gcc/ChangeLog

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