gcc/ChangeLog

   1 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
   2
   3         PR libgomp/60107
   4         * config/i386/sol2-9.h: New file.
   5         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
   6         *-*-solaris2.9*): Use it.
   7
   8 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
   9
  10         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
  11         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
  12
  13 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
  14
  15         * config/microblaze/microblaze.c: Extend mcpu version format
  16
  17 2014-02-10  David Holsgrove <david.holsgrove@xilinx.com>
  18
  19         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
  20
  21 2014-02-10  Richard Henderson <rth@redhat.com>
  22
  23         PR target/59927
  24         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
  25         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
  26         ms-abi vs -mno-accumulate-outgoing-args.
  27         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
  28         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
  29         respect to ms-abi.
  30
  31 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
  32
  33         PR middle-end/60080
  34         * cfgexpand.c (expand_asm_operands): Attach source location to
  35         ASM_INPUT rtx objects.
  36         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
  37
  38 2014-02-10  Nick Clifton  <nickc@redhat.com>
  39
  40         * config/mn10300/mn10300.c (popcount): New function.
  41         (mn10300_expand_prologue): Include saved registers in stack usage
  42         count.
  43
  44 2014-02-10  Jeff Law  <law@redhat.com>
  45
  46         PR middle-end/52306
  47         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
  48         when changing the SET_DEST of a prior insn to avoid an input
  49         reload.
  50
  51 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
  52
  53         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
  54         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
  55         -mcall-openbsd, or -mcall-linux.
  56         (CC1_ENDIAN_BIG_SPEC): Remove.
  57         (CC1_ENDIAN_LITTLE_SPEC): Remove.
  58         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
  59         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
  60         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
  61         and %cc1_endian_default.
  62         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
  63
  64 2014-02-10  Richard Biener  <rguenther@suse.de>
  65
  66         PR tree-optimization/60115
  67         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
  68         MEM_REF handling.  Properly verify that the accesses are not
  69         out of the objects bound.
  70
  71 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
  72
  73         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
  74         coretex to cortex.
  75
  76 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
  77
  78         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
  79         proper constants and fix formatting.
  80         (possible_polymorphic_call_targets): Fix formatting.
  81
  82 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
  83             Ilya Tocar  <ilya.tocar@intel.com>
  84
  85         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
  86         (_mm512_loadu_epi32): Renamed into...
  87         (_mm512_loadu_si512): This.
  88         (_mm512_storeu_epi32): Renamed into...
  89         (_mm512_storeu_si512): This.
  90         (_mm512_maskz_ceil_ps): Removed.
  91         (_mm512_maskz_ceil_pd): Ditto.
  92         (_mm512_maskz_floor_ps): Ditto.
  93         (_mm512_maskz_floor_pd): Ditto.
  94         (_mm512_floor_round_ps): Ditto.
  95         (_mm512_floor_round_pd): Ditto.
  96         (_mm512_ceil_round_ps): Ditto.
  97         (_mm512_ceil_round_pd): Ditto.
  98         (_mm512_mask_floor_round_ps): Ditto.
  99         (_mm512_mask_floor_round_pd): Ditto.
 100         (_mm512_mask_ceil_round_ps): Ditto.
 101         (_mm512_mask_ceil_round_pd): Ditto.
 102         (_mm512_maskz_floor_round_ps): Ditto.
 103         (_mm512_maskz_floor_round_pd): Ditto.
 104         (_mm512_maskz_ceil_round_ps): Ditto.
 105         (_mm512_maskz_ceil_round_pd): Ditto.
 106         (_mm512_expand_pd): Ditto.
 107         (_mm512_expand_ps): Ditto.
 108         * config/i386/i386.c (ix86_builtins): Remove
 109         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
 110         (bdesc_args): Ditto.
 111         * config/i386/predicates.md (const1256_operand): New.
 112         (const_1_to_2_operand): Ditto.
 113         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
 114         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
 115         (*avx512pf_gatherpf<mode>sf): Ditto.
 116         (avx512pf_gatherpf<mode>df): Ditto.
 117         (*avx512pf_gatherpf<mode>df_mask): Ditto.
 118         (*avx512pf_gatherpf<mode>df): Ditto.
 119         (avx512pf_scatterpf<mode>sf): Ditto.
 120         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
 121         (*avx512pf_scatterpf<mode>sf): Ditto.
 122         (avx512pf_scatterpf<mode>df): Ditto.
 123         (*avx512pf_scatterpf<mode>df_mask): Ditto.
 124         (*avx512pf_scatterpf<mode>df): Ditto.
 125         (avx512f_expand<mode>): Removed.
 126         (<shift_insn><mode>3<mask_name>): Change predicate type.
 127
 128 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
 129
 130         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
 131         not at the end of datarefs vector use ordered_remove to avoid
 132         reordering datarefs vector.
 133
 134         PR c/59984
 135         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
 136         mark local addressable non-static vars as GOVD_PRIVATE
 137         instead of GOVD_LOCAL.
 138         * omp-low.c (lower_omp_for): Move gimple_bind_vars
 139         and BLOCK_VARS of gimple_bind_block to new_stmt rather
 140         than copying them.
 141
 142         PR middle-end/60092
 143         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
 144         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
 145         assume_aligned or alloc_align attributes.
 146         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
 147         arguments.  Handle also assume_aligned and alloc_align attributes.
 148         (evaluate_stmt): Adjust bit_value_assume_aligned caller.
 149         Handle calls to functions with assume_aligned or alloc_align
 150         attributes.
 151         * doc/extend.texi: Document assume_aligned and alloc_align
 152         attributes.
 153
 154 2014-02-08  Terry Guo  <terry.guo@arm.com>
 155
 156         * doc/invoke.texi: Document ARM -march=armv7e-m.
 157
 158 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
 159
 160         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
 161         flag on __cilkrts_rethrow builtin.
 162
 163         PR ipa/60026
 164         * ipa-cp.c (determine_versionability): Fail at -O0
 165         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
 166         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
 167
 168         Revert:
 169         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
 170
 171         PR ipa/60026
 172         * tree-inline.c (copy_forbidden): Fail for
 173         __attribute__((optimize (0))) functions.
 174
 175 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
 176
 177         * varpool.c: Include pointer-set.h.
 178         (varpool_remove_unreferenced_decls): Variables in other partitions
 179         will not be output; be however careful to not lose information
 180         about partitioning.
 181
 182 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
 183
 184         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
 185         lookup in the vtable constructor.
 186
 187 2014-02-07  Jeff Law  <law@redhat.com>
 188
 189         PR target/40977
 190         * config/m68k/m68k.md (ashldi_extsi): Turn into a
 191         define_insn_and_split.
 192
 193         * ipa-inline.c (inline_small_functions): Fix typos.
 194
 195 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
 196
 197         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
 198         (s390_can_use_return_insn): Declare.
 199         * config/s390/s390.h (EPILOGUE_USES): Define.
 200         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
 201         instructions.
 202         (s390_chunkify_start): Handle return JUMP_LABELs.
 203         (s390_early_mach): Emit a main_pool instruction on the entry edge.
 204         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
 205         (s390_can_use_return_insn): New functions.
 206         (s390_fix_long_loop_prediction): Handle conditional returns.
 207         (TARGET_SET_UP_BY_PROLOGUE): Define.
 208         * config/s390/s390.md (ANY_RETURN): New code iterator.
 209         (*creturn, *csimple_return, return, simple_return): New patterns.
 210
 211 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
 212
 213         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
 214         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
 215         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
 216         REG_CFA_RESTORE list when deciding not to restore a register.
 217
 218 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
 219
 220         * config/s390/s390.c: Include tree-pass.h and context.h.
 221         (s390_early_mach): New function, split out from...
 222         (s390_emit_prologue): ...here.
 223         (pass_data_s390_early_mach): New pass structure.
 224         (pass_s390_early_mach): New class.
 225         (s390_option_override): Create and register early_mach pass.
 226         Move to end of file.
 227
 228 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
 229
 230         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
 231         to match for the exit block.
 232
 233 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
 234
 235         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
 236         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
 237         Reject misaligned operands.
 238
 239 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
 240
 241         * optabs.c (expand_atomic_compare_and_swap): Allow expander to
 242         fail.
 243
 244 2014-02-07  Richard Biener  <rguenther@suse.de>
 245
 246         PR middle-end/60092
 247         * gimple-low.c (lower_builtin_posix_memalign): New function.
 248         (lower_stmt): Call it to lower posix_memalign in a way
 249         to make alignment info accessible.
 250
 251 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
 252
 253         PR c++/60082
 254         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
 255         __builtin_setjmp_receiver.
 256
 257 2014-02-07  Richard Biener  <rguenther@suse.de>
 258
 259         PR middle-end/60092
 260         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
 261         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
 262         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
 263         Handle BUILT_IN_POSIX_MEMALIGN.
 264         (find_func_clobbers): Likewise.
 265         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
 266         (call_may_clobber_ref_p_1): Likewise.
 267
 268 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
 269
 270         PR ipa/59918
 271         * ipa-devirt.c (record_target_from_binfo): Remove overactive sanity check.
 272
 273 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
 274
 275         PR ipa/59469
 276         * lto-cgraph.c (lto_output_node): Use
 277         symtab_get_symbol_partitioning_class.
 278         (lto_output_varpool_node): likewise.
 279         (symtab_get_symbol_partitioning_class): Move here from
 280         lto/lto-partition.c
 281         * cgraph.h (symbol_partitioning_class): Likewise.
 282         (symtab_get_symbol_partitioning_class): Declare.
 283
 284 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
 285
 286         * ggc.h (ggc_internal_cleared_alloc): New macro.
 287         * vec.h (vec_safe_copy): Handle memory stats.
 288         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
 289         * target-globals.c (save_target_globals): Likewise.
 290
 291 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
 292
 293         PR target/60077
 294         * expr.c (emit_move_resolve_push): Export; be bit more selective
 295         on when to clear alias set.
 296         * expr.h (emit_move_resolve_push): Declare.
 297         * function.h (struct function): Add tail_call_marked.
 298         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
 299         * config/i386/i386-protos.h (ix86_expand_push): Remove.
 300         * config/i386/i386.md (TImode move expander): De not call
 301         ix86_expand_push.
 302         (FP push expanders): Preserve memory attributes.
 303         * config/i386/sse.md (push<mode>1): Remove.
 304         * config/i386/i386.c (ix86_expand_vector_move): Handle push
 305         operation.
 306         (ix86_expand_push): Remove.
 307         * config/i386/mmx.md (push<mode>1): Remove.
 308
 309 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
 310
 311         PR rtl-optimization/60030
 312         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
 313         lopart with paradoxical subreg before shifting it up by hprec.
 314
 315 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 316
 317         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
 318         Remove extra newline at end of file.
 319         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
 320         (arm_issue_rate): Handle cortexa57.
 321         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
 322         (cortex-a57.cortex-a53): Likewise.
 323
 324 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
 325
 326         PR target/59575
 327         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
 328         don't record in REG_FRAME_RELATED_EXPR registers not set in that
 329         bitmask.
 330         (arm_expand_prologue): Adjust all callers.
 331         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
 332         info, registers also at the lowest numbered registers side.  Use
 333         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
 334         XEXP.
 335
 336         PR debug/59992
 337         * var-tracking.c (adjust_mems): Before adding a SET
 338         to amd->side_effects, adjust it's SET_SRC using
 339         simplify_replace_fn_rtx.
 340
 341 2014-02-06  Alan Modra  <amodra@gmail.com>
 342
 343         PR target/60032
 344         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
 345         change SDmode to DDmode when lra_in_progress.
 346
 347 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
 348
 349         PR middle-end/59150
 350         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
 351         free_data_ref on the dr first, and before goto again also set dr
 352         to the next dr.  For simd_lane_access, free old datarefs[i] before
 353         overwriting it.  For get_vectype_for_scalar_type failure, don't
 354         free_data_ref if simd_lane_access.
 355
 356         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
 357
 358         PR target/60062
 359         * tree.h (opts_for_fn): New inline function.
 360         (opt_for_fn): Define.
 361         * config/i386/i386.c (ix86_function_regparm): Use
 362         opt_for_fn (decl, optimize) instead of optimize.
 363
 364 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
 365
 366         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
 367         for SYMBOL_REF in large memory model.
 368
 369 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 370
 371         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
 372         and crypto support.
 373         (cortex-a57): Likewise.
 374         (cortex-a57.cortex-a53): Likewise.
 375
 376 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
 377             Kugan Vivekanandarajah  <kuganv@linaro.org>
 378
 379         * config/arm/arm.c (arm_vector_alignment_reachable): Check
 380         unaligned_access.
 381         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
 382
 383 2014-02-06  Richard Biener  <rguenther@suse.de>
 384
 385         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
 386         set_loop_copy and initialize_original_copy_tables.
 387
 388 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
 389
 390         * config/aarch64/aarch64-simd.md
 391         (aarch64_ashr_simddi): Change QI to SI.
 392
 393 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
 394             Jakub Jelinek  <jakub@redhat.com>
 395
 396         PR middle-end/60013
 397         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
 398         of the dataflow.
