         (simplify_mult): New function.
3287
3288 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
3289
3290         PR middle-end/59917
3291         PR tree-optimization/59920
3292         * tree.c (build_common_builtin_nodes): Remove
3293         __builtin_setjmp_dispatcher initialization.
3294         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
3295         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
3296         instead of gsi_after_labels + manually skipping debug stmts.
3297         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
3298         ignore bbs with IFN_ABNORMAL_DISPATCHER.
3299         * tree-inline.c (copy_edges_for_bb): Remove
3300         can_make_abnormal_goto argument, instead add abnormal_goto_dest
3301         argument.  Ignore computed_goto_p stmts.  Don't call
3302         make_abnormal_goto_edges.  If a call might need abnormal edges
3303         for non-local gotos, see if it already has an edge to
3304         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
3305         with true argument, don't do anything then, otherwise add
3306         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
3307         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
3308         caller.
3309         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
3310         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
3311         (lower_stmt): Don't set data->calls_builtin_setjmp.
3312         (lower_builtin_setjmp): Adjust comment.
3313         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
3314         * tree-cfg.c (found_computed_goto): Remove.
3315         (factor_computed_gotos): Remove.
3316         (make_goto_expr_edges): Return bool, true for computed gotos.
3317         Don't call make_abnormal_goto_edges.
3318         (build_gimple_cfg): Don't set found_computed_goto, don't call
3319         factor_computed_gotos.
3320         (computed_goto_p): No longer static.
3321         (make_blocks): Don't set found_computed_goto.
3322         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
3323         (make_edges): If make_goto_expr_edges returns true, push bb
3324         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
3325         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
3326         vector.  Record mapping between bbs and OpenMP regions if there
3327         are any, adjust make_gimple_omp_edges caller.  Call
3328         handle_abnormal_edges.
3329         (make_abnormal_goto_edges): Remove.
3330         * tree-cfg.h (make_abnormal_goto_edges): Remove.
3331         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
3332         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
3333         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
3334         * internal-fn.def (ABNORMAL_DISPATCHER): New.
3335         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
3336         filling *region also set *region_idx to (*region)->entry->index.
3337
3338         PR other/58712
3339         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
3340         For REGs set ORIGINAL_REGNO.
3341
3342 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
3343
3344         * doc/md.texi: Mention that a target shouldn't implement
3345         vec_widen_(s|u)mul_even/odd pair if it is less efficient
3346         than hi/lo pair.
3347
3348 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
3349
3350         PR tree-optimization/59594
3351         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
3352         a copy of the datarefs vector rather than the vector itself.
3353
3354 2014-01-28  Jason Merrill  <jason@redhat.com>
3355
3356         PR c++/53756
3357         * dwarf2out.c (auto_die): New static.
3358         (gen_type_die_with_usage): Handle C++1y 'auto'.
3359         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
3360         on definition.
3361
3362 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
3363
3364         PR target/59672
3365         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
3366         (SPEC_X32): Likewise.
3367         (SPEC_64): Likewise.
3368         * config/i386/i386.c (ix86_option_override_internal): Turn off
3369         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
3370         for TARGET_16BIT.
3371         (x86_file_start): Output .code16gcc for TARGET_16BIT.
3372         * config/i386/i386.h (TARGET_16BIT): New macro.
3373         (TARGET_16BIT_P): Likewise.
3374         * config/i386/i386.opt: Add m16.
3375         * doc/invoke.texi: Document -m16.
3376
3377 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
3378
3379         PR preprocessor/59935
3380         * input.c (location_get_source_line): Bail out on when line number
3381         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
3382
3383 2014-01-28  Richard Biener  <rguenther@suse.de>
3384
3385         PR tree-optimization/58742
3386         * tree-ssa-forwprop.c (associate_plusminus): Handle
3387         pointer subtraction of the form (T)(P + A) - (T)P.
3388
3389 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3390
3391         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
3392         at const_int_cost.
3393
3394 2014-01-28  Richard Biener  <rguenther@suse.de>
3395
3396         Revert
3397         2014-01-28  Richard Biener  <rguenther@suse.de>
3398
3399         PR rtl-optimization/45364
3400         PR rtl-optimization/59890
3401         * var-tracking.c (local_get_addr_clear_given_value): Handle
3402         already cleared slot.
3403         (val_reset): Handle not allocated local_get_addr_cache.
3404         (vt_find_locations): Use post-order on the inverted CFG.
3405
3406 2014-01-28  Richard Biener  <rguenther@suse.de>
3407
3408         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
3409
3410 2014-01-28  Richard Biener  <rguenther@suse.de>
3411
3412         PR rtl-optimization/45364
3413         PR rtl-optimization/59890
3414         * var-tracking.c (local_get_addr_clear_given_value): Handle
3415         already cleared slot.
3416         (val_reset): Handle not allocated local_get_addr_cache.
3417         (vt_find_locations): Use post-order on the inverted CFG.
3418
3419 2014-01-28  Alan Modra  <amodra@gmail.com>
3420
3421         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
3422         * configure.ac <recursive call for build != host>: Define
3423         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
3424         and LD_FOR_BUILD too.
3425         * configure: Regenerate.
3426
3427 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
3428
3429         * config/i386/i386.c (get_builtin_code_for_version): Separate
3430         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
3431         Broadwell from Haswell.
3432
3433 2014-01-27  Steve Ellcey  <sellcey@mips.com>
3434
3435         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3436         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
3437         * config/mips/mips.c (mips_option_override): Change setting
3438         of TARGET_DSP.
3439         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
3440         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
3441         Change from Mask to Var.
3442
3443 2014-01-27  Jeff Law  <law@redhat.com>
3444
3445         * ipa-inline.c (inline_small_functions): Fix typo.
3446
3447 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
3448
3449         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
3450         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
3451         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
3452         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
3453         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
3454         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
3455         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
3456         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
3457         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
3458         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
3459         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
3460         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
3461         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
3462         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
3463         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
3464         (_mm512_storeu_epi64): Ditto.
3465         (_mm512_cmpge_epi32_mask): Ditto.
3466         (_mm512_cmpge_epu32_mask): Ditto.
3467         (_mm512_cmpge_epi64_mask): Ditto.
3468         (_mm512_cmpge_epu64_mask): Ditto.
3469         (_mm512_cmple_epi32_mask): Ditto.
3470         (_mm512_cmple_epu32_mask): Ditto.
3471         (_mm512_cmple_epi64_mask): Ditto.
3472         (_mm512_cmple_epu64_mask): Ditto.
3473         (_mm512_cmplt_epi32_mask): Ditto.
3474         (_mm512_cmplt_epu32_mask): Ditto.
3475         (_mm512_cmplt_epi64_mask): Ditto.
3476         (_mm512_cmplt_epu64_mask): Ditto.
3477         (_mm512_cmpneq_epi32_mask): Ditto.
3478         (_mm512_cmpneq_epu32_mask): Ditto.
3479         (_mm512_cmpneq_epi64_mask): Ditto.
3480         (_mm512_cmpneq_epu64_mask): Ditto.
3481         (_mm512_expand_pd): Ditto.
3482         (_mm512_expand_ps): Ditto.
3483         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
3484         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
3485         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
3486         * config/i386/i386.c (ix86_builtins): Add
3487         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
3488         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
3489         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
3490         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
3491         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
3492         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
3493         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
3494         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
3495         IX86_BUILTIN_PMOVUSQW512_MEM.
3496         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
3497         __builtin_ia32_pmovsqd512mem_mask,
3498         __builtin_ia32_pmovqd512mem_mask,
3499         __builtin_ia32_pmovusqw512mem_mask,
3500         __builtin_ia32_pmovsqw512mem_mask,
3501         __builtin_ia32_pmovqw512mem_mask,
3502         __builtin_ia32_pmovusdw512mem_mask,
3503         __builtin_ia32_pmovsdw512mem_mask,
3504         __builtin_ia32_pmovdw512mem_mask,
3505         __builtin_ia32_pmovqb512mem_mask,
3506         __builtin_ia32_pmovusqb512mem_mask,
3507         __builtin_ia32_pmovsqb512mem_mask,
3508         __builtin_ia32_pmovusdb512mem_mask,
3509         __builtin_ia32_pmovsdb512mem_mask,
3510         __builtin_ia32_pmovdb512mem_mask.
3511         (bdesc_args): Add __builtin_ia32_expanddf512,
3512         __builtin_ia32_expandsf512.
