gcc/ChangeLog

   1 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
   2
   3         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
   4         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
   5         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
   6         (intel_memset): New.  Duplicate slm_memset.
   7         (intel_cost): New.  Duplicate slm_cost.
   8         (m_INTEL): New macro.
   9         (processor_target_table): Add "intel".
  10         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
  11         with PROCESSOR_INTEL for "intel".
  12         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
  13         PROCESSOR_SILVERMONT.
  14         (ix86_issue_rate): Likewise.
  15         (ix86_adjust_cost): Likewise.
  16         (ia32_multipass_dfa_lookahead): Likewise.
  17         (swap_top_of_ready_list): Likewise.
  18         (ix86_sched_reorder): Likewise.
  19         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
  20         instead of TARGET_OPT_AGU.
  21         * config/i386/i386.h (TARGET_INTEL): New.
  22         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
  23         (processor_type): Add PROCESSOR_INTEL.
  24         * config/i386/x86-tune.def: Support m_INTEL. Duplicate
  25         m_SILVERMONT.  Add X86_TUNE_AVOID_LEA_FOR_ADDR.
  26
  27 2014-01-17  Marek Polacek  <polacek@redhat.com>
  28
  29         PR c/58346
  30         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
  31         size is zero.
  32
  33 2014-01-17  Richard Biener  <rguenther@suse.de>
  34
  35         PR tree-optimization/46590
  36         * opts.c (default_options_table): Add entries for
  37         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
  38         all enabled at -O1 but not for -Og.
  39         * common.opt (fbranch-count-reg): Remove Init(1).
  40         (fmove-loop-invariants): Likewise.
  41         (ftree-pta): Likewise.
  42
  43 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
  44
  45         * config/i386/i386.c (ix86_data_alignment): For compatibility with
  46         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
  47         decls to at least the GCC 4.8 used alignments.
  48
  49         PR fortran/59440
  50         * tree-nested.c (convert_nonlocal_reference_stmt,
  51         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
  52         of GIMPLE_BIND stmts, adjust associated decls.
  53
  54 2014-01-17  Richard Biener  <rguenther@suse.de>
  55
  56         PR tree-optimization/46590
  57         * vec.h (vec<>::bseach): New member function implementing
  58         binary search according to C89 bsearch.
  59         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
  60         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
  61         bitmap pointer again.  Make accesses_in_loop a flat array.
  62         (mem_ref_obstack): New global.
  63         (outermost_indep_loop): Adjust for mem_ref->stored changes.
  64         (mark_ref_stored): Likewise.
  65         (ref_indep_loop_p_2): Likewise.
  66         (set_ref_stored_in_loop): New helper function.
  67         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
  68         (memref_free): Adjust.
  69         (record_mem_ref_loc): Simplify.
  70         (gather_mem_refs_stmt): Adjust.
  71         (sort_locs_in_loop_postorder_cmp): New function.
  72         (analyze_memory_references): Sort accesses_in_loop after
  73         loop postorder number.
  74         (find_ref_loc_in_loop_cmp): New function.
  75         (for_all_locs_in_loop): Find relevant cluster of locs in
  76         accesses_in_loop and iterate without recursion.
  77         (execute_sm): Avoid uninit warning.
  78         (struct ref_always_accessed): Simplify.
  79         (ref_always_accessed::operator ()): Likewise.
  80         (ref_always_accessed_p): Likewise.
  81         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
  82         loop postorder numbers here.
  83         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
  84         numbers.
  85
  86 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
  87
  88         PR c++/57945
  89         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
  90         on decls for which assemble_alias has been called.
  91
  92 2014-01-17  Nick Clifton  <nickc@redhat.com>
  93
  94         * config/msp430/msp430.opt: (mcpu): New option.
  95         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
  96         (msp430_option_override): Parse target_cpu.  If the MCU name
  97         matches a generic string, clear target_mcu.
  98         (msp430_attr): Allow numeric interrupt values up to 63.
  99         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
 100         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
 101         option.
 102         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
 103         Add mcpu matches.
 104         * config/msp430/msp430.md (popm): Use %J rather than %I.