 399
 400 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
 401
 402         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
 403         CODE_FOR_altivec_vpku[hw]um to
 404         CODE_FOR_altivec_vpku[hw]um_direct.
 405         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
 406         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
 407         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
 408         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
 409
 410 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
 411
 412         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
 413         generation for -maltivec=be.
 414         (altivec_vsumsws): Simplify redundant test.
 415
 416 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
 417
 418         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
 419         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
 420         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
 421         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
 422         gen_altivec_vpkuwum.
 423         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
 424         BYTES_BIG_ENDIAN.
 425         (altivec_vpks<VI_char>ss): Likewise.
 426         (altivec_vpks<VI_char>us): Likewise.
 427         (altivec_vpku<VI_char>us): Likewise.
 428         (altivec_vpku<VI_char>um): Likewise.
 429         (altivec_vpku<VI_char>um_direct): New (copy of
 430         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
 431         internal use).
 432         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
 433         target is little endian and -maltivec=be is not specified.
 434         (*altivec_vupkhs<VU_char>_direct): New (copy of
 435         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal
 436         use).
 437         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
 438         target is little endian and -maltivec=be is not specified.
 439         (*altivec_vupkls<VU_char>_direct): New (copy of
 440         altivec_vupkls<VU_char> that always emits vupkls*, for internal
 441         use).
 442         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
 443         little endian and -maltivec=be is not specified.
 444         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
 445         little endian and -maltivec=be is not specified.
 446
 447 2014-02-05  Richard Henderson <rth@redhat.com>
 448
 449         PR debug/52727
 450         * combine-stack-adj.c: Revert r206943.
 451         * sched-int.h (struct deps_desc): Add last_args_size.
 452         * sched-deps.c (init_deps): Initialize it.
 453         (sched_analyze_insn): Add OUTPUT dependencies between insns that
 454         contain REG_ARGS_SIZE notes.
 455
 456 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
 457
 458         * lto-cgraph.c (asm_nodes_output): Make global.
 459         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
 460         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA
 461         parameter
 462         (driver_handle_option): Handle OPT_fwpa.
 463
 464 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
 465
 466         PR ipa/59947
 467         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
 468         a comment typo and formatting issue.  If odr_hash hasn't been
 469         created, return vNULL and set *completep to false.
 470
 471         PR middle-end/57499
 472         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
 473         bb with no successors.
 474
 475 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
 476
 477         PR target/59718
 478         * doc/invoke.texi (-march): Clarify documentation for ARM.
 479         (-mtune): Likewise.
 480         (-mcpu): Likewise.
 481
 482 2014-02-05  Richard Biener  <rguenther@suse.de>
 483
 484         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
 485         when not vectorizing because of too many alias checks.
 486         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
 487         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
 488
 489 2014-02-05  Nick Clifton  <nickc@redhat.com>
 490
 491         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
 492         accept extended registers in any mode when compiling for the
 493         MN10300.
 494
 495 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
 496
 497         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
 498         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
 499         sanitization attributes.
 500         (can_inline_edge_p): Likewise.
 501         (sanitize_attrs_match_for_inline_p): New function.
 502
 503 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
 504
 505         * ipa-prop.c (detect_type_change): Shor circuit testing of
 506         type changes on THIS pointer.
 507
 508 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
 509
 510         PR target/59777
 511         * config/pa/pa.c (legitimize_tls_address): Return original address
 512         if not passed a SYMBOL_REF rtx.
 513         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
 514         addresses.
 515         (pa_emit_move_sequence): Simplify TLS source operands.
 516         (pa_legitimate_constant_p): Reject all TLS constants.
 517         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
 518         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
 519
 520 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
 521
 522         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
 523         groups when we know they are controlled by LTO.
 524         * varasm.c (default_binds_local_p_1): If object is in other partition,
 525         it will be resolved locally.
 526
 527 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
 528
 529         * config/host-linux.c (linux_gt_pch_use_address): Don't
 530         use SSIZE_MAX because it is not always defined.
 531
 532 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
 533
 534         PR bootstrap/59913
 535         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
 536         threshold for pseudo splitting.
 537         (update_ebb_live_info): Process call argument hard registers and
 538         hard registers from insn definition too.
 539         (max_small_class_regs_num): New constant.
 540         (inherit_in_ebb): Update live hard regs through EBBs.  Update
 541         reloads_num only for small register classes.  Don't split for
 542         outputs of jumps.
 543
 544 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
 545
 546         PR ipa/60058
 547         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
 548         is non-null.
 549
 550 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
 551
 552         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default visibility is safe.
 553
 554 2014-02-04  Marek Polacek  <polacek@redhat.com>
 555
 556         * gdbinit.in (pel): Define.
 557
 558 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
 559
 560         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
 561         behavior.
 562
 563 2014-02-04  Richard Biener  <rguenther@suse.de>
 564
 565         PR lto/59723
 566         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
 567         in function context local.
 568         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
 569         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
 570         similar to LTO_imported_decl_ref.
 571
 572 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
 573
 574         PR tree-optimization/60002
 575         * cgraphclones.c (build_function_decl_skip_args): Clear
 576         DECL_LANG_SPECIFIC.
 577
 578         PR tree-optimization/60023
 579         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
 580         false to gsi_replace.
 581         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
 582         has been in some EH region and vec_stmt could throw, add
 583         vec_stmt into the same EH region.
 584         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
 585         has no lhs, ignore it.
 586         * internal-fn.c (expand_MASK_LOAD): Likewise.
 587
 588         PR ipa/60026
 589         * tree-inline.c (copy_forbidden): Fail for
 590         __attribute__((optimize (0))) functions.
 591
 592         PR other/58712
 593         * omp-low.c (simd_clone_struct_copy): If from->inbranch
 594         is set, copy one less argument.
 595         (expand_simd_clones): Don't subtract clone_info->inbranch
 596         from simd_clone_struct_alloc argument.
 597
 598         PR rtl-optimization/57915
 599         * recog.c (simplify_while_replacing): If all unary/binary/relational
 600         operation arguments are constant, attempt to simplify those.
 601
 602         PR middle-end/59261
 603         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
 604         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
 605
 606 2014-02-04  Richard Biener  <rguenther@suse.de>
 607
 608         PR tree-optimization/60012
 609         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
 610         TBAA disambiguation to all DDRs.
 611
 612 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
 613
 614         PR target/59788
 615         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
 616         (LINK_SPEC): Use it for -shared, -shared-libgcc.
 617
 618 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
 619
 620         PR ipa/59882
 621         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
 622
 623 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
 624
 625         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
 626         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
 627
 628 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
 629
 630         PR ipa/59831
 631         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
 632         to figure out targets of polymorphic calls with known decl.
 633         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
 634         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
 635         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
 636         (get_polymorphic_call_info): ... here.
 637         (get_polymorphic_call_info_from_invariant): New function.
 638
 639 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
 640
 641         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
 642         lookup via vtable pointer; check for type consistency
 643         and turn inconsitent facts into UNREACHABLE.
 644         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
 645         * gimple-fold.c (gimple_get_virt_method_for_vtable):
 646         Do not ICE on type inconsistent querries; return UNREACHABLE
 647         instead.
 648
 649 2014-02-03  Richard Henderson  <rth@twiddle.net>
 650
 651         PR tree-opt/59924
 652         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
 653         already processed this node.
 654         (normalize_one_pred_1): Pass along mark_set.
 655         (normalize_one_pred): Create and destroy a pointer_set_t.
 656         (normalize_one_pred_chain): Likewise.
 657
 658 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
 659
 660         PR gcov-profile/58602
 661         * gcc/gcov-io.c (gcov_open): Open with truncation when mode<0
 662
 663 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
 664
 665         PR ipa/59831
 666         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on -fno-devirtualize;
 667         Try to devirtualize by the knowledge of virtual table pointer given by
 668         aggregate propagation.
 669         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
 670         ipa_print_node_jump_functions): Dump also offset that
 671         is relevant for polymorphic calls.
 672         (determine_known_aggregate_parts): Add arg_type parameter; use it
 673         instead of determining the type from pointer type.
 674         (ipa_compute_jump_functions_for_edge): Update call of
 675         determine_known_aggregate_parts.
 676         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
 677         (gimple_get_virt_method_for_binfo): ... here; simplify using
 678         vtable_pointer_value_to_vtable.
 679         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
 680         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
 681         to unsigned HOST_WIDE_INT; Use vtable_pointer_value_to_vtable.
 682         (vtable_pointer_value_to_vtable): Break out from ...; handle also
 683         POINTER_PLUS_EXPR.
 684         (vtable_pointer_value_to_binfo): ... here.
 685         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
 686
 687 2014-02-03  Teresa Johnson  <tejohnson@google.com>
 688
 689         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
 690         redef of outer loop index variable.
 691
 692 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
 693
 694         PR c++/53017
 695         PR c++/59211
 696         * doc/extend.texi (Function Attributes): Typo.
 697
 698 2014-02-03  Cong Hou  <congh@google.com>
 699
 700         PR tree-optimization/60000
 701         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
 702         if the vectorized statement is a store.  A store statement can only
 703         appear at the end of pattern statements.
 704
 705 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
 706
 707         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
 708         (ix86_option_override_internal): Default long double to 64-bit for
 709         32-bit Bionic and to 128-bit for 64-bit Bionic.
 710
 711         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
 712         TARGET_LONG_DOUBLE_128 is true.
 713         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
 714
 715         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
 716         (mlong-double-64): Negate -mlong-double-128.
 717         (mlong-double-128): New option.
 718
 719         * config/i386/i386-c.c (ix86_target_macros): Define
 720         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
 721
 722         * doc/invoke.texi: Document -mlong-double-128.
 723
 724 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
 725
 726         PR rtl-optimization/60024
 727         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
 728
 729 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
 730
 731         * doc/invoke.texi: (fprofile-reorder-functions): Fix typo.
 732
 733 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
 734
 735         PR rtl-optimization/57662
 736         * sel-sched.c (code_motion_path_driver): Do not mark already not
 737         existing blocks in the visiting bitmap.
 738
 739 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
 740
 741         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
 742         on the insn being emitted.
 743
 744 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
 745             Will Deacon  <will.deacon@arm.com>
 746
 747         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
 748
 749 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 750
 751         * config/arm/arm-tables.opt: Regenerate.
 752
 753 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
 754
 755         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
 756         for vector types other than V16QImode.
 757         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
 758         define_expand, and call altivec_expand_vec_perm_le when producing
 759         code with little endian element order.
 760         (*altivec_vperm_<mode>_internal): New insn having previous
 761         behavior of altivec_vperm_<mode>.
 762         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
 763         altivec_expand_vec_perm_le when producing code with little endian
 764         element order.
 765         (*altivec_vperm_<mode>_uns_internal): New insn having previous
 766         behavior of altivec_vperm_<mode>_uns.
 767
 768 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
 769
 770         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
 771         (altivec_vsumsws): Add handling for -maltivec=be with a little
 772         endian target.
 773         (altivec_vsumsws_direct): New.
 774         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
 775         gen_altivec_vsumsws.
 776
 777 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
 778
 779         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
 780         vtable_pointer_value_to_binfo): New functions.
 781         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
 782         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
 783
 784 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
 785
 786         * config/nios2/nios2.md (load_got_register): Initialize GOT
 787         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
 788         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
 789
 790 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
 791
 792         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
 793         preserverd by passthrough, do not propagate the type.
 794
 795 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
 796
 797         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
 798         (mips_atomic_assign_expand_fenv): New function.
 799         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
 800
 801 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
 802
 803         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
 804         (__builtin_mips_set_fcsr): Likewise.
 805         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
 806         MIPS_USI_FTYPE_VOID.
 807         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
 808         (mips16_expand_set_fcsr): Likewise.
 809         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
 810         (mips16_set_fcsr_stub): Likewise.
 811         (mips16_get_fcsr_one_only_stub): New class.
 812         (mips16_set_fcsr_one_only_stub): Likewise.
 813         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
 814         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
 815         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
 816         (hard_float): New availability predicate.
 817         (mips_builtins): Add get_fcsr and set_fcsr.
 818         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
 819         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
 820         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
 821         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
 822         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
 823         patterns.
 824
 825 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
 826
 827         * config/mips/mips.c (mips_one_only_stub): New class.
 828         (mips_need_mips16_rdhwr_p): Replace with...
 829         (mips16_rdhwr_stub): ...this new variable.
 830         (mips16_stub_call_address): New function.
 831         (mips16_rdhwr_one_only_stub): New class.
 832         (mips_expand_thread_pointer): Use mips16_stub_call_address.
 833         (mips_output_mips16_rdhwr): Delete.
 834         (mips_finish_stub): New function.
 835         (mips_code_end): Use it to handle rdhwr stubs.
 836
 837 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
 838
 839         PR target/60017
 840         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
 841         when calculating size of integer atomic types.