3513         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
3514         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
3515         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
3516         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
3517         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
3518         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
3519         (avx512f_<code>v8div16qi2_mask_store): This.
3520         (avx512f_expand<mode>): New.
3521
3522 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
3523
3524         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
3525         New.
3526         (_mm512_mask_prefetch_i64gather_pd): Ditto.
3527         (_mm512_prefetch_i32scatter_pd): Ditto.
3528         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
3529         (_mm512_prefetch_i64scatter_pd): Ditto.
3530         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
3531         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
3532         (_mm512_mask_prefetch_i64gather_ps): Ditto.
3533         (_mm512_prefetch_i32scatter_ps): Ditto.
3534         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
3535         (_mm512_prefetch_i64scatter_ps): Ditto.
3536         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
3537         * config/i386/i386-builtin-types.def: Define
3538         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
3539         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
3540         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
3541         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
3542         IX86_BUILTIN_SCATTERPFQPD.
3543         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
3544         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
3545         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
3546         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
3547         __builtin_ia32_scatterpfqps.
3548         (ix86_expand_builtin): Expand new built-ins.
3549         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
3550         fix memory access data type.
3551         (*avx512pf_gatherpf<mode>_mask): Ditto.
3552         (*avx512pf_gatherpf<mode>): Ditto.
3553         (avx512pf_scatterpf<mode>): Ditto.
3554         (*avx512pf_scatterpf<mode>_mask): Ditto.
3555         (*avx512pf_scatterpf<mode>): Ditto.
3556         (GATHER_SCATTER_SF_MEM_MODE): New.
3557         (avx512pf_gatherpf<mode>df): Ditto.
3558         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3559         (*avx512pf_scatterpf<mode>df): Ditto.
3560
3561 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
3562
3563         PR bootstrap/59934
3564         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
3565         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
3566         reached.
3567
3568 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
3569
3570         * common/config/arm/arm-common.c
3571         (arm_rewrite_mcpu): Handle multiple names.
3572         * config/arm/arm.h
3573         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
3574
3575 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
3576
3577         * gimple-builder.h (create_gimple_tmp): Delete.
3578
3579 2014-01-27  Christian Bruel  <christian.bruel@st.com>
3580
3581         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
3582         words comparisons.
3583
3584 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
3585
3586         * config/pa/pa.md (call): Generate indirect long calls to non-local
3587         functions when outputing 32-bit code.
3588         (call_value): Likewise except for special call to buggy powf function.
3589
3590         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
3591         portable runtime and PIC indirect calls.
3592         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
3593         and PIC call sequences.  Use ldo instead of blr to set return register
3594         in PIC call sequence.
3595
3596 2014-01-25  Walter Lee  <walt@tilera.com>
3597
3598         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
3599         avoid clobbering a live register.
3600
3601 2014-01-25  Walter Lee  <walt@tilera.com>
3602
3603         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
3604         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
3605         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
3606         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
3607
3608 2014-01-25  Walter Lee  <walt@tilera.com>
3609
3610         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
3611         arguments on even registers.
3612         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
3613         STACK_BOUNDARY.
3614         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
3615         (BIGGEST_ALIGNMENT): Ditto.
3616         (BIGGEST_FIELD_ALIGNMENT): Ditto.
3617
3618 2014-01-25  Walter Lee  <walt@tilera.com>
3619
3620         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
3621         insns before bundling.
3622         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
3623
3624 2014-01-25  Walter Lee  <walt@tilera.com>
3625
3626         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
3627         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
3628         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
3629
3630 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
3631
3632         * config/mips/constraints.md (kl): Delete.
3633         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
3634         define expands, using...
3635         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
3636         instructions for MIPS16.
3637         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
3638         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
3639
3640 2014-01-25  Walter Lee  <walt@tilera.com>
3641
3642         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
3643         (clzdi2): Ditto.
3644         (ffsdi2): Ditto.
3645
3646 2014-01-25  Walter Lee  <walt@tilera.com>
3647
3648         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
3649         (TARGET_EXPAND_TO_RTL_HOOK): Define.
3650
3651 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
3652
3653         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
3654         Handle XOR.
3655
3656 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
3657
3658         * print-rtl.c (in_call_function_usage): New var.
3659         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
3660         EXPR_LIST mode as mode and not as reg note name.
3661
3662         PR middle-end/59561
3663         * cfgloopmanip.c (copy_loop_info): If
3664         loop->warned_aggressive_loop_optimizations, make sure
3665         the flag is set in target loop too.
3666
3667 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
3668
3669         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
3670         flag_cilkplus.
3671         * builtins.def: Likewise.
3672         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
3673         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
3674         * ira.c (ira_setup_eliminable_regset): Likewise.
3675         * omp-low.c (gate_expand_omp): Likewise.
3676         (execute_lower_omp): Likewise.
3677         (diagnose_sb_0): Likewise.
3678         (gate_diagnose_omp_blocks): Likewise.
3679         (simd_clone_clauses_extract): Likewise.
3680         (gate): Likewise.
3681
3682 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3683
3684         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
3685         correction for little endian...
3686         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
3687         here.
3688
3689 2014-01-24  Jeff Law  <law@redhat.com>
3690
3691         PR tree-optimization/59919
3692         * tree-vrp.c (find_assert_locations_1): Do not register asserts
3693         for non-returning calls.
3694
3695 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
3696
3697         * common/config/aarch64/aarch64-common.c
3698         (aarch64_rewrite_mcpu): Handle multiple names.
3699         * config/aarch64/aarch64.h
3700         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
3701
3702 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
3703
3704         * input.c (add_file_to_cache_tab): Handle the case where fopen
3705         returns NULL.
3706
3707 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
3708
3709         PR target/59929
3710         * config/i386/i386.md (pushsf splitter): Get stack adjustment
3711         from push operand if code of push isn't PRE_DEC.
3712
3713 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
3714
3715         PR target/59909
3716         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3717         -mquad-memory-atomic.  Update -mquad-memory documentation to say
3718         it is only used for non-atomic loads/stores.
3719
3720         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
3721         -mquad-memory or -mquad-memory-atomic switches.
3722
3723         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
3724         -mquad-memory-atomic to ISA 2.07 support.
3725
3726         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
3727         to separate support of normal quad word memory operations (ldq, stq)
3728         from the atomic quad word memory operations.
3729
3730         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
3731         support to separate non-atomic quad word operations from atomic
3732         quad word operations.  Disable non-atomic quad word operations in
3733         little endian mode so that we don't have to swap words after the
3734         load and before the store.
3735         (quad_load_store_p): Add comment about atomic quad word support.
3736         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
3737         options printed with -mdebug=reg.
3738
3739         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
3740         -mquad-memory-atomic as the test for whether we have quad word
3741         atomic instructions.
3742         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
3743         or -mp8-vector are used, allow byte/half-word atomic operations.
3744
3745         * config/rs6000/sync.md (load_lockedti): Insure that the address
3746         is a proper indexed or indirect address for the lqarx instruction.
3747         On little endian systems, swap the hi/lo registers after the lqarx
3748         instruction.
3749         (load_lockedpti): Use indexed_or_indirect_operand predicate to
3750         insure the address is valid for the lqarx instruction.
3751         (store_conditionalti): Insure that the address is a proper indexed
3752         or indirect address for the stqcrx. instruction.  On little endian
3753         systems, swap the hi/lo registers before doing the stqcrx.
3754         instruction.
3755         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
3756         insure the address is valid for the stqcrx. instruction.
3757
3758         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
3759         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
3760         type of quad memory support is available.
3761
3762 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
3763
3764         PR regression/59915
3765         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
3766         there is a danger of looping.
3767
3768 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
3769
3770         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3771         force flag_ira_loop_pressure if set via command line.
3772
3773 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3774
3775         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
3776         (ashr_simd): New builtin handling DI mode.
3777         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
3778         (aarch64_sshr_simddi): New match pattern.
3779         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
3780         (vshrd_n_s64): Likewise.
3781         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
3782
3783 2014-01-23  Nick Clifton  <nickc@redhat.com>
3784
3785         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
3786         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
3787         favour of mcu specific scripts.
3788         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
3789         430x multilibs.
3790
3791 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
3792             Alex Velenko  <Alex.Velenko@arm.com>
3793
3794         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
3795         (vaddv_s16): Likewise.
3796         (vaddv_s32): Likewise.
3797         (vaddv_u8): Likewise.
3798         (vaddv_u16): Likewise.