 105         (addsi3): Use msp430_nonimmediate_operand for operand 2.
 106         (addhi_cy_i): Use immediate_operand for operand 2.
 107         * doc/invoke.texi: Document -mcpu option.
 108
 109 2014-01-17  Richard Biener  <rguenther@suse.de>
 110
 111         PR rtl-optimization/38518
 112         * df.h (df_analyze_loop): Declare.
 113         * df-core.c: Include cfgloop.h.
 114         (df_analyze_1): Split out main part of df_analyze.
 115         (df_analyze): Adjust.
 116         (loop_inverted_post_order_compute): New function.
 117         (loop_post_order_compute): Likewise.
 118         (df_analyze_loop): New function avoiding whole-function
 119         postorder computes.
 120         * loop-invariant.c (find_defs): Use df_analyze_loop.
 121         (find_invariants): Adjust.
 122         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
 123
 124 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
 125
 126         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
 127         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
 128
 129 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
 130
 131         * ipa-ref.c (ipa_remove_stmt_references): Fix references
 132         traversal when removing references.
 133
 134 2014-01-16  Jan Hubicka  <jh@suse.cz>
 135
 136         PR ipa/59775
 137         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
 138
 139 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
 140
 141         PR middle-end/56791
 142         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
 143         pushing a reload for an autoinc when we had previously reloaded an
 144         inner part of the address.
 145
 146 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
 147
 148         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
 149         field.
 150         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
 151         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
 152         when not giving up or versioning for alias only because of
 153         loop->safelen.
 154         (vect_analyze_data_ref_dependences): Set to true.
 155         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
 156         is a GIMPLE_PHI.
 157         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
 158         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
 159         to the condition.
 160
 161         PR middle-end/58344
 162         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
 163
 164         PR target/59839
 165         * config/i386/i386.c (ix86_expand_builtin): If target doesn't
 166         satisfy operand 0 predicate for gathers, use a new pseudo as
 167         subtarget.
 168
 169 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
 170
 171         PR middle-end/59609
 172         * lra-constraints.c (process_alt_operands): Add printing debug
 173         info.  Check absence of input/output reloads for matched operands
 174         too.
 175
 176 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
 177
 178         PR rtl-optimization/59835
 179         * ira.c (ira_init_register_move_cost): Increase cost for
 180         impossible modes.
 181
 182 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
 183
 184         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
 185
 186 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
 187
 188         PR target/59780
 189         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
 190         non-register objects.  Use gen_(high/low)part more consistently.
 191         Fix assertions.
 192
 193 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
 194
 195         PR target/59844
 196         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
 197         endian support, remove tests for WORDS_BIG_ENDIAN.
 198         (p8_mfvsrd_3_<mode>): Likewise.
 199         (reload_gpr_from_vsx<mode>): Likewise.
 200         (reload_gpr_from_vsxsf): Likewise.
 201         (p8_mfvsrd_4_disf): Likewise.
 202
 203 2014-01-16  Richard Biener  <rguenther@suse.de>
 204
 205         PR rtl-optimization/46590
 206         * lcm.c (compute_antinout_edge): Use postorder iteration.
 207         (compute_laterin): Use inverted postorder iteration.
 208
 209 2014-01-16  Nick Clifton  <nickc@redhat.com>
 210
 211         PR middle-end/28865
 212         * varasm.c (output_constant): Return the number of bytes actually
 213         emitted.
 214         (output_constructor_array_range): Update the field size with the
 215         number of bytes emitted by output_constant.
 216         (output_constructor_regular_field): Likewise.  Also do not
 217         complain if the total number of bytes emitted is now greater
 218         than the expected fieldpos.
 219         * output.h (output_constant): Update prototype and descriptive
 220         comment.
 221
 222 2014-01-16  Marek Polacek  <polacek@redhat.com>
 223
 224         PR middle-end/59827
 225         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
 226         it is error_mark_node.
 227
 228 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
 229
 230         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
 231         VALID_AVX256_REG_OR_OI_MODE.
 232
 233 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
 234
 235         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
 236         current procedure should be profiled.