 842
 843 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
 844
 845         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
 846
 847 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
 848
 849         PR tree-optimization/60003
 850         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
 851         * profile.c (branch_prob): Use gimple_call_builtin_p
 852         to check for BUILT_IN_SETJMP_RECEIVER.
 853         * tree-inline.c (copy_bb): Call notice_special_calls.
 854
 855 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
 856
 857         PR bootstrap/59985
 858         * lra-constraints.c (process_alt_operands): Update reload_sum only
 859         on the first pass.
 860
 861 2014-01-31  Richard Henderson  <rth@redhat.com>
 862
 863         PR middle-end/60004
 864         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
 865         until after else_eh is processed.
 866
 867 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
 868
 869         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
 870         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
 871         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
 872         in smmintrin.h, remove them.
 873         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
 874         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
 875         * config/i386/i386.md (ROUND_SAE): Fix value.
 876         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
 877         (const48_operand): New.
 878         * config/i386/subst.md (round), (round_expand): Use
 879         const_4_or_8_to_11_operand.
 880         (round_saeonly), (round_saeonly_expand): Use const48_operand.
 881
 882 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
 883
 884         * config/i386/constraints.md (Yk): Swap meaning with k.
 885         * config/i386/i386.md (movhi_internal): Change Yk to k.
 886         (movqi_internal): Ditto.
 887         (*k<logic><mode>): Ditto.
 888         (*andhi_1): Ditto.
 889         (*andqi_1): Ditto.
 890         (kandn<mode>): Ditto.
 891         (*<code>hi_1): Ditto.
 892         (*<code>qi_1): Ditto.
 893         (kxnor<mode>): Ditto.
 894         (kortestzhi): Ditto.
 895         (kortestchi): Ditto.
 896         (kunpckhi): Ditto.
 897         (*one_cmplhi2_1): Ditto.
 898         (*one_cmplqi2_1): Ditto.
 899         * config/i386/sse.md (): Change k to Yk.
 900         (avx512f_load<mode>_mask): Ditto.
 901         (avx512f_blendm<mode>): Ditto.
 902         (avx512f_store<mode>_mask): Ditto.
 903         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
 904         (avx512f_storedqu<mode>_mask): Ditto.
 905         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
 906         Ditto.
 907         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
 908         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
 909         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
 910         (avx512f_maskcmp<mode>3): Ditto.
 911         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
 912         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
 913         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
 914         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
 915         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
 916         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
 917         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
 918         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
 919         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
 920         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
 921         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
 922         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
 923         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
 924         (vec_extract_lo_<mode>_maskm): Ditto.
 925         (vec_extract_hi_<mode>_maskm): Ditto.
 926         (avx512f_vternlog<mode>_mask): Ditto.
 927         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
 928         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
 929         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
 930         (avx512f_<code>v8div16qi2_mask): Ditto.
 931         (avx512f_<code>v8div16qi2_mask_store): Ditto.
 932         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
 933         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
 934         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
 935         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
 936         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
 937         (*avx512pf_gatherpf<mode>df_mask): Ditto.
 938         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
 939         (*avx512pf_scatterpf<mode>df_mask): Ditto.
 940         (avx512cd_maskb_vec_dupv8di): Ditto.
 941         (avx512cd_maskw_vec_dupv16si): Ditto.
 942         (avx512f_vpermi2var<mode>3_maskz): Ditto.
 943         (avx512f_vpermi2var<mode>3_mask): Ditto.
 944         (avx512f_vpermi2var<mode>3_mask): Ditto.
 945         (avx512f_vpermt2var<mode>3_maskz): Ditto.
 946         (*avx512f_gathersi<mode>): Ditto.
 947         (*avx512f_gathersi<mode>_2): Ditto.
 948         (*avx512f_gatherdi<mode>): Ditto.
 949         (*avx512f_gatherdi<mode>_2): Ditto.
 950         (*avx512f_scattersi<mode>): Ditto.
 951         (*avx512f_scatterdi<mode>): Ditto.
 952         (avx512f_compress<mode>_mask): Ditto.
 953         (avx512f_compressstore<mode>_mask): Ditto.
 954         (avx512f_expand<mode>_mask): Ditto.
 955         * config/i386/subst.md (mask): Change k to Yk.
 956         (mask_scalar_merge): Ditto.
 957         (sd): Ditto.
 958
 959 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
 960
 961         * doc/extend.texi (Vector Extensions): Document ?: in C++.
 962
 963 2014-01-31  Richard Biener  <rguenther@suse.de>
 964
 965         PR middle-end/59990
 966         * builtins.c (fold_builtin_memory_op): Make sure to not
 967         use a floating-point mode or a boolean or enumeral type for
 968         the copy operation.
 969
 970 2014-01-30  DJ Delorie  <dj@redhat.com>
 971
 972         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
 973         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
 974         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
 975         whenever main() has an epilogue.
 976
 977 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
 978
 979         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
 980         unused variable "field".
 981         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
 982         (vsx_mergeh_<mode>): Likewise.
 983         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
 984         (altivec_vmrghh): Likewise.
 985         (altivec_vmrghw): Likewise.
 986         (altivec_vmrglb): Likewise.
 987         (altivec_vmrglh): Likewise.
 988         (altivec_vmrglw): Likewise.
 989         (altivec_vspltb): Add missing uses.
 990         (altivec_vsplth): Likewise.
 991         (altivec_vspltw): Likewise.
 992         (altivec_vspltsf): Likewise.
 993
 994 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
 995
 996         PR target/59923
 997         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
 998         frame related instructions.
 999
1000 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
1001
1002         PR rtl-optimization/59959
1003         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
1004         any reload of register whose subreg is invalid.
1005
1006 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
1007
1008         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
1009         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
1010         Add missing return type - void.
1011
1012 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1013
1014         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
1015         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
1016         remove element index adjustment for endian (now handled in vsx.md
1017         and altivec.md).
1018         (altivec_expand_vec_perm_const): Use
1019         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
1020         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
1021         (vsx_xxspltw_<mode>): Adjust element index for little endian.
1022         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
1023         define_expand and a new define_insn *altivec_vspltb_internal;
1024         adjust for -maltivec=be on a little endian target.
1025         (altivec_vspltb_direct): New.
1026         (altivec_vsplth): Divide into a define_expand and a new
1027         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
1028         little endian target.
1029         (altivec_vsplth_direct): New.
1030         (altivec_vspltw): Divide into a define_expand and a new
1031         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
1032         little endian target.
1033         (altivec_vspltw_direct): New.
1034         (altivec_vspltsf): Divide into a define_expand and a new
1035         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
1036         a little endian target.
1037
1038 2014-01-30  Richard Biener  <rguenther@suse.de>
1039
1040         PR tree-optimization/59993
1041         * tree-ssa-forwprop.c (associate_pointerplus): Check we
1042         can propagate form the earlier stmt and avoid the transform
1043         when the intermediate result is needed.
1044
1045 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
1046
1047         * README.Portability: Fix typo.
1048
1049 2014-01-30  David Holsgrove <david.holsgrove@xilinx.com>
1050
1051         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
1052         comparison_operator with ordered_comparison_operator.
1053
1054 2014-01-30  Nick Clifton  <nickc@redhat.com>
1055
1056         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
1057         Rename to mn10300_store_multiple_regs.
1058         * config/mn10300/mn10300.c: Likewise.
1059         * config/mn10300/mn10300.md (store_movm): Fix typo: call
1060         store_multiple_regs.
1061         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
1062         Call mn10300_store_multiple_regs.
1063
1064 2014-01-30  Nick Clifton  <nickc@redhat.com>
1065             DJ Delorie  <dj@redhat.com>
1066
1067         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
1068         %fp 2 to keep registers after it properly word-aligned.
1069         (rl78_alloc_physical_registers_umul): Handle the case where both
1070         input operands are the same.
1071
1072 2014-01-30  Richard Biener  <rguenther@suse.de>
1073
1074         PR tree-optimization/59903
1075         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
1076         check properly.
1077
1078 2014-01-30  Jason Merrill  <jason@redhat.com>
1079
1080         PR c++/59633
1081         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
1082
1083         PR c++/59645
1084         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
1085
1086 2014-01-30  Richard Biener  <rguenther@suse.de>
1087
1088         PR tree-optimization/59951
1089         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
1090
1091 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
1092
1093         PR target/59784
1094         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
1095         SFmode to DFmode case.
1096
1097 2014-01-29  DJ Delorie  <dj@redhat.com>
1098
1099         * config/msp430/msp430.opt (-minrt): New.
1100         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
1101         if -minrt given.
1102         (ENDFILE_SPEC): Likewise.
1103
1104 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
1105
1106         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
1107         (estimate_function_body_sizes): Use it.
1108
1109 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
1110
1111         PR c++/58561
1112         * dwarf2out.c (is_cxx_auto): New.
1113         (is_base_type): Use it.
1114         (gen_type_die_with_usage): Likewise.
1115
1116 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1117
1118         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
1119         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
1120         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
1121         -maltivec=be with LE targets.
1122         (vsx_mergeh_<mode>): Likewise.
1123         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
1124         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
1125         (altivec_vmrghb): Replace with define_expand and new
1126         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
1127         (altivec_vmrghb_direct): New define_insn.
1128         (altivec_vmrghh): Replace with define_expand and new
1129         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
1130         (altivec_vmrghh_direct): New define_insn.
1131         (altivec_vmrghw): Replace with define_expand and new
1132         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
1133         (altivec_vmrghw_direct): New define_insn.
1134         (*altivec_vmrghsf): Adjust for endianness.
1135         (altivec_vmrglb): Replace with define_expand and new
1136         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
1137         (altivec_vmrglb_direct): New define_insn.
1138         (altivec_vmrglh): Replace with define_expand and new
1139         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
1140         (altivec_vmrglh_direct): New define_insn.
1141         (altivec_vmrglw): Replace with define_expand and new
1142         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
1143         (altivec_vmrglw_direct): New define_insn.
1144         (*altivec_vmrglsf): Adjust for endianness.
1145         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
1146         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
1147         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
1148         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
1149         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
1150         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
1151         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
1152         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
1153
1154 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
1155
1156         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
1157         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
1158         whitespace.
1159
1160 2014-01-29  Richard Biener  <rguenther@suse.de>
1161
1162         PR tree-optimization/58742
1163         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
1164         associate_pointerplus_align.
1165         (associate_pointerplus_diff): New function.
1166         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
1167         and associate_pointerplus_diff.
1168
1169 2014-01-29  Richard Biener  <rguenther@suse.de>
1170
1171         * lto-streamer.h (LTO_major_version): Bump to 3.
1172         (LTO_minor_version): Reset to 0.
1173
1174 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
1175
1176         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
1177         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
1178         (arm_file_start): Generate correct asm header for armv7ve.
1179         * config/arm/bpabi.h: Add multilib support for armv7ve.
1180         * config/arm/driver-arm.c: Change the architectures of cortex-a7
1181         and cortex-a15 to armv7ve.
1182         * config/arm/t-aprofile: Add multilib support for armv7ve.
1183         * doc/invoke.texi: Document -march=armv7ve.
1184
1185 2014-01-29  Richard Biener  <rguenther@suse.de>
1186
1187         PR tree-optimization/58742
1188         * tree-ssa-forwprop.c (associate_plusminus): Return true
1189         if we changed sth, defer EH cleanup to ...
1190         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
1191         (simplify_mult): New function.
1192
1193 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
1194
1195         PR middle-end/59917
1196         PR tree-optimization/59920
1197         * tree.c (build_common_builtin_nodes): Remove
1198         __builtin_setjmp_dispatcher initialization.
1199         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
1200         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
1201         instead of gsi_after_labels + manually skipping debug stmts.
1202         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
1203         ignore bbs with IFN_ABNORMAL_DISPATCHER.
1204         * tree-inline.c (copy_edges_for_bb): Remove
1205         can_make_abnormal_goto argument, instead add abnormal_goto_dest
1206         argument.  Ignore computed_goto_p stmts.  Don't call
1207         make_abnormal_goto_edges.  If a call might need abnormal edges
1208         for non-local gotos, see if it already has an edge to
1209         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
1210         with true argument, don't do anything then, otherwise add
1211         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
1212         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
1213         caller.
1214         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
1215         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
1216         (lower_stmt): Don't set data->calls_builtin_setjmp.
1217         (lower_builtin_setjmp): Adjust comment.
1218         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
1219         * tree-cfg.c (found_computed_goto): Remove.
1220         (factor_computed_gotos): Remove.
1221         (make_goto_expr_edges): Return bool, true for computed gotos.
1222         Don't call make_abnormal_goto_edges.
1223         (build_gimple_cfg): Don't set found_computed_goto, don't call
1224         factor_computed_gotos.
1225         (computed_goto_p): No longer static.
1226         (make_blocks): Don't set found_computed_goto.
1227         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
1228         (make_edges): If make_goto_expr_edges returns true, push bb
1229         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
1230         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
1231         vector.  Record mapping between bbs and OpenMP regions if there
1232         are any, adjust make_gimple_omp_edges caller.  Call
1233         handle_abnormal_edges.
1234         (make_abnormal_goto_edges): Remove.