3799         (vaddv_u32): Likewise.
3800         (vaddvq_s8): Likewise.
3801         (vaddvq_s16): Likewise.
3802         (vaddvq_s32): Likewise.
3803         (vaddvq_s64): Likewise.
3804         (vaddvq_u8): Likewise.
3805         (vaddvq_u16): Likewise.
3806         (vaddvq_u32): Likewise.
3807         (vaddvq_u64): Likewise.
3808         (vaddv_f32): Likewise.
3809         (vaddvq_f32): Likewise.
3810         (vaddvq_f64): Likewise.
3811         (vmaxv_f32): Likewise.
3812         (vmaxv_s8): Likewise.
3813         (vmaxv_s16): Likewise.
3814         (vmaxv_s32): Likewise.
3815         (vmaxv_u8): Likewise.
3816         (vmaxv_u16): Likewise.
3817         (vmaxv_u32): Likewise.
3818         (vmaxvq_f32): Likewise.
3819         (vmaxvq_f64): Likewise.
3820         (vmaxvq_s8): Likewise.
3821         (vmaxvq_s16): Likewise.
3822         (vmaxvq_s32): Likewise.
3823         (vmaxvq_u8): Likewise.
3824         (vmaxvq_u16): Likewise.
3825         (vmaxvq_u32): Likewise.
3826         (vmaxnmv_f32): Likewise.
3827         (vmaxnmvq_f32): Likewise.
3828         (vmaxnmvq_f64): Likewise.
3829         (vminv_f32): Likewise.
3830         (vminv_s8): Likewise.
3831         (vminv_s16): Likewise.
3832         (vminv_s32): Likewise.
3833         (vminv_u8): Likewise.
3834         (vminv_u16): Likewise.
3835         (vminv_u32): Likewise.
3836         (vminvq_f32): Likewise.
3837         (vminvq_f64): Likewise.
3838         (vminvq_s8): Likewise.
3839         (vminvq_s16): Likewise.
3840         (vminvq_s32): Likewise.
3841         (vminvq_u8): Likewise.
3842         (vminvq_u16): Likewise.
3843         (vminvq_u32): Likewise.
3844         (vminnmv_f32): Likewise.
3845         (vminnmvq_f32): Likewise.
3846         (vminnmvq_f64): Likewise.
3847
3848 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
3849
3850         * config/aarch64/aarch64-simd.md
3851         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
3852         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
3853         (*aarch64_mul3_elt<mode>): Likewise.
3854         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
3855         (*aarch64_mul3_elt_to_64v2df): Likewise.
3856         (*aarch64_mla_elt<mode>): Likewise.
3857         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
3858         (*aarch64_mls_elt<mode>): Likewise.
3859         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
3860         (*aarch64_fma4_elt<mode>): Likewise.
3861         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
3862         (*aarch64_fma4_elt_to_64v2df): Likewise.
3863         (*aarch64_fnma4_elt<mode>): Likewise.
3864         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
3865         (*aarch64_fnma4_elt_to_64v2df): Likewise.
3866         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
3867         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
3868         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
3869         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
3870         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
3871         (aarch64_sqdmull_lane<mode>_internal): Likewise.
3872         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
3873
3874 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3875
3876         * config/aarch64/aarch64-simd.md
3877         (aarch64_be_checked_get_lane<mode>): New define_expand.
3878         * config/aarch64/aarch64-simd-builtins.def
3879         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
3880         New builtin definition.
3881         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
3882         Use new safe be builtin.
3883
3884 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3885
3886         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
3887         New define_insn.
3888         (aarch64_be_st1<mode>): Likewise.
3889         (aarch_ld1<VALL:mode>): Define_expand modified.
3890         (aarch_st1<VALL:mode>): Likewise.
3891         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
3892         (UNSPEC_ST1): Likewise.
3893
3894 2014-01-23  David Holsgrove <david.holsgrove@xilinx.com>
3895
3896         * config/microblaze/microblaze.md: Add trap insn and attribute
3897
3898 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
3899
3900         PR preprocessor/58580
3901         * input.h (location_get_source_line): Take an additional line_size
3902         parameter.
3903         (void diagnostics_file_cache_fini): Declare new function.
3904         * input.c (struct fcache): New type.
3905         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
3906         New static constants.
3907         (diagnostic_file_cache_init, total_lines_num)
3908         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
3909         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
3910         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
3911         (get_next_line, read_next_line, goto_next_line, read_line_num):
3912         New static function definitions.
3913         (diagnostic_file_cache_fini): New function.
3914         (location_get_source_line): Take an additional output line_len
3915         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
3916         read_line_num.
3917         * diagnostic.c (diagnostic_finish): Call
3918         diagnostic_file_cache_fini.
3919         (adjust_line): Take an additional input parameter for the length
3920         of the line, rather than calculating it with strlen.
3921         (diagnostic_show_locus): Adjust the use of
3922         location_get_source_line and adjust_line with respect to their new
3923         signature.  While displaying a line now, do not stop at the first
3924         null byte.  Rather, display the zero byte as a space and keep
3925         going until we reach the size of the line.
3926         * Makefile.in: Add vec.o to OBJS-libcommon
3927
3928 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
3929             Ilya Tocar     <ilya.tocar@intel.com>
3930
3931         * config/i386/avx512fintrin.h (_mm512_kmov): New.
3932         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
3933         (__builtin_ia32_kmov16): Ditto.
3934         * config/i386/i386.md (UNSPEC_KMOV): New.
3935         (kmovw): Ditto.
3936
3937 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
3938
3939         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
3940         (_mm512_storeu_si512): Ditto.
3941
3942 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
3943
3944         PR target/52125
3945         * rtl.h (get_referenced_operands): Declare.
3946         * recog.c (get_referenced_operands): New function.
3947         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
3948         operands have been referenced when recording LO_SUM references.
3949
3950 2014-01-22  David Holsgrove <david.holsgrove@xilinx.com>
3951
3952         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
3953
3954 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
3955
3956         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
3957         Enable for generic and recent AMD targets.
3958
3959 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
3960
3961         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
3962         ARG_SIZE note when adjustment was eliminated.
3963
3964 2014-01-22  Jeff Law  <law@redhat.com>
3965
3966         PR tree-optimization/59597
3967         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
3968         in file.  Accept new argument REGISTERING and use it to modify
3969         dump output appropriately.
3970         (register_jump_thread): Corresponding changes.
3971         (mark_threaded_blocks): Reinstate code to cancel unprofitable
3972         thread paths involving joiner blocks.  Add code to dump cancelled
3973         jump threading paths.
3974
3975 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
3976
3977         PR rtl-optimization/59477
3978         * lra-constraints.c (inherit_in_ebb): Process call for living hard
3979         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
3980
3981 2014-01-22  Tom Tromey  <tromey@redhat.com>
3982
3983         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
3984         PARAMS.
3985         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
3986
3987 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
3988
3989         PR rtl-optimization/59896
3990         * lra-constraints.c (process_alt_operands): Check unused note for
3991         matched operands of insn with no output reloads.
3992
3993 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
3994
3995         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
3996         (mips_move_from_gpr_cost): Likewise.
3997
3998 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
3999
4000         PR rtl-optimization/59858
4001         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
4002         ira_class_hard_regs_num.
4003         (process_alt_operands): Increase reject for dying matched operand.
4004
4005 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
4006
4007         PR target/59003
4008         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
4009         smaller than size, perform several stores or loads and stores
4010         at dst + count - size to store or copy all of size bytes, rather
4011         than just last modesize bytes.
4012
4013 2014-01-20  DJ Delorie  <dj@redhat.com>
4014
4015         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
4016         that CLOBBERs are REGs before propogating their values.
4017
4018 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
4019
4020         PR middle-end/59789
4021         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
4022         (cgraph_inline_failed_type): New function.
4023         * cgraph.h (DEFCIFCODE): Add type.
4024         (cgraph_inline_failed_type_t): New enum.
4025         (cgraph_inline_failed_type): New prototype.
4026         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
4027         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
4028         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
4029         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
4030         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
4031         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
4032         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
4033         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
4034         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
4035         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
4036         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
4037         OPTIMIZATION_MISMATCH.
4038         * tree-inline.c (expand_call_inline): Emit errors during
4039         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
4040
4041 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
4042
4043         PR target/59685
4044         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
4045         mode attribute in insn output.
4046
4047 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
4048
4049         * output.h (output_constant): Delete.
4050         * varasm.c (output_constant): Make private.