 237
 238 2014-01-15  Andrew Pinski  <apinski@cavium.com>
 239
 240         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
 241         of moving from/to the STACK_REG register class.
 242
 243 2014-01-15  Richard Henderson <rth@redhat.com>
 244
 245         PR debug/54694
 246         * reginfo.c (global_regs_decl): Globalize.
 247         * rtl.h (global_regs_decl): Declare.
 248         * ira.c (do_reload): Diagnose frame_pointer_needed and it
 249         reserved via global_regs.
 250
 251 2014-01-15  Teresa Johnson  <tejohnson@google.com>
 252
 253         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
 254
 255 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
 256
 257         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
 258         and vmulosh rather than call gen_vec_widen_smult_*.
 259         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
 260         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
 261         (vec_widen_smult_even_v16qi): Likewise.
 262         (vec_widen_umult_even_v8hi): Likewise.
 263         (vec_widen_smult_even_v8hi): Likewise.
 264         (vec_widen_umult_odd_v16qi): Likewise.
 265         (vec_widen_smult_odd_v16qi): Likewise.
 266         (vec_widen_umult_odd_v8hi): Likewise.
 267         (vec_widen_smult_odd_v8hi): Likewise.
 268         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
 269         vmuloub rather than call gen_vec_widen_umult_*.
 270         (vec_widen_umult_lo_v16qi): Likewise.
 271         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
 272         vmulosb rather than call gen_vec_widen_smult_*.
 273         (vec_widen_smult_lo_v16qi): Likewise.
 274         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
 275         rather than call gen_vec_widen_umult_*.
 276         (vec_widen_umult_lo_v8hi): Likewise.
 277         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
 278         rather than call gen_vec_widen_smult_*.
 279         (vec_widen_smult_lo_v8hi): Likewise.
 280
 281 2014-01-15  Jeff Law  <law@redhat.com>
 282
 283         PR tree-optimization/59747
 284         * ree.c (find_and_remove_re): Properly handle case where a second
 285         eliminated extension requires widening a copy created for elimination
 286         of a prior extension.
 287         (combine_set_extension): Ensure that the number of hard regs needed
 288         for a destination register does not change when we widen it.
 289
 290 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
 291
 292         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
 293         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
 294         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
 295         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
 296         (avr-*-rtems*): Likewise.
 297         (bfin*-rtems*): Likewise.
 298         (moxie-*-rtems*): Likewise.
 299         (h8300-*-rtems*): Likewise.
 300         (i[34567]86-*-rtems*): Likewise.
 301         (lm32-*-rtems*): Likewise.
 302         (m32r-*-rtems*): Likewise.
 303         (m68k-*-rtems*): Likewise.
 304         (microblaze*-*-rtems*): Likewise.
 305         (mips*-*-rtems*): Likewise.
 306         (powerpc-*-rtems*): Likewise.
 307         (sh-*-rtems*): Likewise.
 308         (sparc-*-rtems*): Likewise.
 309         (sparc64-*-rtems*): Likewise.
 310         (v850-*-rtems*): Likewise.
 311         (m32c-*-rtems*): Likewise.
 312
 313 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
 314
 315         PR rtl-optimization/59511
 316         * ira.c (ira_init_register_move_cost): Use memory costs for some
 317         cases of register move cost calculations.
 318         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
 319         instead of BB frequency.
 320         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
 321         * lra-assigns.c (find_hard_regno_for): Ditto.
 322
 323 2014-01-15  Richard Biener  <rguenther@suse.de>
 324
 325         PR tree-optimization/59822
 326         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
 327         (vectorizable_load): Use it to hoist defs of uses of invariant
 328         loads out of the loop.
 329
 330 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
 331             Kugan Vivekanandarajah  <kuganv@linaro.org>
 332
 333         PR target/59695
 334         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
 335         truncation.
 336
 337 2014-01-15  Richard Biener  <rguenther@suse.de>
 338
 339         PR rtl-optimization/59802
 340         * lcm.c (compute_available): Use inverted postorder to seed
 341         the initial worklist.