1235         * tree-cfg.h (make_abnormal_goto_edges): Remove.
1236         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
1237         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
1238         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
1239         * internal-fn.def (ABNORMAL_DISPATCHER): New.
1240         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
1241         filling *region also set *region_idx to (*region)->entry->index.
1242
1243         PR other/58712
1244         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
1245         For REGs set ORIGINAL_REGNO.
1246
1247 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
1248
1249         * doc/md.texi: Mention that a target shouldn't implement
1250         vec_widen_(s|u)mul_even/odd pair if it is less efficient
1251         than hi/lo pair.
1252
1253 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
1254
1255         PR tree-optimization/59594
1256         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
1257         a copy of the datarefs vector rather than the vector itself.
1258
1259 2014-01-28  Jason Merrill  <jason@redhat.com>
1260
1261         PR c++/53756
1262         * dwarf2out.c (auto_die): New static.
1263         (gen_type_die_with_usage): Handle C++1y 'auto'.
1264         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
1265         on definition.
1266
1267 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
1268
1269         PR target/59672
1270         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
1271         (SPEC_X32): Likewise.
1272         (SPEC_64): Likewise.
1273         * config/i386/i386.c (ix86_option_override_internal): Turn off
1274         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
1275         for TARGET_16BIT.
1276         (x86_file_start): Output .code16gcc for TARGET_16BIT.
1277         * config/i386/i386.h (TARGET_16BIT): New macro.
1278         (TARGET_16BIT_P): Likewise.
1279         * config/i386/i386.opt: Add m16.
1280         * doc/invoke.texi: Document -m16.
1281
1282 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
1283
1284         PR preprocessor/59935
1285         * input.c (location_get_source_line): Bail out on when line number
1286         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
1287
1288 2014-01-28  Richard Biener  <rguenther@suse.de>
1289
1290         PR tree-optimization/58742
1291         * tree-ssa-forwprop.c (associate_plusminus): Handle
1292         pointer subtraction of the form (T)(P + A) - (T)P.
1293
1294 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1295
1296         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
1297         at const_int_cost.
1298
1299 2014-01-28  Richard Biener  <rguenther@suse.de>
1300
1301         Revert
1302         2014-01-28  Richard Biener  <rguenther@suse.de>
1303
1304         PR rtl-optimization/45364
1305         PR rtl-optimization/59890
1306         * var-tracking.c (local_get_addr_clear_given_value): Handle
1307         already cleared slot.
1308         (val_reset): Handle not allocated local_get_addr_cache.
1309         (vt_find_locations): Use post-order on the inverted CFG.
1310
1311 2014-01-28  Richard Biener  <rguenther@suse.de>
1312
1313         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
1314
1315 2014-01-28  Richard Biener  <rguenther@suse.de>
1316
1317         PR rtl-optimization/45364
1318         PR rtl-optimization/59890
1319         * var-tracking.c (local_get_addr_clear_given_value): Handle
1320         already cleared slot.
1321         (val_reset): Handle not allocated local_get_addr_cache.
1322         (vt_find_locations): Use post-order on the inverted CFG.
1323
1324 2014-01-28  Alan Modra  <amodra@gmail.com>
1325
1326         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
1327         * configure.ac <recursive call for build != host>: Define
1328         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
1329         and LD_FOR_BUILD too.
1330         * configure: Regenerate.
1331
1332 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
1333
1334         * config/i386/i386.c (get_builtin_code_for_version): Separate
1335         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
1336         Broadwell from Haswell.
1337
1338 2014-01-27  Steve Ellcey  <sellcey@mips.com>
1339
1340         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1341         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
1342         * config/mips/mips.c (mips_option_override): Change setting
1343         of TARGET_DSP.
1344         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
1345         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
1346         Change from Mask to Var.
1347
1348 2014-01-27  Jeff Law  <law@redhat.com>
1349
1350         * ipa-inline.c (inline_small_functions): Fix typo.
1351
1352 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
1353
1354         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
1355         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
1356         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
1357         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
1358         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
1359         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
1360         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
1361         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
1362         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
1363         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
1364         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
1365         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
1366         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
1367         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
1368         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
1369         (_mm512_storeu_epi64): Ditto.
1370         (_mm512_cmpge_epi32_mask): Ditto.
1371         (_mm512_cmpge_epu32_mask): Ditto.
1372         (_mm512_cmpge_epi64_mask): Ditto.
1373         (_mm512_cmpge_epu64_mask): Ditto.
1374         (_mm512_cmple_epi32_mask): Ditto.
1375         (_mm512_cmple_epu32_mask): Ditto.
1376         (_mm512_cmple_epi64_mask): Ditto.
1377         (_mm512_cmple_epu64_mask): Ditto.
1378         (_mm512_cmplt_epi32_mask): Ditto.
1379         (_mm512_cmplt_epu32_mask): Ditto.
1380         (_mm512_cmplt_epi64_mask): Ditto.
1381         (_mm512_cmplt_epu64_mask): Ditto.
1382         (_mm512_cmpneq_epi32_mask): Ditto.
1383         (_mm512_cmpneq_epu32_mask): Ditto.
1384         (_mm512_cmpneq_epi64_mask): Ditto.
1385         (_mm512_cmpneq_epu64_mask): Ditto.
1386         (_mm512_expand_pd): Ditto.
1387         (_mm512_expand_ps): Ditto.
1388         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
1389         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
1390         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
1391         * config/i386/i386.c (ix86_builtins): Add
1392         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
1393         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
1394         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
1395         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
1396         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
1397         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
1398         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
1399         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
1400         IX86_BUILTIN_PMOVUSQW512_MEM.
1401         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
1402         __builtin_ia32_pmovsqd512mem_mask,
1403         __builtin_ia32_pmovqd512mem_mask,
1404         __builtin_ia32_pmovusqw512mem_mask,
1405         __builtin_ia32_pmovsqw512mem_mask,
1406         __builtin_ia32_pmovqw512mem_mask,
1407         __builtin_ia32_pmovusdw512mem_mask,
1408         __builtin_ia32_pmovsdw512mem_mask,
1409         __builtin_ia32_pmovdw512mem_mask,
1410         __builtin_ia32_pmovqb512mem_mask,
1411         __builtin_ia32_pmovusqb512mem_mask,
1412         __builtin_ia32_pmovsqb512mem_mask,
1413         __builtin_ia32_pmovusdb512mem_mask,
1414         __builtin_ia32_pmovsdb512mem_mask,
1415         __builtin_ia32_pmovdb512mem_mask.
1416         (bdesc_args): Add __builtin_ia32_expanddf512,
1417         __builtin_ia32_expandsf512.
1418         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
1419         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
1420         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
1421         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
1422         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
1423         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
1424         (avx512f_<code>v8div16qi2_mask_store): This.
1425         (avx512f_expand<mode>): New.
1426
1427 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
1428
1429         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
1430         New.
1431         (_mm512_mask_prefetch_i64gather_pd): Ditto.
1432         (_mm512_prefetch_i32scatter_pd): Ditto.
1433         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
1434         (_mm512_prefetch_i64scatter_pd): Ditto.
1435         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
1436         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
1437         (_mm512_mask_prefetch_i64gather_ps): Ditto.
1438         (_mm512_prefetch_i32scatter_ps): Ditto.
1439         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
1440         (_mm512_prefetch_i64scatter_ps): Ditto.
1441         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
1442         * config/i386/i386-builtin-types.def: Define
1443         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
1444         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
1445         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
1446         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
1447         IX86_BUILTIN_SCATTERPFQPD.
1448         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
1449         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
1450         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
1451         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
1452         __builtin_ia32_scatterpfqps.
1453         (ix86_expand_builtin): Expand new built-ins.
1454         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
1455         fix memory access data type.
1456         (*avx512pf_gatherpf<mode>_mask): Ditto.
1457         (*avx512pf_gatherpf<mode>): Ditto.
1458         (avx512pf_scatterpf<mode>): Ditto.
1459         (*avx512pf_scatterpf<mode>_mask): Ditto.
1460         (*avx512pf_scatterpf<mode>): Ditto.
1461         (GATHER_SCATTER_SF_MEM_MODE): New.
1462         (avx512pf_gatherpf<mode>df): Ditto.
1463         (*avx512pf_gatherpf<mode>df_mask): Ditto.
1464         (*avx512pf_scatterpf<mode>df): Ditto.
1465
1466 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
1467
1468         PR bootstrap/59934
1469         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
1470         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
1471         reached.
1472
1473 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
1474
1475         * common/config/arm/arm-common.c
1476         (arm_rewrite_mcpu): Handle multiple names.
1477         * config/arm/arm.h
1478         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
1479
1480 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
1481
1482         * gimple-builder.h (create_gimple_tmp): Delete.
1483
1484 2014-01-27  Christian Bruel  <christian.bruel@st.com>
1485
1486         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
1487         words comparisons.
1488
1489 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
1490
1491         * config/pa/pa.md (call): Generate indirect long calls to non-local
1492         functions when outputing 32-bit code.
1493         (call_value): Likewise except for special call to buggy powf function.
1494
1495         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
1496         portable runtime and PIC indirect calls.
1497         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
1498         and PIC call sequences.  Use ldo instead of blr to set return register
1499         in PIC call sequence.
1500
1501 2014-01-25  Walter Lee  <walt@tilera.com>
1502
1503         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
1504         avoid clobbering a live register.
1505
1506 2014-01-25  Walter Lee  <walt@tilera.com>
1507
1508         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
1509         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
1510         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
1511         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
1512
1513 2014-01-25  Walter Lee  <walt@tilera.com>
1514
1515         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
1516         arguments on even registers.
1517         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
1518         STACK_BOUNDARY.
1519         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
1520         (BIGGEST_ALIGNMENT): Ditto.
1521         (BIGGEST_FIELD_ALIGNMENT): Ditto.
1522
1523 2014-01-25  Walter Lee  <walt@tilera.com>
1524
1525         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
1526         insns before bundling.
1527         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
1528
1529 2014-01-25  Walter Lee  <walt@tilera.com>
1530
1531         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
1532         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
1533         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
1534
1535 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
1536
1537         * config/mips/constraints.md (kl): Delete.
1538         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
1539         define expands, using...
1540         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
1541         instructions for MIPS16.
1542         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
1543         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
1544
1545 2014-01-25  Walter Lee  <walt@tilera.com>
1546
1547         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
1548         (clzdi2): Ditto.
1549         (ffsdi2): Ditto.
1550
1551 2014-01-25  Walter Lee  <walt@tilera.com>
1552
1553         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
1554         (TARGET_EXPAND_TO_RTL_HOOK): Define.
1555
1556 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
1557
1558         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
1559         Handle XOR.
1560
1561 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
1562
1563         * print-rtl.c (in_call_function_usage): New var.
1564         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
1565         EXPR_LIST mode as mode and not as reg note name.
1566
1567         PR middle-end/59561
1568         * cfgloopmanip.c (copy_loop_info): If
1569         loop->warned_aggressive_loop_optimizations, make sure
1570         the flag is set in target loop too.
1571
1572 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
1573
1574         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
1575         flag_cilkplus.
1576         * builtins.def: Likewise.
1577         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
1578         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
1579         * ira.c (ira_setup_eliminable_regset): Likewise.
1580         * omp-low.c (gate_expand_omp): Likewise.
1581         (execute_lower_omp): Likewise.
1582         (diagnose_sb_0): Likewise.
1583         (gate_diagnose_omp_blocks): Likewise.
1584         (simd_clone_clauses_extract): Likewise.
1585         (gate): Likewise.
1586
1587 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1588
1589         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
1590         correction for little endian...
1591         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
1592         here.
1593
1594 2014-01-24  Jeff Law  <law@redhat.com>
1595
1596         PR tree-optimization/59919
1597         * tree-vrp.c (find_assert_locations_1): Do not register asserts
1598         for non-returning calls.
1599
1600 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
1601
1602         * common/config/aarch64/aarch64-common.c
1603         (aarch64_rewrite_mcpu): Handle multiple names.
1604         * config/aarch64/aarch64.h
1605         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
1606
1607 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
1608
1609         * input.c (add_file_to_cache_tab): Handle the case where fopen
1610         returns NULL.
1611
1612 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
1613
1614         PR target/59929
1615         * config/i386/i386.md (pushsf splitter): Get stack adjustment
1616         from push operand if code of push isn't PRE_DEC.
1617
1618 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
1619
1620         PR target/59909
1621         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1622         -mquad-memory-atomic.  Update -mquad-memory documentation to say
1623         it is only used for non-atomic loads/stores.
1624
1625         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
1626         -mquad-memory or -mquad-memory-atomic switches.
1627
1628         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
1629         -mquad-memory-atomic to ISA 2.07 support.
1630
1631         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
1632         to separate support of normal quad word memory operations (ldq, stq)
1633         from the atomic quad word memory operations.
1634
1635         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
1636         support to separate non-atomic quad word operations from atomic
1637         quad word operations.  Disable non-atomic quad word operations in
1638         little endian mode so that we don't have to swap words after the
1639         load and before the store.