4051
4052 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
4053
4054         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
4055
4056 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
4057
4058         PR middle-end/59860
4059         * tree.h (fold_builtin_strcat): New prototype.
4060         * builtins.c (fold_builtin_strcat): No longer static.  Add len
4061         argument, if non-NULL, don't call c_strlen.  Optimize
4062         directly into __builtin_memcpy instead of __builtin_strcpy.
4063         (fold_builtin_2): Adjust fold_builtin_strcat caller.
4064         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
4065
4066 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
4067
4068         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
4069         for SImode_address_operand operands, having only a REG argument.
4070
4071 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
4072
4073         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
4074         loader name using mbig-endian.
4075         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
4076
4077 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
4078
4079         * doc/invoke.texi (-march): Clarify documentation for AArch64.
4080         (-mtune): Likewise.
4081         (-mcpu): Likewise.
4082
4083 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
4084
4085         * config/aarch64/aarch64-protos.h
4086         (aarch64_cannot_change_mode_class_ptr): Declare.
4087         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
4088         aarch64_cannot_change_mode_class_ptr): New.
4089         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
4090         backend hook aarch64_cannot_change_mode_class.
4091
4092 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
4093
4094         * common/config/aarch64/aarch64-common.c
4095         (aarch64_handle_option): Don't handle any option order logic here.
4096         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
4097         selected_cpu, warn on architecture version mismatch.
4098         (aarch64_override_options): Fix parsing order for option strings.
4099
4100 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
4101             Iain Sandoe  <iain@codesourcery.com>
4102
4103         PR bootstrap/59496
4104         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
4105         warning.  Amend comment to reflect current functionality.
4106
4107 2014-01-20  Richard Biener  <rguenther@suse.de>
4108
4109         PR middle-end/59860
4110         * builtins.c (fold_builtin_strcat): Remove case better handled
4111         by tree-ssa-strlen.c.
4112
4113 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
4114
4115         * config/aarch64/aarch64.opt
4116         (mcpu, march, mtune): Make case-insensitive.
4117
4118 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
4119
4120         PR target/59880
4121         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
4122         if operands[1] is a REG or ZERO_EXTEND of a REG.
4123
4124 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
4125
4126         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
4127
4128 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
4129
4130         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
4131         long non-pic millicode calls.
4132
4133 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
4134
4135         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
4136
4137 2014-01-19  Kito Cheng  <kito@0xlab.org>
4138
4139         * builtins.c (expand_movstr): Check movstr expand done or fail.
4140
4141 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
4142             H.J. Lu  <hongjiu.lu@intel.com>
4143
4144         PR target/59379
4145         * config/i386/i386.md (*lea<mode>): Zero-extend return register
4146         to DImode for zero-extended addresses.
4147
4148 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
4149
4150         PR rtl-optimization/57763
4151         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
4152         on the new indirect jump_insn and increment LABEL_NUSES (label).
4153
4154 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
4155
4156         PR bootstrap/59580
4157         PR bootstrap/59583
4158         * config.gcc (x86_archs): New variable.
4159         (x86_64_archs): Likewise.
4160         (x86_cpus): Likewise.
4161         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
4162         --with-arch/--with-cpu= options.
4163         Support --with-arch=/--with-cpu={nehalem,westmere,
4164         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
4165
4166 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
4167
4168         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
4169         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
4170
4171 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
4172
4173         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
4174
4175 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
4176
4177         PR target/58944
4178         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
4179         clear cpp_get_options (parse_in)->warn_unused_macros for
4180         ix86_target_macros_internal with cpp_define.
4181
4182 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
4183
4184         * jump.c (delete_related_insns): Keep (use (insn))s.
4185         * reorg.c (redundant_insn): Check for barriers too.
4186
4187 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
4188
4189         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
4190
4191 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
4192
4193         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
4194         call to $$dyncall when TARGET_LONG_CALLS is true.
4195
4196 2014-01-17  Jeff Law  <law@redhat.com>
4197
4198         * ree.c (combine_set_extension): Temporarily disable test for
4199         changing number of hard registers.
4200
4201 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
4202
4203         PR middle-end/58125
4204         * ipa-inline-analysis.c (inline_free_summary):
4205         Do not free summary of aliases.
4206
4207 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
4208
4209         PR middle-end/59706
4210         * gimplify.c (gimplify_expr): Use create_tmp_var
4211         instead of create_tmp_var_raw.  If cond doesn't have
4212         integral type, don't add the IFN_ANNOTATE builtin at all.
4213
4214 2014-01-17  Martin Jambor  <mjambor@suse.cz>
4215
4216         PR ipa/59736
4217         * ipa-cp.c (prev_edge_clone): New variable.
4218         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
4219         Also resize prev_edge_clone vector.
4220         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
4221         (ipcp_edge_removal_hook): New function.
4222         (ipcp_driver): Register ipcp_edge_removal_hook.
4223
4224 2014-01-17  Andrew Pinski <apinski@cavium.com>
4225             Steve Ellcey  <sellcey@mips.com>
4226
4227         PR target/59462
4228         * config/mips/mips.c (mips_print_operand): Check operand mode instead
4229         of operator mode.
4230
4231 2014-01-17  Jeff Law  <law@redhat.com>
4232
4233         PR middle-end/57904
4234         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
4235         so that pass_ccp runs first.
4236
4237 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
4238
4239         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
4240         (ix86_adjust_cost): Use !TARGET_XXX.
4241         (do_reorder_for_imul): Likewise.
4242         (swap_top_of_ready_list): Likewise.
4243         (ix86_sched_reorder): Likewise.
4244
4245 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
4246
4247         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4248         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
4249         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
4250         (intel_memset): New.  Duplicate slm_memset.
4251         (intel_cost): New.  Duplicate slm_cost.
4252         (m_INTEL): New macro.
4253         (processor_target_table): Add "intel".
4254         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
4255         with PROCESSOR_INTEL for "intel".
4256         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
4257         PROCESSOR_SILVERMONT.
4258         (ix86_issue_rate): Likewise.
4259         (ix86_adjust_cost): Likewise.
4260         (ia32_multipass_dfa_lookahead): Likewise.
4261         (swap_top_of_ready_list): Likewise.
4262         (ix86_sched_reorder): Likewise.
4263         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
4264         instead of TARGET_OPT_AGU.
4265         * config/i386/i386.h (TARGET_INTEL): New.
4266         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
4267         (processor_type): Add PROCESSOR_INTEL.
4268         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
4269         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
4270
4271 2014-01-17  Marek Polacek  <polacek@redhat.com>
4272
4273         PR c/58346
4274         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
4275         size is zero.
4276
4277 2014-01-17  Richard Biener  <rguenther@suse.de>
4278
4279         PR tree-optimization/46590
4280         * opts.c (default_options_table): Add entries for
4281         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
4282         all enabled at -O1 but not for -Og.
4283         * common.opt (fbranch-count-reg): Remove Init(1).
4284         (fmove-loop-invariants): Likewise.
4285         (ftree-pta): Likewise.
4286
4287 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
4288
4289         * config/i386/i386.c (ix86_data_alignment): For compatibility with
4290         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
4291         decls to at least the GCC 4.8 used alignments.
4292
4293         PR fortran/59440
4294         * tree-nested.c (convert_nonlocal_reference_stmt,
4295         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
4296         of GIMPLE_BIND stmts, adjust associated decls.
4297
4298 2014-01-17  Richard Biener  <rguenther@suse.de>
4299
4300         PR tree-optimization/46590
4301         * vec.h (vec<>::bseach): New member function implementing
4302         binary search according to C89 bsearch.
4303         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
4304         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
4305         bitmap pointer again.  Make accesses_in_loop a flat array.
4306         (mem_ref_obstack): New global.
4307         (outermost_indep_loop): Adjust for mem_ref->stored changes.
4308         (mark_ref_stored): Likewise.
4309         (ref_indep_loop_p_2): Likewise.
4310         (set_ref_stored_in_loop): New helper function.
4311         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
4312         (memref_free): Adjust.
4313         (record_mem_ref_loc): Simplify.
4314         (gather_mem_refs_stmt): Adjust.
4315         (sort_locs_in_loop_postorder_cmp): New function.
4316         (analyze_memory_references): Sort accesses_in_loop after
4317         loop postorder number.
4318         (find_ref_loc_in_loop_cmp): New function.
4319         (for_all_locs_in_loop): Find relevant cluster of locs in
4320         accesses_in_loop and iterate without recursion.