 342
 343 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
 344
 345         PR target/59803
 346         * config/s390/s390.c (s390_preferred_reload_class): Don't return
 347         ADDR_REGS for invalid symrefs in non-PIC code.
 348
 349 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
 350
 351         PR other/58712
 352         * builtins.c (determine_block_size): Initialize *probable_max_size
 353         even if len_rtx is CONST_INT.
 354
 355 2014-01-14  Andrew Pinski  <apinski@cavium.com>
 356
 357         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
 358         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
 359         (cortexa53_tunings): Likewise.
 360         (aarch64_sched_issue_rate): New function.
 361         (TARGET_SCHED_ISSUE_RATE): Define.
 362
 363 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
 364
 365         * ira-costs.c (find_costs_and_classes): Add missed
 366         ira_init_register_move_cost_if_necessary.
 367
 368 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
 369
 370         PR target/59787
 371         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
 372
 373 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
 374
 375         PR target/59794
 376         * config/i386/i386.c (type_natural_mode): Add a bool parameter
 377         to indicate if type is used for function return value.  Warn ABI
 378         change if the vector mode isn't available for function return value.
 379         (ix86_function_arg_advance): Pass false to type_natural_mode.
 380         (ix86_function_arg): Likewise.
 381         (ix86_gimplify_va_arg): Likewise.
 382         (function_arg_32): Don't warn ABI change.
 383         (ix86_function_value): Pass true to type_natural_mode.
 384         (ix86_return_in_memory): Likewise.
 385         (ix86_struct_value_rtx): Removed.
 386         (TARGET_STRUCT_VALUE_RTX): Likewise.
 387
 388 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
 389
 390         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
 391         converting a conditional jump into a conditional return.
 392
 393 2014-01-14  Richard Biener  <rguenther@suse.de>
 394
 395         PR tree-optimization/58921
 396         PR tree-optimization/59006
 397         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
 398         hoisting invariant stmts.
 399         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
 400         invariant loads on the preheader edge if possible.
 401
 402 2014-01-14  Joey Ye  <joey.ye@arm.com>
 403
 404         * doc/plugin.texi (Building GCC plugins): Update to C++.
 405
 406 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
 407
 408         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
 409         (_mm_rcp28_round_ss): Ditto.
 410         (_mm_rsqrt28_round_sd): Ditto.
 411         (_mm_rsqrt28_round_ss): Ditto.
 412         (_mm_rcp28_sd): Ditto.
 413         (_mm_rcp28_ss): Ditto.
 414         (_mm_rsqrt28_sd): Ditto.
 415         (_mm_rsqrt28_ss): Ditto.
 416         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
 417         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
 418         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
 419         (IX86_BUILTIN_RCP28SD): Ditto.
 420         (IX86_BUILTIN_RCP28SS): Ditto.
 421         (IX86_BUILTIN_RSQRT28SD): Ditto.
 422         (IX86_BUILTIN_RSQRT28SS): Ditto.
 423         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
 424         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
 425         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
 426         (ix86_expand_special_args_builtin): Expand new FTYPE.
 427         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
 428         (srcp14<mode>): Make insn unary.
 429         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
 430         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
 431         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
 432         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
 433         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
 434         Fix rounding: make it SAE only.
 435         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
 436         Ditto.
 437         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
 438         Ditto.
 439         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
 440         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
 441         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
 442         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
 443         (round_saeonly_mask_scalar_operand4): Ditto.
 444         (round_saeonly_mask_scalar_op3): Ditto.
 445         (round_saeonly_mask_scalar_op4): Ditto.
 446
 447 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
 448
 449         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
 450         Implement -maltivec=be for vec_insert and vec_extract.
 451
 452 2014-01-10  DJ Delorie  <dj@redhat.com>
 453
 454         * config/msp430/msp430.md (call_internal): Don't allow memory
 455         references with SP as the base register.
 456         (call_value_internal): Likewise.
 457         * config/msp430/constraints.md (Yc): New.  For memory references
 458         that don't use SP as a base register.
 459
 460         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
 461         "an integer without a # prefix"
 462         * config/msp430/msp430.md (epilogue_helper): Use it.