1640         (quad_load_store_p): Add comment about atomic quad word support.
1641         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
1642         options printed with -mdebug=reg.
1643
1644         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
1645         -mquad-memory-atomic as the test for whether we have quad word
1646         atomic instructions.
1647         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
1648         or -mp8-vector are used, allow byte/half-word atomic operations.
1649
1650         * config/rs6000/sync.md (load_lockedti): Insure that the address
1651         is a proper indexed or indirect address for the lqarx instruction.
1652         On little endian systems, swap the hi/lo registers after the lqarx
1653         instruction.
1654         (load_lockedpti): Use indexed_or_indirect_operand predicate to
1655         insure the address is valid for the lqarx instruction.
1656         (store_conditionalti): Insure that the address is a proper indexed
1657         or indirect address for the stqcrx. instruction.  On little endian
1658         systems, swap the hi/lo registers before doing the stqcrx.
1659         instruction.
1660         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
1661         insure the address is valid for the stqcrx. instruction.
1662
1663         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
1664         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
1665         type of quad memory support is available.
1666
1667 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
1668
1669         PR regression/59915
1670         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
1671         there is a danger of looping.
1672
1673 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
1674
1675         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1676         force flag_ira_loop_pressure if set via command line.
1677
1678 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
1679
1680         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
1681         (ashr_simd): New builtin handling DI mode.
1682         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
1683         (aarch64_sshr_simddi): New match pattern.
1684         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
1685         (vshrd_n_s64): Likewise.
1686         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
1687
1688 2014-01-23  Nick Clifton  <nickc@redhat.com>
1689
1690         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
1691         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
1692         favour of mcu specific scripts.
1693         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
1694         430x multilibs.
1695
1696 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
1697             Alex Velenko  <Alex.Velenko@arm.com>
1698
1699         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
1700         (vaddv_s16): Likewise.
1701         (vaddv_s32): Likewise.
1702         (vaddv_u8): Likewise.
1703         (vaddv_u16): Likewise.
1704         (vaddv_u32): Likewise.
1705         (vaddvq_s8): Likewise.
1706         (vaddvq_s16): Likewise.
1707         (vaddvq_s32): Likewise.
1708         (vaddvq_s64): Likewise.
1709         (vaddvq_u8): Likewise.
1710         (vaddvq_u16): Likewise.
1711         (vaddvq_u32): Likewise.
1712         (vaddvq_u64): Likewise.
1713         (vaddv_f32): Likewise.
1714         (vaddvq_f32): Likewise.
1715         (vaddvq_f64): Likewise.
1716         (vmaxv_f32): Likewise.
1717         (vmaxv_s8): Likewise.
1718         (vmaxv_s16): Likewise.
1719         (vmaxv_s32): Likewise.
1720         (vmaxv_u8): Likewise.
1721         (vmaxv_u16): Likewise.
1722         (vmaxv_u32): Likewise.
1723         (vmaxvq_f32): Likewise.
1724         (vmaxvq_f64): Likewise.
1725         (vmaxvq_s8): Likewise.
1726         (vmaxvq_s16): Likewise.
1727         (vmaxvq_s32): Likewise.
1728         (vmaxvq_u8): Likewise.
1729         (vmaxvq_u16): Likewise.
1730         (vmaxvq_u32): Likewise.
1731         (vmaxnmv_f32): Likewise.
1732         (vmaxnmvq_f32): Likewise.
1733         (vmaxnmvq_f64): Likewise.
1734         (vminv_f32): Likewise.
1735         (vminv_s8): Likewise.
1736         (vminv_s16): Likewise.
1737         (vminv_s32): Likewise.
1738         (vminv_u8): Likewise.
1739         (vminv_u16): Likewise.
1740         (vminv_u32): Likewise.
1741         (vminvq_f32): Likewise.
1742         (vminvq_f64): Likewise.
1743         (vminvq_s8): Likewise.
1744         (vminvq_s16): Likewise.
1745         (vminvq_s32): Likewise.
1746         (vminvq_u8): Likewise.
1747         (vminvq_u16): Likewise.
1748         (vminvq_u32): Likewise.
1749         (vminnmv_f32): Likewise.
1750         (vminnmvq_f32): Likewise.
1751         (vminnmvq_f64): Likewise.
1752
1753 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
1754
1755         * config/aarch64/aarch64-simd.md
1756         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
1757         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
1758         (*aarch64_mul3_elt<mode>): Likewise.
1759         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
1760         (*aarch64_mul3_elt_to_64v2df): Likewise.
1761         (*aarch64_mla_elt<mode>): Likewise.
1762         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
1763         (*aarch64_mls_elt<mode>): Likewise.
1764         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
1765         (*aarch64_fma4_elt<mode>): Likewise.
1766         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
1767         (*aarch64_fma4_elt_to_64v2df): Likewise.
1768         (*aarch64_fnma4_elt<mode>): Likewise.
1769         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
1770         (*aarch64_fnma4_elt_to_64v2df): Likewise.
1771         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
1772         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
1773         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
1774         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
1775         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
1776         (aarch64_sqdmull_lane<mode>_internal): Likewise.
1777         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
1778
1779 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
1780
1781         * config/aarch64/aarch64-simd.md
1782         (aarch64_be_checked_get_lane<mode>): New define_expand.
1783         * config/aarch64/aarch64-simd-builtins.def
1784         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
1785         New builtin definition.
1786         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
1787         Use new safe be builtin.
1788
1789 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
1790
1791         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
1792         New define_insn.
1793         (aarch64_be_st1<mode>): Likewise.
1794         (aarch_ld1<VALL:mode>): Define_expand modified.
1795         (aarch_st1<VALL:mode>): Likewise.
1796         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
1797         (UNSPEC_ST1): Likewise.
1798
1799 2014-01-23  David Holsgrove <david.holsgrove@xilinx.com>
1800
1801         * config/microblaze/microblaze.md: Add trap insn and attribute
1802
1803 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
1804
1805         PR preprocessor/58580
1806         * input.h (location_get_source_line): Take an additional line_size
1807         parameter.
1808         (void diagnostics_file_cache_fini): Declare new function.
1809         * input.c (struct fcache): New type.
1810         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
1811         New static constants.
1812         (diagnostic_file_cache_init, total_lines_num)
1813         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
1814         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
1815         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
1816         (get_next_line, read_next_line, goto_next_line, read_line_num):
1817         New static function definitions.
1818         (diagnostic_file_cache_fini): New function.
1819         (location_get_source_line): Take an additional output line_len
1820         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
1821         read_line_num.
1822         * diagnostic.c (diagnostic_finish): Call
1823         diagnostic_file_cache_fini.
1824         (adjust_line): Take an additional input parameter for the length
1825         of the line, rather than calculating it with strlen.
1826         (diagnostic_show_locus): Adjust the use of
1827         location_get_source_line and adjust_line with respect to their new
1828         signature.  While displaying a line now, do not stop at the first
1829         null byte.  Rather, display the zero byte as a space and keep
1830         going until we reach the size of the line.
1831         * Makefile.in: Add vec.o to OBJS-libcommon
1832
1833 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
1834             Ilya Tocar     <ilya.tocar@intel.com>
1835
1836         * config/i386/avx512fintrin.h (_mm512_kmov): New.
1837         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
1838         (__builtin_ia32_kmov16): Ditto.
1839         * config/i386/i386.md (UNSPEC_KMOV): New.
1840         (kmovw): Ditto.
1841
1842 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
1843
1844         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
1845         (_mm512_storeu_si512): Ditto.
1846
1847 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
1848
1849         PR target/52125
1850         * rtl.h (get_referenced_operands): Declare.
1851         * recog.c (get_referenced_operands): New function.
1852         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
1853         operands have been referenced when recording LO_SUM references.
1854
1855 2014-01-22  David Holsgrove <david.holsgrove@xilinx.com>
1856
1857         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
1858
1859 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
1860
1861         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
1862         Enable for generic and recent AMD targets.
1863
1864 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
1865
1866         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
1867         ARG_SIZE note when adjustment was eliminated.
1868
1869 2014-01-22  Jeff Law  <law@redhat.com>
1870
1871         PR tree-optimization/59597
1872         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
1873         in file.  Accept new argument REGISTERING and use it to modify
1874         dump output appropriately.
1875         (register_jump_thread): Corresponding changes.
1876         (mark_threaded_blocks): Reinstate code to cancel unprofitable
1877         thread paths involving joiner blocks.  Add code to dump cancelled
1878         jump threading paths.
1879
1880 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
1881
1882         PR rtl-optimization/59477
1883         * lra-constraints.c (inherit_in_ebb): Process call for living hard
1884         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
1885
1886 2014-01-22  Tom Tromey  <tromey@redhat.com>
1887
1888         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
1889         PARAMS.
1890         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
1891
1892 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
1893
1894         PR rtl-optimization/59896
1895         * lra-constraints.c (process_alt_operands): Check unused note for
1896         matched operands of insn with no output reloads.
1897
1898 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
1899
1900         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
1901         (mips_move_from_gpr_cost): Likewise.
1902
1903 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
1904
1905         PR rtl-optimization/59858
1906         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
1907         ira_class_hard_regs_num.
1908         (process_alt_operands): Increase reject for dying matched operand.
1909
1910 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
1911
1912         PR target/59003
1913         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
1914         smaller than size, perform several stores or loads and stores
1915         at dst + count - size to store or copy all of size bytes, rather
1916         than just last modesize bytes.
1917
1918 2014-01-20  DJ Delorie  <dj@redhat.com>
1919
1920         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
1921         that CLOBBERs are REGs before propogating their values.
1922
1923 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
1924
1925         PR middle-end/59789
1926         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
1927         (cgraph_inline_failed_type): New function.
1928         * cgraph.h (DEFCIFCODE): Add type.
1929         (cgraph_inline_failed_type_t): New enum.
1930         (cgraph_inline_failed_type): New prototype.
1931         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
1932         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
1933         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
1934         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
1935         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
1936         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
1937         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
1938         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
1939         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
1940         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
1941         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
1942         OPTIMIZATION_MISMATCH.
1943         * tree-inline.c (expand_call_inline): Emit errors during
1944         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
1945
1946 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
1947
1948         PR target/59685
1949         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
1950         mode attribute in insn output.
1951
1952 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
1953
1954         * output.h (output_constant): Delete.
1955         * varasm.c (output_constant): Make private.
1956
1957 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
1958
1959         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
1960
1961 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
1962
1963         PR middle-end/59860
1964         * tree.h (fold_builtin_strcat): New prototype.
1965         * builtins.c (fold_builtin_strcat): No longer static.  Add len
1966         argument, if non-NULL, don't call c_strlen.  Optimize
1967         directly into __builtin_memcpy instead of __builtin_strcpy.
1968         (fold_builtin_2): Adjust fold_builtin_strcat caller.
1969         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
1970
1971 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
1972
1973         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
1974         for SImode_address_operand operands, having only a REG argument.
1975
1976 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
1977
1978         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
1979         loader name using mbig-endian.
1980         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
1981
1982 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
1983
1984         * doc/invoke.texi (-march): Clarify documentation for AArch64.
1985         (-mtune): Likewise.
1986         (-mcpu): Likewise.
1987
1988 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
1989
1990         * config/aarch64/aarch64-protos.h
1991         (aarch64_cannot_change_mode_class_ptr): Declare.
1992         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
1993         aarch64_cannot_change_mode_class_ptr): New.
1994         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
1995         backend hook aarch64_cannot_change_mode_class.
1996
1997 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
1998
1999         * common/config/aarch64/aarch64-common.c
2000         (aarch64_handle_option): Don't handle any option order logic here.
2001         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
2002         selected_cpu, warn on architecture version mismatch.
2003         (aarch64_override_options): Fix parsing order for option strings.
2004
2005 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
2006             Iain Sandoe  <iain@codesourcery.com>
2007
2008         PR bootstrap/59496
2009         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
2010         warning.  Amend comment to reflect current functionality.
2011
2012 2014-01-20  Richard Biener  <rguenther@suse.de>
2013
2014         PR middle-end/59860
2015         * builtins.c (fold_builtin_strcat): Remove case better handled
2016         by tree-ssa-strlen.c.
2017
2018 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
2019
2020         * config/aarch64/aarch64.opt
2021         (mcpu, march, mtune): Make case-insensitive.
2022
2023 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
2024
2025         PR target/59880
2026         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
2027         if operands[1] is a REG or ZERO_EXTEND of a REG.
2028
2029 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
2030
2031         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
2032
2033 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
2034
2035         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
2036         long non-pic millicode calls.
2037
2038 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
2039
2040         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
2041
2042 2014-01-19  Kito Cheng  <kito@0xlab.org>
2043
2044         * builtins.c (expand_movstr): Check movstr expand done or fail.
2045
2046 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
2047             H.J. Lu  <hongjiu.lu@intel.com>
2048
2049         PR target/59379
2050         * config/i386/i386.md (*lea<mode>): Zero-extend return register
2051         to DImode for zero-extended addresses.