4321         (execute_sm): Avoid uninit warning.
4322         (struct ref_always_accessed): Simplify.
4323         (ref_always_accessed::operator ()): Likewise.
4324         (ref_always_accessed_p): Likewise.
4325         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
4326         loop postorder numbers here.
4327         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
4328         numbers.
4329
4330 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
4331
4332         PR c++/57945
4333         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
4334         on decls for which assemble_alias has been called.
4335
4336 2014-01-17  Nick Clifton  <nickc@redhat.com>
4337
4338         * config/msp430/msp430.opt: (mcpu): New option.
4339         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
4340         (msp430_option_override): Parse target_cpu.  If the MCU name
4341         matches a generic string, clear target_mcu.
4342         (msp430_attr): Allow numeric interrupt values up to 63.
4343         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
4344         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
4345         option.
4346         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
4347         Add mcpu matches.
4348         * config/msp430/msp430.md (popm): Use %J rather than %I.
4349         (addsi3): Use msp430_nonimmediate_operand for operand 2.
4350         (addhi_cy_i): Use immediate_operand for operand 2.
4351         * doc/invoke.texi: Document -mcpu option.
4352
4353 2014-01-17  Richard Biener  <rguenther@suse.de>
4354
4355         PR rtl-optimization/38518
4356         * df.h (df_analyze_loop): Declare.
4357         * df-core.c: Include cfgloop.h.
4358         (df_analyze_1): Split out main part of df_analyze.
4359         (df_analyze): Adjust.
4360         (loop_inverted_post_order_compute): New function.
4361         (loop_post_order_compute): Likewise.
4362         (df_analyze_loop): New function avoiding whole-function
4363         postorder computes.
4364         * loop-invariant.c (find_defs): Use df_analyze_loop.
4365         (find_invariants): Adjust.
4366         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
4367
4368 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
4369
4370         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
4371         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
4372
4373 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
4374
4375         * ipa-ref.c (ipa_remove_stmt_references): Fix references
4376         traversal when removing references.
4377
4378 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
4379
4380         PR ipa/59775
4381         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
4382
4383 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
4384
4385         PR middle-end/56791
4386         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
4387         pushing a reload for an autoinc when we had previously reloaded an
4388         inner part of the address.
4389
4390 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
4391
4392         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
4393         field.
4394         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
4395         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
4396         when not giving up or versioning for alias only because of
4397         loop->safelen.
4398         (vect_analyze_data_ref_dependences): Set to true.
4399         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
4400         is a GIMPLE_PHI.
4401         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
4402         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
4403         to the condition.
4404
4405         PR middle-end/58344
4406         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
4407
4408         PR target/59839
4409         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
4410         operand 0 predicate for gathers, use a new pseudo as subtarget.
4411
4412 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
4413
4414         PR middle-end/59609
4415         * lra-constraints.c (process_alt_operands): Add printing debug info.
4416         Check absence of input/output reloads for matched operands too.
4417
4418 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
4419
4420         PR rtl-optimization/59835
4421         * ira.c (ira_init_register_move_cost): Increase cost for
4422         impossible modes.
4423
4424 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
4425
4426         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
4427
4428 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
4429
4430         PR target/59780
4431         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
4432         non-register objects.  Use gen_(high/low)part more consistently.
4433         Fix assertions.
4434
4435 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
4436
4437         PR target/59844
4438         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
4439         endian support, remove tests for WORDS_BIG_ENDIAN.
4440         (p8_mfvsrd_3_<mode>): Likewise.
4441         (reload_gpr_from_vsx<mode>): Likewise.
4442         (reload_gpr_from_vsxsf): Likewise.
4443         (p8_mfvsrd_4_disf): Likewise.
4444
4445 2014-01-16  Richard Biener  <rguenther@suse.de>
4446
4447         PR rtl-optimization/46590
4448         * lcm.c (compute_antinout_edge): Use postorder iteration.
4449         (compute_laterin): Use inverted postorder iteration.
4450
4451 2014-01-16  Nick Clifton  <nickc@redhat.com>
4452
4453         PR middle-end/28865
4454         * varasm.c (output_constant): Return the number of bytes actually
4455         emitted.
4456         (output_constructor_array_range): Update the field size with the
4457         number of bytes emitted by output_constant.
4458         (output_constructor_regular_field): Likewise.  Also do not
4459         complain if the total number of bytes emitted is now greater
4460         than the expected fieldpos.
4461         * output.h (output_constant): Update prototype and descriptive comment.
4462
4463 2014-01-16  Marek Polacek  <polacek@redhat.com>
4464
4465         PR middle-end/59827
4466         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
4467         it is error_mark_node.
4468
4469 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
4470
4471         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
4472         VALID_AVX256_REG_OR_OI_MODE.
4473
4474 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
4475
4476         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
4477         current procedure should be profiled.
4478
4479 2014-01-15  Andrew Pinski  <apinski@cavium.com>
4480
4481         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
4482         of moving from/to the STACK_REG register class.
4483
4484 2014-01-15  Richard Henderson  <rth@redhat.com>
4485
4486         PR debug/54694
4487         * reginfo.c (global_regs_decl): Globalize.
4488         * rtl.h (global_regs_decl): Declare.
4489         * ira.c (do_reload): Diagnose frame_pointer_needed and it
4490         reserved via global_regs.
4491
4492 2014-01-15  Teresa Johnson  <tejohnson@google.com>
4493
4494         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
4495
4496 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
4497
4498         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
4499         and vmulosh rather than call gen_vec_widen_smult_*.
4500         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
4501         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
4502         (vec_widen_smult_even_v16qi): Likewise.
4503         (vec_widen_umult_even_v8hi): Likewise.
4504         (vec_widen_smult_even_v8hi): Likewise.
4505         (vec_widen_umult_odd_v16qi): Likewise.
4506         (vec_widen_smult_odd_v16qi): Likewise.
4507         (vec_widen_umult_odd_v8hi): Likewise.
4508         (vec_widen_smult_odd_v8hi): Likewise.
4509         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
4510         vmuloub rather than call gen_vec_widen_umult_*.
4511         (vec_widen_umult_lo_v16qi): Likewise.
4512         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
4513         vmulosb rather than call gen_vec_widen_smult_*.
4514         (vec_widen_smult_lo_v16qi): Likewise.
4515         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
4516         rather than call gen_vec_widen_umult_*.
4517         (vec_widen_umult_lo_v8hi): Likewise.
4518         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
4519         rather than call gen_vec_widen_smult_*.
4520         (vec_widen_smult_lo_v8hi): Likewise.
4521
4522 2014-01-15  Jeff Law  <law@redhat.com>
4523
4524         PR tree-optimization/59747
4525         * ree.c (find_and_remove_re): Properly handle case where a second
4526         eliminated extension requires widening a copy created for elimination
4527         of a prior extension.
4528         (combine_set_extension): Ensure that the number of hard regs needed
4529         for a destination register does not change when we widen it.
4530
4531 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
4532
4533         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
4534         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
4535         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
4536         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
4537         (avr-*-rtems*): Likewise.
4538         (bfin*-rtems*): Likewise.
4539         (moxie-*-rtems*): Likewise.
4540         (h8300-*-rtems*): Likewise.
4541         (i[34567]86-*-rtems*): Likewise.
4542         (lm32-*-rtems*): Likewise.
4543         (m32r-*-rtems*): Likewise.
4544         (m68k-*-rtems*): Likewise.
4545         (microblaze*-*-rtems*): Likewise.
4546         (mips*-*-rtems*): Likewise.
4547         (powerpc-*-rtems*): Likewise.
4548         (sh-*-rtems*): Likewise.
4549         (sparc-*-rtems*): Likewise.
4550         (sparc64-*-rtems*): Likewise.
4551         (v850-*-rtems*): Likewise.
4552         (m32c-*-rtems*): Likewise.
4553
4554 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
4555
4556         PR rtl-optimization/59511
4557         * ira.c (ira_init_register_move_cost): Use memory costs for some
4558         cases of register move cost calculations.
4559         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
4560         instead of BB frequency.
4561         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
4562         * lra-assigns.c (find_hard_regno_for): Ditto.
4563
4564 2014-01-15  Richard Biener  <rguenther@suse.de>
4565
4566         PR tree-optimization/59822
4567         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
4568         (vectorizable_load): Use it to hoist defs of uses of invariant
4569         loads out of the loop.