2052
2053 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
2054
2055         PR rtl-optimization/57763
2056         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
2057         on the new indirect jump_insn and increment LABEL_NUSES (label).
2058
2059 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
2060
2061         PR bootstrap/59580
2062         PR bootstrap/59583
2063         * config.gcc (x86_archs): New variable.
2064         (x86_64_archs): Likewise.
2065         (x86_cpus): Likewise.
2066         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
2067         --with-arch/--with-cpu= options.
2068         Support --with-arch=/--with-cpu={nehalem,westmere,
2069         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
2070
2071 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
2072
2073         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
2074         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
2075
2076 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
2077
2078         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
2079
2080 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
2081
2082         PR target/58944
2083         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
2084         clear cpp_get_options (parse_in)->warn_unused_macros for
2085         ix86_target_macros_internal with cpp_define.
2086
2087 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
2088
2089         * jump.c (delete_related_insns): Keep (use (insn))s.
2090         * reorg.c (redundant_insn): Check for barriers too.
2091
2092 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
2093
2094         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
2095
2096 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
2097
2098         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
2099         call to $$dyncall when TARGET_LONG_CALLS is true.
2100
2101 2014-01-17  Jeff Law  <law@redhat.com>
2102
2103         * ree.c (combine_set_extension): Temporarily disable test for
2104         changing number of hard registers.
2105
2106 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
2107
2108         PR middle-end/58125
2109         * ipa-inline-analysis.c (inline_free_summary):
2110         Do not free summary of aliases.
2111
2112 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
2113
2114         PR middle-end/59706
2115         * gimplify.c (gimplify_expr): Use create_tmp_var
2116         instead of create_tmp_var_raw.  If cond doesn't have
2117         integral type, don't add the IFN_ANNOTATE builtin at all.
2118
2119 2014-01-17  Martin Jambor  <mjambor@suse.cz>
2120
2121         PR ipa/59736
2122         * ipa-cp.c (prev_edge_clone): New variable.
2123         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
2124         Also resize prev_edge_clone vector.
2125         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
2126         (ipcp_edge_removal_hook): New function.
2127         (ipcp_driver): Register ipcp_edge_removal_hook.
2128
2129 2014-01-17  Andrew Pinski <apinski@cavium.com>
2130             Steve Ellcey  <sellcey@mips.com>
2131
2132         PR target/59462
2133         * config/mips/mips.c (mips_print_operand): Check operand mode instead
2134         of operator mode.
2135
2136 2014-01-17  Jeff Law  <law@redhat.com>
2137
2138         PR middle-end/57904
2139         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
2140         so that pass_ccp runs first.
2141
2142 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
2143
2144         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
2145         (ix86_adjust_cost): Use !TARGET_XXX.
2146         (do_reorder_for_imul): Likewise.
2147         (swap_top_of_ready_list): Likewise.
2148         (ix86_sched_reorder): Likewise.
2149
2150 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
2151
2152         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2153         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
2154         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
2155         (intel_memset): New.  Duplicate slm_memset.
2156         (intel_cost): New.  Duplicate slm_cost.
2157         (m_INTEL): New macro.
2158         (processor_target_table): Add "intel".
2159         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
2160         with PROCESSOR_INTEL for "intel".
2161         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
2162         PROCESSOR_SILVERMONT.
2163         (ix86_issue_rate): Likewise.
2164         (ix86_adjust_cost): Likewise.
2165         (ia32_multipass_dfa_lookahead): Likewise.
2166         (swap_top_of_ready_list): Likewise.
2167         (ix86_sched_reorder): Likewise.
2168         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
2169         instead of TARGET_OPT_AGU.
2170         * config/i386/i386.h (TARGET_INTEL): New.
2171         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
2172         (processor_type): Add PROCESSOR_INTEL.
2173         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
2174         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
2175
2176 2014-01-17  Marek Polacek  <polacek@redhat.com>
2177
2178         PR c/58346
2179         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
2180         size is zero.
2181
2182 2014-01-17  Richard Biener  <rguenther@suse.de>
2183
2184         PR tree-optimization/46590
2185         * opts.c (default_options_table): Add entries for
2186         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
2187         all enabled at -O1 but not for -Og.
2188         * common.opt (fbranch-count-reg): Remove Init(1).
2189         (fmove-loop-invariants): Likewise.
2190         (ftree-pta): Likewise.
2191
2192 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
2193
2194         * config/i386/i386.c (ix86_data_alignment): For compatibility with
2195         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
2196         decls to at least the GCC 4.8 used alignments.
2197
2198         PR fortran/59440
2199         * tree-nested.c (convert_nonlocal_reference_stmt,
2200         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
2201         of GIMPLE_BIND stmts, adjust associated decls.
2202
2203 2014-01-17  Richard Biener  <rguenther@suse.de>
2204
2205         PR tree-optimization/46590
2206         * vec.h (vec<>::bseach): New member function implementing
2207         binary search according to C89 bsearch.
2208         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
2209         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
2210         bitmap pointer again.  Make accesses_in_loop a flat array.
2211         (mem_ref_obstack): New global.
2212         (outermost_indep_loop): Adjust for mem_ref->stored changes.
2213         (mark_ref_stored): Likewise.
2214         (ref_indep_loop_p_2): Likewise.
2215         (set_ref_stored_in_loop): New helper function.
2216         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
2217         (memref_free): Adjust.
2218         (record_mem_ref_loc): Simplify.
2219         (gather_mem_refs_stmt): Adjust.
2220         (sort_locs_in_loop_postorder_cmp): New function.
2221         (analyze_memory_references): Sort accesses_in_loop after
2222         loop postorder number.
2223         (find_ref_loc_in_loop_cmp): New function.
2224         (for_all_locs_in_loop): Find relevant cluster of locs in
2225         accesses_in_loop and iterate without recursion.
2226         (execute_sm): Avoid uninit warning.
2227         (struct ref_always_accessed): Simplify.
2228         (ref_always_accessed::operator ()): Likewise.
2229         (ref_always_accessed_p): Likewise.
2230         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
2231         loop postorder numbers here.
2232         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
2233         numbers.
2234
2235 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
2236
2237         PR c++/57945
2238         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
2239         on decls for which assemble_alias has been called.
2240
2241 2014-01-17  Nick Clifton  <nickc@redhat.com>
2242
2243         * config/msp430/msp430.opt: (mcpu): New option.
2244         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
2245         (msp430_option_override): Parse target_cpu.  If the MCU name
2246         matches a generic string, clear target_mcu.
2247         (msp430_attr): Allow numeric interrupt values up to 63.
2248         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
2249         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
2250         option.
2251         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
2252         Add mcpu matches.
2253         * config/msp430/msp430.md (popm): Use %J rather than %I.
2254         (addsi3): Use msp430_nonimmediate_operand for operand 2.
2255         (addhi_cy_i): Use immediate_operand for operand 2.
2256         * doc/invoke.texi: Document -mcpu option.
2257
2258 2014-01-17  Richard Biener  <rguenther@suse.de>
2259
2260         PR rtl-optimization/38518
2261         * df.h (df_analyze_loop): Declare.
2262         * df-core.c: Include cfgloop.h.
2263         (df_analyze_1): Split out main part of df_analyze.
2264         (df_analyze): Adjust.
2265         (loop_inverted_post_order_compute): New function.
2266         (loop_post_order_compute): Likewise.
2267         (df_analyze_loop): New function avoiding whole-function
2268         postorder computes.
2269         * loop-invariant.c (find_defs): Use df_analyze_loop.
2270         (find_invariants): Adjust.
2271         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
2272
2273 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
2274
2275         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
2276         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
2277
2278 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
2279
2280         * ipa-ref.c (ipa_remove_stmt_references): Fix references
2281         traversal when removing references.
2282
2283 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
2284
2285         PR ipa/59775
2286         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
2287
2288 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
2289
2290         PR middle-end/56791
2291         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
2292         pushing a reload for an autoinc when we had previously reloaded an
2293         inner part of the address.
2294
2295 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
2296
2297         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
2298         field.
2299         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
2300         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
2301         when not giving up or versioning for alias only because of
2302         loop->safelen.
2303         (vect_analyze_data_ref_dependences): Set to true.
2304         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
2305         is a GIMPLE_PHI.
2306         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
2307         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
2308         to the condition.
2309
2310         PR middle-end/58344
2311         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
2312
2313         PR target/59839
2314         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
2315         operand 0 predicate for gathers, use a new pseudo as subtarget.
2316
2317 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
2318
2319         PR middle-end/59609
2320         * lra-constraints.c (process_alt_operands): Add printing debug info.
2321         Check absence of input/output reloads for matched operands too.
2322
2323 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
2324
2325         PR rtl-optimization/59835
2326         * ira.c (ira_init_register_move_cost): Increase cost for
2327         impossible modes.
2328
2329 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
2330
2331         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
2332
2333 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
2334
2335         PR target/59780
2336         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
2337         non-register objects.  Use gen_(high/low)part more consistently.
2338         Fix assertions.
2339
2340 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
2341
2342         PR target/59844
2343         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
2344         endian support, remove tests for WORDS_BIG_ENDIAN.
2345         (p8_mfvsrd_3_<mode>): Likewise.
2346         (reload_gpr_from_vsx<mode>): Likewise.
2347         (reload_gpr_from_vsxsf): Likewise.
2348         (p8_mfvsrd_4_disf): Likewise.
2349
2350 2014-01-16  Richard Biener  <rguenther@suse.de>
2351
2352         PR rtl-optimization/46590
2353         * lcm.c (compute_antinout_edge): Use postorder iteration.
2354         (compute_laterin): Use inverted postorder iteration.
2355
2356 2014-01-16  Nick Clifton  <nickc@redhat.com>
2357
2358         PR middle-end/28865
2359         * varasm.c (output_constant): Return the number of bytes actually
2360         emitted.
2361         (output_constructor_array_range): Update the field size with the
2362         number of bytes emitted by output_constant.
2363         (output_constructor_regular_field): Likewise.  Also do not
2364         complain if the total number of bytes emitted is now greater
2365         than the expected fieldpos.
2366         * output.h (output_constant): Update prototype and descriptive comment.
2367
2368 2014-01-16  Marek Polacek  <polacek@redhat.com>
2369
2370         PR middle-end/59827
2371         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
2372         it is error_mark_node.
2373
2374 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
2375
2376         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
2377         VALID_AVX256_REG_OR_OI_MODE.
2378
2379 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
2380
2381         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
2382         current procedure should be profiled.
2383
2384 2014-01-15  Andrew Pinski  <apinski@cavium.com>
2385
2386         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
2387         of moving from/to the STACK_REG register class.
2388
2389 2014-01-15  Richard Henderson <rth@redhat.com>
2390
2391         PR debug/54694
2392         * reginfo.c (global_regs_decl): Globalize.
2393         * rtl.h (global_regs_decl): Declare.
2394         * ira.c (do_reload): Diagnose frame_pointer_needed and it
2395         reserved via global_regs.
2396
2397 2014-01-15  Teresa Johnson  <tejohnson@google.com>
2398
2399         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
2400
2401 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
2402
2403         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
2404         and vmulosh rather than call gen_vec_widen_smult_*.
2405         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
2406         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
2407         (vec_widen_smult_even_v16qi): Likewise.
2408         (vec_widen_umult_even_v8hi): Likewise.
2409         (vec_widen_smult_even_v8hi): Likewise.
2410         (vec_widen_umult_odd_v16qi): Likewise.
2411         (vec_widen_smult_odd_v16qi): Likewise.
2412         (vec_widen_umult_odd_v8hi): Likewise.
2413         (vec_widen_smult_odd_v8hi): Likewise.
2414         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
2415         vmuloub rather than call gen_vec_widen_umult_*.
2416         (vec_widen_umult_lo_v16qi): Likewise.
2417         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
2418         vmulosb rather than call gen_vec_widen_smult_*.
2419         (vec_widen_smult_lo_v16qi): Likewise.
2420         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
2421         rather than call gen_vec_widen_umult_*.
2422         (vec_widen_umult_lo_v8hi): Likewise.
2423         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
2424         rather than call gen_vec_widen_smult_*.
2425         (vec_widen_smult_lo_v8hi): Likewise.
2426
2427 2014-01-15  Jeff Law  <law@redhat.com>
2428
2429         PR tree-optimization/59747
2430         * ree.c (find_and_remove_re): Properly handle case where a second
2431         eliminated extension requires widening a copy created for elimination
2432         of a prior extension.
2433         (combine_set_extension): Ensure that the number of hard regs needed
2434         for a destination register does not change when we widen it.
2435
2436 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
2437
2438         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
2439         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
2440         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
2441         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
2442         (avr-*-rtems*): Likewise.
2443         (bfin*-rtems*): Likewise.
2444         (moxie-*-rtems*): Likewise.
2445         (h8300-*-rtems*): Likewise.
2446         (i[34567]86-*-rtems*): Likewise.
2447         (lm32-*-rtems*): Likewise.
2448         (m32r-*-rtems*): Likewise.
2449         (m68k-*-rtems*): Likewise.
2450         (microblaze*-*-rtems*): Likewise.
2451         (mips*-*-rtems*): Likewise.
2452         (powerpc-*-rtems*): Likewise.
2453         (sh-*-rtems*): Likewise.
2454         (sparc-*-rtems*): Likewise.
2455         (sparc64-*-rtems*): Likewise.
2456         (v850-*-rtems*): Likewise.
2457         (m32c-*-rtems*): Likewise.
2458
2459 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
2460
2461         PR rtl-optimization/59511
2462         * ira.c (ira_init_register_move_cost): Use memory costs for some
2463         cases of register move cost calculations.
2464         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
2465         instead of BB frequency.
2466         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
2467         * lra-assigns.c (find_hard_regno_for): Ditto.
2468
2469 2014-01-15  Richard Biener  <rguenther@suse.de>
2470
2471         PR tree-optimization/59822
2472         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
2473         (vectorizable_load): Use it to hoist defs of uses of invariant
2474         loads out of the loop.
2475
2476 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
2477             Kugan Vivekanandarajah  <kuganv@linaro.org>
2478
2479         PR target/59695
2480         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
2481         truncation.
2482
2483 2014-01-15  Richard Biener  <rguenther@suse.de>
2484
2485         PR rtl-optimization/59802
2486         * lcm.c (compute_available): Use inverted postorder to seed
2487         the initial worklist.
2488
2489 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2490
2491         PR target/59803
2492         * config/s390/s390.c (s390_preferred_reload_class): Don't return
2493         ADDR_REGS for invalid symrefs in non-PIC code.
2494
2495 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
2496
2497         PR other/58712
2498         * builtins.c (determine_block_size): Initialize *probable_max_size
2499         even if len_rtx is CONST_INT.
2500
2501 2014-01-14  Andrew Pinski  <apinski@cavium.com>
2502
2503         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
2504         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
2505         (cortexa53_tunings): Likewise.
2506         (aarch64_sched_issue_rate): New function.
2507         (TARGET_SCHED_ISSUE_RATE): Define.
2508
2509 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
2510
2511         * ira-costs.c (find_costs_and_classes): Add missed
2512         ira_init_register_move_cost_if_necessary.
2513
2514 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
2515
2516         PR target/59787
2517         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
2518
2519 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
2520
2521         PR target/59794
2522         * config/i386/i386.c (type_natural_mode): Add a bool parameter
2523         to indicate if type is used for function return value.  Warn ABI
2524         change if the vector mode isn't available for function return value.
2525         (ix86_function_arg_advance): Pass false to type_natural_mode.
2526         (ix86_function_arg): Likewise.
2527         (ix86_gimplify_va_arg): Likewise.
2528         (function_arg_32): Don't warn ABI change.
2529         (ix86_function_value): Pass true to type_natural_mode.
2530         (ix86_return_in_memory): Likewise.
2531         (ix86_struct_value_rtx): Removed.
2532         (TARGET_STRUCT_VALUE_RTX): Likewise.
2533
2534 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2535
2536         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
2537         converting a conditional jump into a conditional return.
2538
2539 2014-01-14  Richard Biener  <rguenther@suse.de>
2540
2541         PR tree-optimization/58921
2542         PR tree-optimization/59006
2543         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
2544         hoisting invariant stmts.
2545         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
2546         invariant loads on the preheader edge if possible.
2547
2548 2014-01-14  Joey Ye  <joey.ye@arm.com>
2549
2550         * doc/plugin.texi (Building GCC plugins): Update to C++.
2551
2552 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
2553
2554         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
2555         (_mm_rcp28_round_ss): Ditto.
2556         (_mm_rsqrt28_round_sd): Ditto.
2557         (_mm_rsqrt28_round_ss): Ditto.
2558         (_mm_rcp28_sd): Ditto.
2559         (_mm_rcp28_ss): Ditto.
2560         (_mm_rsqrt28_sd): Ditto.
2561         (_mm_rsqrt28_ss): Ditto.
2562         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
2563         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
2564         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
2565         (IX86_BUILTIN_RCP28SD): Ditto.
2566         (IX86_BUILTIN_RCP28SS): Ditto.
2567         (IX86_BUILTIN_RSQRT28SD): Ditto.
2568         (IX86_BUILTIN_RSQRT28SS): Ditto.
2569         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
2570         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
2571         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
2572         (ix86_expand_special_args_builtin): Expand new FTYPE.
2573         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
2574         (srcp14<mode>): Make insn unary.
2575         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
2576         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
2577         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
2578         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
2579         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
2580         Fix rounding: make it SAE only.
2581         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
2582         Ditto.
2583         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
2584         Ditto.
2585         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
2586         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
2587         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
2588         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
2589         (round_saeonly_mask_scalar_operand4): Ditto.
2590         (round_saeonly_mask_scalar_op3): Ditto.
2591         (round_saeonly_mask_scalar_op4): Ditto.
2592
2593 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2594
2595         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2596         Implement -maltivec=be for vec_insert and vec_extract.
2597
2598 2014-01-10  DJ Delorie  <dj@redhat.com>
2599
2600         * config/msp430/msp430.md (call_internal): Don't allow memory
2601         references with SP as the base register.
2602         (call_value_internal): Likewise.
2603         * config/msp430/constraints.md (Yc): New.  For memory references
2604         that don't use SP as a base register.
2605
2606         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
2607         "an integer without a # prefix"
2608         * config/msp430/msp430.md (epilogue_helper): Use it.
2609
2610 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
2611
2612         PR target/59617
2613         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
2614         AVX512F gather builtins.
2615         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
2616         on gather decls with INTEGER_TYPE masktype.
2617         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
2618         directly into the builtin rather than hoisting it before loop.
2619
2620         PR tree-optimization/59387
2621         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
2622         (scev_const_prop): If folded_casts and type has undefined overflow,
2623         use force_gimple_operand instead of force_gimple_operand_gsi and
2624         for each added stmt if it is assign with
2625         arith_code_with_undefined_signed_overflow, call
2626         rewrite_to_defined_overflow.
2627         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
2628         gimple-fold.h instead.
2629         (arith_code_with_undefined_signed_overflow,
2630         rewrite_to_defined_overflow): Moved to ...
2631         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
2632         rewrite_to_defined_overflow): ... here.  No longer static.
2633         Include gimplify-me.h.
2634         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
2635         rewrite_to_defined_overflow): New prototypes.
2636
2637 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2638
2639         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
2640
2641 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
2642
2643         * builtins.c (get_object_alignment_2): Minor tweak.
2644         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
2645
2646 2014-01-13  Christian Bruel  <christian.bruel@st.com>
2647
2648         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
2649         optimized non constant lengths.
2650
2651 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
2652
2653         PR libgomp/59194
2654         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
2655         load as __atomic_load_N if possible.
2656
2657 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
2658
2659         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
2660         target parameter.
2661         (rs6000_expand_builtin): Adjust call.
2662
2663 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
2664
2665         PR target/58115
2666         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
2667         * config/rs6000/rs6000.c: Include target-globals.h.
2668         (rs6000_set_current_function): Instead of doing target_reinit
2669         unconditionally, use save_target_globals_default_opts and
2670         restore_target_globals.
2671
2672         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
2673         FPSCR.
2674         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
2675         (rs6000_expand_builtin): Handle mffs and mtfsf.
2676         (rs6000_init_builtins): Define mffs and mtfsf.
2677         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
2678         (rs6000_mffs): New pattern.
2679         (rs6000_mtfsf): New pattern.
2680
2681 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
2682
2683         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
2684         Start narrowing with START.  Apply candidate-use pair
2685         and check overall cost in narrowing.
2686         (iv_ca_prune): Pass new argument.
2687
2688 2014-01-10  Jeff Law  <law@redhat.com>
2689
2690         PR middle-end/59743
2691         * ree.c (combine_reaching_defs): Ensure the defining statement
2692         occurs before the extension when optimizing extensions with
2693         different source and destination hard registers.
2694
2695 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
2696
2697         PR ipa/58585
2698         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
2699         vtables into the type inheritance graph.
2700
2701 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
2702
2703         PR rtl-optimization/59754
2704         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
2705         modes in the REGNO != REGNO case.
2706
2707 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2708
2709         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
2710
2711 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
2712
2713         PR tree-optimization/59745
2714         * tree-predcom.c (tree_predictive_commoning_loop): Call
2715         free_affine_expand_cache if giving up because components is NULL.
2716
2717         * target-globals.c (save_target_globals): Allocate < 4KB structs using
2718         GC in payload of target_globals struct instead of allocating them on
2719         the heap and the larger structs separately using GC.
2720         * target-globals.h (struct target_globals): Make regs, hard_regs,
2721         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
2722         of GTY((skip)) and change type to void *.
2723         (reset_target_globals): Cast loads from those fields to corresponding
2724         types.
2725
2726 2014-01-10  Steve Ellcey  <sellcey@mips.com>
2727
2728         PR plugins/59335
2729         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
2730         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
2731         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
2732
2733 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
2734
2735         PR target/59744
2736         * aarch64-modes.def (CC_Zmode): New flags mode.
2737         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
2738         represents an equality.
2739         (aarch64_get_condition_code): Handle CC_Zmode.
2740         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
2741
2742 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2743
2744         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
2745         extraction in good case.
2746
2747 2014-01-10  Richard Biener  <rguenther@suse.de>
2748
2749         PR tree-optimization/59374
2750         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
2751         checking after SLP discovery.  Mark stmts not participating
2752         in any SLP instance properly.
2753
2754 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2755
2756         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
2757         when handling a SET rtx.
2758
2759 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2760
2761         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
2762         (cortex-a57): Likewise.
2763         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
2764
2765 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2766
2767         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
2768         non-iwmmxt builtins.
2769
2770 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
2771
2772         PR ipa/58252
2773         PR ipa/59226
2774         * ipa-devirt.c record_target_from_binfo): Take as argument
2775         stack of binfos and lookup matching one for virtual inheritance.
2776         (possible_polymorphic_call_targets_1): Update.
2777
2778 2014-01-10  Huacai Chen  <chenhc@lemote.com>
2779
2780         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
2781         kernel strings for Loongson-2E/2F/3A.
2782
2783 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
2784
2785         PR middle-end/59670
2786         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
2787         is_gimple_call before calling gimple_call_internal_p.
2788
2789 2014-01-09  Steve Ellcey  <sellcey@mips.com>
2790
2791         * Makefile.in (TREE_FLOW_H): Remove.
2792         (TREE_SSA_H): Add file names from tree-flow.h.
2793         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
2794         * tree.h: Remove tree-flow.h reference.
2795         * hash-table.h: Remove tree-flow.h reference.
2796         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
2797         reference with tree-ssa-loop.h.
2798
2799 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2800
2801         * doc/invoke.texi: Add -maltivec={be,le} options, and document
2802         default element-order behavior for -maltivec.
2803         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
2804         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
2805         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
2806         when targeting big endian, at least for now.
2807         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
2808
2809 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
2810
2811         PR middle-end/47735
2812         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
2813         var satisfies use_register_for_decl, just take into account type
2814         alignment, rather than decl alignment.
2815
2816         PR tree-optimization/59622
2817         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
2818         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
2819         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
2820         Don't devirtualize for inplace at all.  For targets.length () == 1,
2821         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
2822
2823 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
2824
2825         * config/i386/i386.md (cpu): Remove the unused btver1.
2826
2827 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
2828
2829         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
2830
2831 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
2832
2833         PR target/58115
2834         * tree-core.h (struct target_globals): New forward declaration.
2835         (struct tree_target_option): Add globals field.
2836         * tree.h (TREE_TARGET_GLOBALS): Define.
2837         (prepare_target_option_nodes_for_pch): New prototype.
2838         * target-globals.h (struct target_globals): Define even if
2839         !SWITCHABLE_TARGET.
2840         * tree.c (prepare_target_option_node_for_pch,
2841         prepare_target_option_nodes_for_pch): New functions.
2842         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
2843         * config/i386/i386.c: Include target-globals.h.
2844         (ix86_set_current_function): Instead of doing target_reinit
2845         unconditionally, use save_target_globals_default_opts and
2846         restore_target_globals.
2847
2848 2014-01-09  Richard Biener  <rguenther@suse.de>
2849
2850         PR tree-optimization/59715
2851         * tree-cfg.h (split_critical_edges): Declare.
2852         * tree-cfg.c (split_critical_edges): Export.
2853         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
2854
2855 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
2856
2857         * cfgexpand.c (expand_stack_vars): Optionally disable
2858         asan stack protection.
2859         (expand_used_vars): Likewise.
2860         (partition_stack_vars): Likewise.
2861         * asan.c (asan_emit_stack_protection): Optionally disable
2862         after return stack usage.
2863         (instrument_derefs): Optionally disable memory access instrumentation.
2864         (instrument_builtin_call): Likewise.
2865         (instrument_strlen_call): Likewise.
2866         (asan_protect_global): Optionally disable global variables protection.
2867         * doc/invoke.texi: Added doc for new options.
2868         * params.def: Added new options.