4570
4571 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
4572             Kugan Vivekanandarajah  <kuganv@linaro.org>
4573
4574         PR target/59695
4575         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
4576         truncation.
4577
4578 2014-01-15  Richard Biener  <rguenther@suse.de>
4579
4580         PR rtl-optimization/59802
4581         * lcm.c (compute_available): Use inverted postorder to seed
4582         the initial worklist.
4583
4584 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
4585
4586         PR target/59803
4587         * config/s390/s390.c (s390_preferred_reload_class): Don't return
4588         ADDR_REGS for invalid symrefs in non-PIC code.
4589
4590 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
4591
4592         PR other/58712
4593         * builtins.c (determine_block_size): Initialize *probable_max_size
4594         even if len_rtx is CONST_INT.
4595
4596 2014-01-14  Andrew Pinski  <apinski@cavium.com>
4597
4598         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
4599         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
4600         (cortexa53_tunings): Likewise.
4601         (aarch64_sched_issue_rate): New function.
4602         (TARGET_SCHED_ISSUE_RATE): Define.
4603
4604 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
4605
4606         * ira-costs.c (find_costs_and_classes): Add missed
4607         ira_init_register_move_cost_if_necessary.
4608
4609 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
4610
4611         PR target/59787
4612         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
4613
4614 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
4615
4616         PR target/59794
4617         * config/i386/i386.c (type_natural_mode): Add a bool parameter
4618         to indicate if type is used for function return value.  Warn ABI
4619         change if the vector mode isn't available for function return value.
4620         (ix86_function_arg_advance): Pass false to type_natural_mode.
4621         (ix86_function_arg): Likewise.
4622         (ix86_gimplify_va_arg): Likewise.
4623         (function_arg_32): Don't warn ABI change.
4624         (ix86_function_value): Pass true to type_natural_mode.
4625         (ix86_return_in_memory): Likewise.
4626         (ix86_struct_value_rtx): Removed.
4627         (TARGET_STRUCT_VALUE_RTX): Likewise.
4628
4629 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
4630
4631         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
4632         converting a conditional jump into a conditional return.
4633
4634 2014-01-14  Richard Biener  <rguenther@suse.de>
4635
4636         PR tree-optimization/58921
4637         PR tree-optimization/59006
4638         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
4639         hoisting invariant stmts.
4640         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
4641         invariant loads on the preheader edge if possible.
4642
4643 2014-01-14  Joey Ye  <joey.ye@arm.com>
4644
4645         * doc/plugin.texi (Building GCC plugins): Update to C++.
4646
4647 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
4648
4649         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
4650         (_mm_rcp28_round_ss): Ditto.
4651         (_mm_rsqrt28_round_sd): Ditto.
4652         (_mm_rsqrt28_round_ss): Ditto.
4653         (_mm_rcp28_sd): Ditto.
4654         (_mm_rcp28_ss): Ditto.
4655         (_mm_rsqrt28_sd): Ditto.
4656         (_mm_rsqrt28_ss): Ditto.
4657         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
4658         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
4659         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
4660         (IX86_BUILTIN_RCP28SD): Ditto.
4661         (IX86_BUILTIN_RCP28SS): Ditto.
4662         (IX86_BUILTIN_RSQRT28SD): Ditto.
4663         (IX86_BUILTIN_RSQRT28SS): Ditto.
4664         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
4665         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
4666         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
4667         (ix86_expand_special_args_builtin): Expand new FTYPE.
4668         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
4669         (srcp14<mode>): Make insn unary.
4670         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
4671         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
4672         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
4673         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
4674         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
4675         Fix rounding: make it SAE only.
4676         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
4677         Ditto.
4678         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
4679         Ditto.
4680         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
4681         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
4682         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
4683         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
4684         (round_saeonly_mask_scalar_operand4): Ditto.
4685         (round_saeonly_mask_scalar_op3): Ditto.
4686         (round_saeonly_mask_scalar_op4): Ditto.
4687
4688 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4689
4690         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4691         Implement -maltivec=be for vec_insert and vec_extract.
4692
4693 2014-01-10  DJ Delorie  <dj@redhat.com>
4694
4695         * config/msp430/msp430.md (call_internal): Don't allow memory
4696         references with SP as the base register.
4697         (call_value_internal): Likewise.
4698         * config/msp430/constraints.md (Yc): New.  For memory references
4699         that don't use SP as a base register.
4700
4701         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
4702         "an integer without a # prefix"
4703         * config/msp430/msp430.md (epilogue_helper): Use it.
4704
4705 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
4706
4707         PR target/59617
4708         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
4709         AVX512F gather builtins.
4710         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
4711         on gather decls with INTEGER_TYPE masktype.
4712         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
4713         directly into the builtin rather than hoisting it before loop.
4714
4715         PR tree-optimization/59387
4716         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
4717         (scev_const_prop): If folded_casts and type has undefined overflow,
4718         use force_gimple_operand instead of force_gimple_operand_gsi and
4719         for each added stmt if it is assign with
4720         arith_code_with_undefined_signed_overflow, call
4721         rewrite_to_defined_overflow.
4722         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
4723         gimple-fold.h instead.
4724         (arith_code_with_undefined_signed_overflow,
4725         rewrite_to_defined_overflow): Moved to ...
4726         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
4727         rewrite_to_defined_overflow): ... here.  No longer static.
4728         Include gimplify-me.h.
4729         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
4730         rewrite_to_defined_overflow): New prototypes.
4731
4732 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4733
4734         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
4735
4736 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
4737
4738         * builtins.c (get_object_alignment_2): Minor tweak.
4739         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
4740
4741 2014-01-13  Christian Bruel  <christian.bruel@st.com>
4742
4743         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
4744         optimized non constant lengths.
4745
4746 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
4747
4748         PR libgomp/59194
4749         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
4750         load as __atomic_load_N if possible.
4751
4752 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
4753
4754         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
4755         target parameter.
4756         (rs6000_expand_builtin): Adjust call.
4757
4758 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
4759
4760         PR target/58115
4761         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
4762         * config/rs6000/rs6000.c: Include target-globals.h.
4763         (rs6000_set_current_function): Instead of doing target_reinit
4764         unconditionally, use save_target_globals_default_opts and
4765         restore_target_globals.
4766
4767         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
4768         FPSCR.
4769         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
4770         (rs6000_expand_builtin): Handle mffs and mtfsf.
4771         (rs6000_init_builtins): Define mffs and mtfsf.
4772         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
4773         (rs6000_mffs): New pattern.
4774         (rs6000_mtfsf): New pattern.
4775
4776 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
4777
4778         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
4779         Start narrowing with START.  Apply candidate-use pair
4780         and check overall cost in narrowing.
4781         (iv_ca_prune): Pass new argument.
4782
4783 2014-01-10  Jeff Law  <law@redhat.com>
4784
4785         PR middle-end/59743
4786         * ree.c (combine_reaching_defs): Ensure the defining statement
4787         occurs before the extension when optimizing extensions with
4788         different source and destination hard registers.
4789
4790 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
4791
4792         PR ipa/58585
4793         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
4794         vtables into the type inheritance graph.
4795
4796 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4797
4798         PR rtl-optimization/59754
4799         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
4800         modes in the REGNO != REGNO case.
4801
4802 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4803
4804         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
4805
4806 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4807
4808         PR tree-optimization/59745
4809         * tree-predcom.c (tree_predictive_commoning_loop): Call
4810         free_affine_expand_cache if giving up because components is NULL.
4811
4812         * target-globals.c (save_target_globals): Allocate < 4KB structs using
4813         GC in payload of target_globals struct instead of allocating them on
4814         the heap and the larger structs separately using GC.
4815         * target-globals.h (struct target_globals): Make regs, hard_regs,
4816         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
4817         of GTY((skip)) and change type to void *.
4818         (reset_target_globals): Cast loads from those fields to corresponding
4819         types.
4820
4821 2014-01-10  Steve Ellcey  <sellcey@mips.com>
4822
4823         PR plugins/59335
4824         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
4825         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
4826         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
4827
4828 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
4829
4830         PR target/59744
4831         * aarch64-modes.def (CC_Zmode): New flags mode.
4832         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
4833         represents an equality.
4834         (aarch64_get_condition_code): Handle CC_Zmode.
4835         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
4836
4837 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
4838
4839         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
4840         extraction in good case.