2869         * params.h: Likewise.
2870
2871 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
2872
2873         PR rtl-optimization/59724
2874         * ifcvt.c (cond_exec_process_if_block): Don't call
2875         flow_find_head_matching_sequence with 0 longest_match.
2876         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
2877         non-active insns if !stop_after.
2878         (try_head_merge_bb): Revert 2014-01-07 changes.
2879
2880 2014-01-08  Jeff Law  <law@redhat.com>
2881
2882         * ree.c (get_sub_rtx): New function, extracted from...
2883         (merge_def_and_ext): Here.
2884         (combine_reaching_defs): Use get_sub_rtx.
2885
2886 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
2887
2888         * cgraph.h (varpool_variable_node): Do not choke on null node.
2889
2890 2014-01-08  Catherine Moore  <clm@codesourcery.com>
2891
2892         * config/mips/mips.md (simple_return): Attempt to use JRC
2893         for microMIPS.
2894         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
2895
2896 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
2897
2898         PR rtl-optimization/59137
2899         * reorg.c (steal_delay_list_from_target): Call update_block for
2900         elided insns.
2901         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
2902
2903 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2904
2905         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2906         two duplicate entries.
2907
2908 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
2909
2910         Revert:
2911         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
2912
2913         * config/mips/mips.c (mips_truncated_op_cost): New function.
2914         (mips_rtx_costs): Adjust test for BADDU.
2915         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
2916
2917         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
2918
2919         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
2920         (*baddu_si): ...this new pattern.
2921
2922 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
2923
2924         PR ipa/59722
2925         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
2926
2927 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2928
2929         PR middle-end/57748
2930         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
2931         inner_reference_p.
2932         (expand_expr, expand_normal): Adjust.
2933         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
2934         inner_reference_p. Use inner_reference_p to expand inner references.
2935         (store_expr): Adjust.
2936         * cfgexpand.c (expand_call_stmt): Adjust.
2937
2938 2014-01-08  Rong Xu  <xur@google.com>
2939
2940         * gcov-io.c (gcov_var): Move from gcov-io.h.
2941         (gcov_position): Ditto.
2942         (gcov_is_error): Ditto.
2943         (gcov_rewrite): Ditto.
2944         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
2945         only part to libgcc/libgcov.h.
2946
2947 2014-01-08  Marek Polacek  <polacek@redhat.com>
2948
2949         PR middle-end/59669
2950         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
2951
2952 2014-01-08  Marek Polacek  <polacek@redhat.com>
2953
2954         PR sanitizer/59667
2955         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
2956
2957 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
2958
2959         PR rtl-optimization/59649
2960         * stor-layout.c (get_mode_bounds): For BImode return
2961         0 and STORE_FLAG_VALUE.
2962
2963 2014-01-08  Richard Biener  <rguenther@suse.de>
2964
2965         PR middle-end/59630
2966         * gimple.h (is_gimple_builtin_call): Remove.
2967         (gimple_builtin_call_types_compatible_p): New.
2968         (gimple_call_builtin_p): New overload.
2969         * gimple.c (is_gimple_builtin_call): Remove.
2970         (validate_call): Rename to ...
2971         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
2972         check return types.
2973         (validate_type): New static function.
2974         (gimple_call_builtin_p): New overload and adjust.
2975         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
2976         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
2977         (gimple_fold_stmt_to_constant_1): Likewise.
2978         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
2979
2980 2014-01-08  Richard Biener  <rguenther@suse.de>
2981
2982         PR middle-end/59471
2983         * gimplify.c (gimplify_expr): Gimplify register-register type
2984         VIEW_CONVERT_EXPRs to separate stmts.
2985
2986 2014-01-07  Jeff Law  <law@redhat.com>
2987
2988         PR middle-end/53623
2989         * ree.c (combine_set_extension): Handle case where source
2990         and destination registers in an extension insn are different.
2991         (combine_reaching_defs): Allow source and destination registers
2992         in extension to be different under limited circumstances.
2993         (add_removable_extension): Remove restriction that the
2994         source and destination registers in the extension are the same.
2995         (find_and_remove_re): Emit a copy from the extension's
2996         destination to its source after the defining insn if
2997         the source and destination registers are different.
2998
2999         PR middle-end/59285
3000         * ifcvt.c (merge_if_block): If we are merging a block with more than
3001         one successor with a block with no successors, remove any BARRIER
3002         after the second block.
3003
3004 2014-01-07  Dan Xio Qiang <ziyan01@163.com>
3005
3006         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
3007
3008 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
3009
3010         PR target/59652
3011         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
3012         for 14-bit register offsets when INT14_OK_STRICT is false.
3013
3014 2014-01-07  Roland Stigge  <stigge@antcom.de>
3015             Michael Meissner  <meissner@linux.vnet.ibm.com>
3016
3017         PR 57386/target
3018         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
3019         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
3020
3021 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
3022
3023         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
3024         -mcpu.
3025
3026 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
3027
3028         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
3029         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
3030         rtx is const0_rtx or not.
3031
3032 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
3033
3034         PR target/58115
3035         * target-globals.c (save_target_globals): Remove this_fn_optab
3036         handling.
3037         * toplev.c: Include optabs.h.
3038         (target_reinit): Temporarily restore the global options if another
3039         set of options are in force.
3040
3041 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
3042
3043         PR rtl-optimization/58668
3044         * cfgcleanup.c (flow_find_cross_jump): Don't count
3045         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
3046         to determine what is counted.
3047         (flow_find_head_matching_sequence): Use active_insn_p to determine
3048         what is counted.
3049         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
3050         counting change.
3051         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
3052         determine what is counted.
3053
3054         PR tree-optimization/59643
3055         * tree-predcom.c (split_data_refs_to_components): If one dr is
3056         read and one write, determine_offset fails and the write isn't
3057         in the bad component, just put the read into the bad component.
3058
3059 2014-01-07  Mike Stump  <mikestump@comcast.net>
3060             Jakub Jelinek  <jakub@redhat.com>
3061
3062         PR pch/59436
3063         * tree-core.h (struct tree_optimization_option): Change optabs
3064         type from unsigned char * to void *.
3065         * optabs.c (init_tree_optimization_optabs): Adjust
3066         TREE_OPTIMIZATION_OPTABS initialization.
3067
3068 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
3069
3070         PR target/59644
3071         * config/i386/i386.h (struct machine_function): Add
3072         no_drap_save_restore field.
3073         * config/i386/i386.c (ix86_save_reg): Use
3074         !cfun->machine->no_drap_save_restore instead of
3075         crtl->stack_realign_needed.
3076         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
3077         this function clears frame_pointer_needed.  Set
3078         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
3079         and DRAP reg is needed.
3080
3081 2014-01-06  Marek Polacek  <polacek@redhat.com>
3082
3083         PR c/57773
3084         * doc/implement-c.texi: Mention that other integer types are
3085         permitted as bit-field types in strictly conforming mode.
3086
3087 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
3088
3089         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
3090         is newly allocated.
3091
3092 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
3093
3094         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
3095
3096 2014-01-06  Martin Jambor  <mjambor@suse.cz>
3097
3098         PR ipa/59008
3099         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
3100         to int.
3101         * ipa-prop.c (ipa_print_node_params): Fix indentation.
3102
3103 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
3104
3105         PR debug/59350
3106         PR debug/59510
3107         * var-tracking.c (add_stores): Preserve the value of the source even if
3108         we don't record the store.
3109
3110 2014-01-06  Terry Guo  <terry.guo@arm.com>
3111
3112         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
3113
3114 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
3115
3116         PR bootstrap/59541
3117         * config/darwin.c (darwin_function_section): Adjust return values to
3118         correspond to optimisation changes made in r206070.
3119
3120 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
3121
3122         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
3123         from prefetch_block tune setting.
3124         (nocona_cost): Correct size of prefetch block to 64.
3125
3126 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
3127
3128         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
3129         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
3130         used to save the static chain register in the computation of the offset
3131         from which the FP registers need to be restored.
3132
3133 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
3134
3135         PR tree-optimization/59519
3136         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
3137         ICE if get_current_def (current_new_name) is already non-NULL, as long
3138         as it is a phi result of some other phi in *new_exit_bb that has
3139         the same argument.
3140
3141         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
3142         or vmovdqu* for misaligned_operand.
3143         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
3144         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
3145         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
3146         aligned_mem for AVX512F masked aligned load and store builtins and for
3147         non-temporal moves.
3148
3149 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
3150
3151         PR tree-optimization/59651
3152         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
3153         Address range for negative step should be added by TYPE_SIZE_UNIT.
3154
3155 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
3156
3157         * config/m68k/m68k.c (handle_move_double): Handle pushes with
3158         overlapping registers also for registers other than the stack pointer.
3159
3160 2014-01-03  Marek Polacek  <polacek@redhat.com>
3161
3162         PR other/59661
3163         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
3164         __builtin_FILE.
3165
3166 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
3167
3168         PR target/59625
3169         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
3170         asm goto as jump.
3171
3172         * config/i386/i386.md (MODE_SIZE): New mode attribute.
3173         (push splitter): Use <P:MODE_SIZE> instead of
3174         GET_MODE_SIZE (<P:MODE>mode).
3175         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
3176         (mov -1, reg peephole2): Likewise.
3177         * config/i386/sse.md (*mov<mode>_internal,
3178         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
3179         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
3180         *<code><mode>3, *andnot<mode>3<mask_name>,
3181         <mask_codefor><code><mode>3<mask_name>): Likewise.
3182         * config/i386/subst.md (mask_mode512bit_condition,
3183         sd_mask_mode512bit_condition): Likewise.
3184
3185 2014-01-02  Xinliang David Li  <davidxl@google.com>
3186
3187         PR tree-optimization/59303
3188         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
3189         (dump_predicates): Better output format.
3190         (pred_equal_p): New function.
3191         (is_neq_relop_p): Ditto.
3192         (is_neq_zero_form_p): Ditto.
3193         (pred_expr_equal_p): Ditto.
3194         (pred_neg_p): Ditto.
3195         (simplify_pred): Ditto.
3196         (simplify_preds_2): Ditto.
3197         (simplify_preds_3): Ditto.
3198         (simplify_preds_4): Ditto.
3199         (simplify_preds): Ditto.
3200         (push_pred): Ditto.
3201         (push_to_worklist): Ditto.
3202         (get_pred_info_from_cmp): Ditto.
3203         (is_degenerated_phi): Ditto.
3204         (normalize_one_pred_1): Ditto.
3205         (normalize_one_pred): Ditto.
3206         (normalize_one_pred_chain): Ditto.
3207         (normalize_preds): Ditto.
3208         (normalize_cond_1): Remove function.
3209         (normalize_cond): Ditto.
3210         (is_gcond_subset_of): Ditto.
3211         (is_subset_of_any): Ditto.
3212         (is_or_set_subset_of): Ditto.
3213         (is_and_set_subset_of): Ditto.
3214         (is_norm_cond_subset_of): Ditto.
3215         (pred_chain_length_cmp): Ditto.
3216         (convert_control_dep_chain_into_preds): Type change.
3217         (find_predicates): Ditto.
3218         (find_def_preds): Ditto.
3219         (destroy_predicates_vecs): Ditto.
3220         (find_matching_predicates_in_rest_chains): Ditto.
3221         (use_pred_not_overlap_with_undef_path_pred): Ditto.
3222         (is_pred_expr_subset): Ditto.
3223         (is_pred_chain_subset_of): Ditto.
3224         (is_included_in): Ditto.
3225         (is_superset_of): Ditto.
3226
3227 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
3228
3229         Update copyright years.
3230
3231 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
3232
3233         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
3234         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
3235         config/arc/arc.md, config/arc/arc.opt,
3236         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
3237         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
3238         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
3239         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
3240         config/linux-protos.h, config/linux.c, config/winnt-c.c,
3241         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
3242         vtable-verify.c, vtable-verify.h: Use the standard form for the
3243         copyright notice.
3244
3245 2014-01-02  Tobias Burnus  <burnus@net-b.de>
3246
3247         * gcc.c (process_command): Update copyright notice dates.
3248         * gcov-dump.c: Ditto.
3249         * gcov.c: Ditto.
3250         * doc/cpp.texi: Bump @copying's copyright year.
3251         * doc/cppinternals.texi: Ditto.
3252         * doc/gcc.texi: Ditto.
3253         * doc/gccint.texi: Ditto.
3254         * doc/gcov.texi: Ditto.
3255         * doc/install.texi: Ditto.
3256         * doc/invoke.texi: Ditto.
3257
3258 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
3259
3260         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
3261
3262 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
3263
3264         * config/i386/sse.md (*mov<mode>_internal): Guard
3265         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
3266
3267         PR rtl-optimization/59647
3268         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
3269         new_rtx into UNSIGNED_FLOAT rtxes.
3270 \f
3271 Copyright (C) 2014 Free Software Foundation, Inc.
3272
3273 Copying and distribution of this file, with or without modification,
3274 are permitted in any medium without royalty provided the copyright
3275 notice and this notice are preserved.