4841
4842 2014-01-10  Richard Biener  <rguenther@suse.de>
4843
4844         PR tree-optimization/59374
4845         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
4846         checking after SLP discovery.  Mark stmts not participating
4847         in any SLP instance properly.
4848
4849 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4850
4851         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
4852         when handling a SET rtx.
4853
4854 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4855
4856         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
4857         (cortex-a57): Likewise.
4858         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
4859
4860 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4861
4862         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
4863         non-iwmmxt builtins.
4864
4865 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
4866
4867         PR ipa/58252
4868         PR ipa/59226
4869         * ipa-devirt.c record_target_from_binfo): Take as argument
4870         stack of binfos and lookup matching one for virtual inheritance.
4871         (possible_polymorphic_call_targets_1): Update.
4872
4873 2014-01-10  Huacai Chen  <chenhc@lemote.com>
4874
4875         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
4876         kernel strings for Loongson-2E/2F/3A.
4877
4878 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4879
4880         PR middle-end/59670
4881         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
4882         is_gimple_call before calling gimple_call_internal_p.
4883
4884 2014-01-09  Steve Ellcey  <sellcey@mips.com>
4885
4886         * Makefile.in (TREE_FLOW_H): Remove.
4887         (TREE_SSA_H): Add file names from tree-flow.h.
4888         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
4889         * tree.h: Remove tree-flow.h reference.
4890         * hash-table.h: Remove tree-flow.h reference.
4891         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
4892         reference with tree-ssa-loop.h.
4893
4894 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4895
4896         * doc/invoke.texi: Add -maltivec={be,le} options, and document
4897         default element-order behavior for -maltivec.
4898         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
4899         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
4900         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
4901         when targeting big endian, at least for now.
4902         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
4903
4904 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4905
4906         PR middle-end/47735
4907         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
4908         var satisfies use_register_for_decl, just take into account type
4909         alignment, rather than decl alignment.
4910
4911         PR tree-optimization/59622
4912         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
4913         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
4914         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
4915         Don't devirtualize for inplace at all.  For targets.length () == 1,
4916         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
4917
4918 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
4919
4920         * config/i386/i386.md (cpu): Remove the unused btver1.
4921
4922 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
4923
4924         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
4925
4926 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4927
4928         PR target/58115
4929         * tree-core.h (struct target_globals): New forward declaration.
4930         (struct tree_target_option): Add globals field.
4931         * tree.h (TREE_TARGET_GLOBALS): Define.
4932         (prepare_target_option_nodes_for_pch): New prototype.
4933         * target-globals.h (struct target_globals): Define even if
4934         !SWITCHABLE_TARGET.
4935         * tree.c (prepare_target_option_node_for_pch,
4936         prepare_target_option_nodes_for_pch): New functions.
4937         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
4938         * config/i386/i386.c: Include target-globals.h.
4939         (ix86_set_current_function): Instead of doing target_reinit
4940         unconditionally, use save_target_globals_default_opts and
4941         restore_target_globals.
4942
4943 2014-01-09  Richard Biener  <rguenther@suse.de>
4944
4945         PR tree-optimization/59715
4946         * tree-cfg.h (split_critical_edges): Declare.
4947         * tree-cfg.c (split_critical_edges): Export.
4948         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
4949
4950 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
4951
4952         * cfgexpand.c (expand_stack_vars): Optionally disable
4953         asan stack protection.
4954         (expand_used_vars): Likewise.
4955         (partition_stack_vars): Likewise.
4956         * asan.c (asan_emit_stack_protection): Optionally disable
4957         after return stack usage.
4958         (instrument_derefs): Optionally disable memory access instrumentation.
4959         (instrument_builtin_call): Likewise.
4960         (instrument_strlen_call): Likewise.
4961         (asan_protect_global): Optionally disable global variables protection.
4962         * doc/invoke.texi: Added doc for new options.
4963         * params.def: Added new options.
4964         * params.h: Likewise.
4965
4966 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4967
4968         PR rtl-optimization/59724
4969         * ifcvt.c (cond_exec_process_if_block): Don't call
4970         flow_find_head_matching_sequence with 0 longest_match.
4971         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
4972         non-active insns if !stop_after.
4973         (try_head_merge_bb): Revert 2014-01-07 changes.
4974
4975 2014-01-08  Jeff Law  <law@redhat.com>
4976
4977         * ree.c (get_sub_rtx): New function, extracted from...
4978         (merge_def_and_ext): Here.
4979         (combine_reaching_defs): Use get_sub_rtx.
4980
4981 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
4982
4983         * cgraph.h (varpool_variable_node): Do not choke on null node.
4984
4985 2014-01-08  Catherine Moore  <clm@codesourcery.com>
4986
4987         * config/mips/mips.md (simple_return): Attempt to use JRC
4988         for microMIPS.
4989         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
4990
4991 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
4992
4993         PR rtl-optimization/59137
4994         * reorg.c (steal_delay_list_from_target): Call update_block for
4995         elided insns.
4996         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
4997
4998 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4999
5000         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5001         two duplicate entries.
5002
5003 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
5004
5005         Revert:
5006         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
5007
5008         * config/mips/mips.c (mips_truncated_op_cost): New function.
5009         (mips_rtx_costs): Adjust test for BADDU.
5010         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
5011
5012         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
5013
5014         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
5015         (*baddu_si): ...this new pattern.
5016
5017 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
5018
5019         PR ipa/59722
5020         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
5021
5022 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5023
5024         PR middle-end/57748
5025         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
5026         inner_reference_p.
5027         (expand_expr, expand_normal): Adjust.
5028         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
5029         inner_reference_p. Use inner_reference_p to expand inner references.
5030         (store_expr): Adjust.
5031         * cfgexpand.c (expand_call_stmt): Adjust.
5032
5033 2014-01-08  Rong Xu  <xur@google.com>
5034
5035         * gcov-io.c (gcov_var): Move from gcov-io.h.
5036         (gcov_position): Ditto.
5037         (gcov_is_error): Ditto.
5038         (gcov_rewrite): Ditto.
5039         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
5040         only part to libgcc/libgcov.h.
5041
5042 2014-01-08  Marek Polacek  <polacek@redhat.com>
5043
5044         PR middle-end/59669
5045         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
5046
5047 2014-01-08  Marek Polacek  <polacek@redhat.com>
5048
5049         PR sanitizer/59667
5050         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
5051
5052 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
5053
5054         PR rtl-optimization/59649
5055         * stor-layout.c (get_mode_bounds): For BImode return
5056         0 and STORE_FLAG_VALUE.
5057
5058 2014-01-08  Richard Biener  <rguenther@suse.de>
5059
5060         PR middle-end/59630
5061         * gimple.h (is_gimple_builtin_call): Remove.
5062         (gimple_builtin_call_types_compatible_p): New.
5063         (gimple_call_builtin_p): New overload.
5064         * gimple.c (is_gimple_builtin_call): Remove.
5065         (validate_call): Rename to ...
5066         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
5067         check return types.
5068         (validate_type): New static function.
5069         (gimple_call_builtin_p): New overload and adjust.
5070         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
5071         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
5072         (gimple_fold_stmt_to_constant_1): Likewise.
5073         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
5074
5075 2014-01-08  Richard Biener  <rguenther@suse.de>
5076
5077         PR middle-end/59471
5078         * gimplify.c (gimplify_expr): Gimplify register-register type
5079         VIEW_CONVERT_EXPRs to separate stmts.
5080
5081 2014-01-07  Jeff Law  <law@redhat.com>
5082
5083         PR middle-end/53623
5084         * ree.c (combine_set_extension): Handle case where source
5085         and destination registers in an extension insn are different.
5086         (combine_reaching_defs): Allow source and destination registers
5087         in extension to be different under limited circumstances.
5088         (add_removable_extension): Remove restriction that the
5089         source and destination registers in the extension are the same.
5090         (find_and_remove_re): Emit a copy from the extension's
5091         destination to its source after the defining insn if
5092         the source and destination registers are different.
5093
5094         PR middle-end/59285
5095         * ifcvt.c (merge_if_block): If we are merging a block with more than
5096         one successor with a block with no successors, remove any BARRIER
5097         after the second block.
5098
5099 2014-01-07  Dan Xio Qiang <ziyan01@163.com>
5100
5101         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
5102
5103 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
5104
5105         PR target/59652
5106         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
5107         for 14-bit register offsets when INT14_OK_STRICT is false.
5108
5109 2014-01-07  Roland Stigge  <stigge@antcom.de>
5110             Michael Meissner  <meissner@linux.vnet.ibm.com>
5111
5112         PR 57386/target
5113         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
5114         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
5115
5116 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
5117
5118         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
5119         -mcpu.
5120
5121 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
5122
5123         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
5124         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
5125         rtx is const0_rtx or not.
5126
5127 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
5128
5129         PR target/58115
5130         * target-globals.c (save_target_globals): Remove this_fn_optab
5131         handling.
5132         * toplev.c: Include optabs.h.
5133         (target_reinit): Temporarily restore the global options if another
5134         set of options are in force.
5135
5136 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
5137
5138         PR rtl-optimization/58668
5139         * cfgcleanup.c (flow_find_cross_jump): Don't count
5140         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
5141         to determine what is counted.
5142         (flow_find_head_matching_sequence): Use active_insn_p to determine
5143         what is counted.
5144         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
5145         counting change.
5146         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
5147         determine what is counted.
5148
5149         PR tree-optimization/59643
5150         * tree-predcom.c (split_data_refs_to_components): If one dr is
5151         read and one write, determine_offset fails and the write isn't
5152         in the bad component, just put the read into the bad component.
5153
5154 2014-01-07  Mike Stump  <mikestump@comcast.net>
5155             Jakub Jelinek  <jakub@redhat.com>
5156
5157         PR pch/59436
5158         * tree-core.h (struct tree_optimization_option): Change optabs
5159         type from unsigned char * to void *.
5160         * optabs.c (init_tree_optimization_optabs): Adjust
5161         TREE_OPTIMIZATION_OPTABS initialization.
5162
5163 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
5164
5165         PR target/59644
5166         * config/i386/i386.h (struct machine_function): Add
5167         no_drap_save_restore field.
5168         * config/i386/i386.c (ix86_save_reg): Use
5169         !cfun->machine->no_drap_save_restore instead of
5170         crtl->stack_realign_needed.
5171         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
5172         this function clears frame_pointer_needed.  Set
5173         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
5174         and DRAP reg is needed.
5175
5176 2014-01-06  Marek Polacek  <polacek@redhat.com>
5177
5178         PR c/57773
5179         * doc/implement-c.texi: Mention that other integer types are
5180         permitted as bit-field types in strictly conforming mode.
5181
5182 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
5183
5184         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
5185         is newly allocated.
5186
5187 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
5188
5189         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
5190
5191 2014-01-06  Martin Jambor  <mjambor@suse.cz>
5192
5193         PR ipa/59008
5194         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
5195         to int.
5196         * ipa-prop.c (ipa_print_node_params): Fix indentation.
5197
5198 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
5199
5200         PR debug/59350
5201         PR debug/59510
5202         * var-tracking.c (add_stores): Preserve the value of the source even if
5203         we don't record the store.
5204
5205 2014-01-06  Terry Guo  <terry.guo@arm.com>
5206
5207         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
5208
5209 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
5210
5211         PR bootstrap/59541
5212         * config/darwin.c (darwin_function_section): Adjust return values to
5213         correspond to optimisation changes made in r206070.
5214
5215 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
5216
5217         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
5218         from prefetch_block tune setting.
5219         (nocona_cost): Correct size of prefetch block to 64.
5220
5221 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
5222
5223         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
5224         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
5225         used to save the static chain register in the computation of the offset
5226         from which the FP registers need to be restored.
5227
5228 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
5229
5230         PR tree-optimization/59519
5231         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
5232         ICE if get_current_def (current_new_name) is already non-NULL, as long
5233         as it is a phi result of some other phi in *new_exit_bb that has
5234         the same argument.
5235
5236         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
5237         or vmovdqu* for misaligned_operand.
5238         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
5239         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
5240         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
5241         aligned_mem for AVX512F masked aligned load and store builtins and for
5242         non-temporal moves.
5243
5244 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
5245
5246         PR tree-optimization/59651
5247         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
5248         Address range for negative step should be added by TYPE_SIZE_UNIT.
5249
5250 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
5251
5252         * config/m68k/m68k.c (handle_move_double): Handle pushes with
5253         overlapping registers also for registers other than the stack pointer.
5254
5255 2014-01-03  Marek Polacek  <polacek@redhat.com>
5256
5257         PR other/59661
5258         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
5259         __builtin_FILE.
5260
5261 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
5262
5263         PR target/59625
5264         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
5265         asm goto as jump.
5266
5267         * config/i386/i386.md (MODE_SIZE): New mode attribute.
5268         (push splitter): Use <P:MODE_SIZE> instead of
5269         GET_MODE_SIZE (<P:MODE>mode).
5270         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
5271         (mov -1, reg peephole2): Likewise.
5272         * config/i386/sse.md (*mov<mode>_internal,
5273         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
5274         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
5275         *<code><mode>3, *andnot<mode>3<mask_name>,
5276         <mask_codefor><code><mode>3<mask_name>): Likewise.
5277         * config/i386/subst.md (mask_mode512bit_condition,
5278         sd_mask_mode512bit_condition): Likewise.
5279
5280 2014-01-02  Xinliang David Li  <davidxl@google.com>
5281
5282         PR tree-optimization/59303
5283         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
5284         (dump_predicates): Better output format.
5285         (pred_equal_p): New function.
5286         (is_neq_relop_p): Ditto.
5287         (is_neq_zero_form_p): Ditto.
5288         (pred_expr_equal_p): Ditto.
5289         (pred_neg_p): Ditto.
5290         (simplify_pred): Ditto.
5291         (simplify_preds_2): Ditto.
5292         (simplify_preds_3): Ditto.
5293         (simplify_preds_4): Ditto.
5294         (simplify_preds): Ditto.
5295         (push_pred): Ditto.
5296         (push_to_worklist): Ditto.
5297         (get_pred_info_from_cmp): Ditto.
5298         (is_degenerated_phi): Ditto.
5299         (normalize_one_pred_1): Ditto.
5300         (normalize_one_pred): Ditto.
5301         (normalize_one_pred_chain): Ditto.
5302         (normalize_preds): Ditto.
5303         (normalize_cond_1): Remove function.
5304         (normalize_cond): Ditto.
5305         (is_gcond_subset_of): Ditto.
5306         (is_subset_of_any): Ditto.
5307         (is_or_set_subset_of): Ditto.
5308         (is_and_set_subset_of): Ditto.
5309         (is_norm_cond_subset_of): Ditto.
5310         (pred_chain_length_cmp): Ditto.
5311         (convert_control_dep_chain_into_preds): Type change.
5312         (find_predicates): Ditto.
5313         (find_def_preds): Ditto.
5314         (destroy_predicates_vecs): Ditto.
5315         (find_matching_predicates_in_rest_chains): Ditto.
5316         (use_pred_not_overlap_with_undef_path_pred): Ditto.
5317         (is_pred_expr_subset): Ditto.
5318         (is_pred_chain_subset_of): Ditto.
5319         (is_included_in): Ditto.
5320         (is_superset_of): Ditto.
5321
5322 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
5323
5324         Update copyright years.
5325
5326 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
5327
5328         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
5329         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
5330         config/arc/arc.md, config/arc/arc.opt,
5331         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
5332         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
5333         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
5334         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
5335         config/linux-protos.h, config/linux.c, config/winnt-c.c,
5336         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
5337         vtable-verify.c, vtable-verify.h: Use the standard form for the
5338         copyright notice.
5339
5340 2014-01-02  Tobias Burnus  <burnus@net-b.de>
5341
5342         * gcc.c (process_command): Update copyright notice dates.
5343         * gcov-dump.c: Ditto.
5344         * gcov.c: Ditto.
5345         * doc/cpp.texi: Bump @copying's copyright year.
5346         * doc/cppinternals.texi: Ditto.
5347         * doc/gcc.texi: Ditto.
5348         * doc/gccint.texi: Ditto.
5349         * doc/gcov.texi: Ditto.
5350         * doc/install.texi: Ditto.
5351         * doc/invoke.texi: Ditto.
5352
5353 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
5354
5355         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
5356
5357 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
5358
5359         * config/i386/sse.md (*mov<mode>_internal): Guard
5360         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
5361
5362         PR rtl-optimization/59647
5363         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
5364         new_rtx into UNSIGNED_FLOAT rtxes.
5365 \f
5366 Copyright (C) 2014 Free Software Foundation, Inc.
5367
5368 Copying and distribution of this file, with or without modification,
5369 are permitted in any medium without royalty provided the copyright
5370 notice and this notice are preserved.