 463
 464 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
 465
 466         PR target/59617
 467         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
 468         AVX512F gather builtins.
 469         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
 470         on gather decls with INTEGER_TYPE masktype.
 471         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
 472         directly into the builtin rather than hoisting it before loop.
 473
 474         PR tree-optimization/59387
 475         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
 476         (scev_const_prop): If folded_casts and type has undefined overflow,
 477         use force_gimple_operand instead of force_gimple_operand_gsi and
 478         for each added stmt if it is assign with
 479         arith_code_with_undefined_signed_overflow, call
 480         rewrite_to_defined_overflow.
 481         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
 482         gimple-fold.h instead.
 483         (arith_code_with_undefined_signed_overflow,
 484         rewrite_to_defined_overflow): Moved to ...
 485         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
 486         rewrite_to_defined_overflow): ... here.  No longer static.
 487         Include gimplify-me.h.
 488         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
 489         rewrite_to_defined_overflow): New prototypes.
 490
 491 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 492
 493         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
 494
 495 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
 496
 497         * builtins.c (get_object_alignment_2): Minor tweak.
 498         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
 499
 500 2014-01-13  Christian Bruel  <christian.bruel@st.com>
 501
 502         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
 503         optimized non constant lengths.
 504
 505 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
 506
 507         PR libgomp/59194
 508         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
 509         load as __atomic_load_N if possible.
 510
 511 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
 512
 513         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
 514         target parameter.
 515         (rs6000_expand_builtin): Adjust call.
 516
 517 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
 518
 519         PR target/58115
 520         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
 521         * config/rs6000/rs6000.c: Include target-globals.h.
 522         (rs6000_set_current_function): Instead of doing target_reinit
 523         unconditionally, use save_target_globals_default_opts and
 524         restore_target_globals.
 525
 526         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
 527         FPSCR.
 528         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
 529         (rs6000_expand_builtin): Handle mffs and mtfsf.
 530         (rs6000_init_builtins): Define mffs and mtfsf.
 531         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
 532         (rs6000_mffs): New pattern.
 533         (rs6000_mtfsf): New pattern.
 534
 535 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
 536
 537         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
 538         Start narrowing with START.  Apply candidate-use pair
 539         and check overall cost in narrowing.
 540         (iv_ca_prune): Pass new argument.
 541
 542 2014-01-10  Jeff Law  <law@redhat.com>
 543
 544         PR middle-end/59743
 545         * ree.c (combine_reaching_defs): Ensure the defining statement
 546         occurs before the extension when optimizing extensions with
 547         different source and destination hard registers.
 548
 549 2014-01-10  Jan Hubicka  <jh@suse.cz>
 550
 551         PR ipa/58585
 552         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
 553         vtables into the type inheritance graph.
 554
 555 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
 556
 557         PR rtl-optimization/59754
 558         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
 559         modes in the REGNO != REGNO case.
 560
 561 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
 562
 563         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
 564
 565 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
 566
 567         PR tree-optimization/59745
 568         * tree-predcom.c (tree_predictive_commoning_loop): Call
 569         free_affine_expand_cache if giving up because components is NULL.
 570
 571         * target-globals.c (save_target_globals): Allocate < 4KB structs using
 572         GC in payload of target_globals struct instead of allocating them on
 573         the heap and the larger structs separately using GC.
 574         * target-globals.h (struct target_globals): Make regs, hard_regs,
 575         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
 576         of GTY((skip)) and change type to void *.
 577         (reset_target_globals): Cast loads from those fields to corresponding
 578         types.
 579
 580 2014-01-10  Steve Ellcey  <sellcey@mips.com>
 581
 582         PR plugins/59335
 583         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
 584         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
 585         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
 586
 587 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
 588
 589         PR target/59744
 590         * aarch64-modes.def (CC_Zmode): New flags mode.
 591         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
 592         represents an equality.
 593         (aarch64_get_condition_code): Handle CC_Zmode.
 594         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
 595
 596 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
 597
 598         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
 599         extraction in good case.
 600
 601 2014-01-10  Richard Biener  <rguenther@suse.de>
 602
 603         PR tree-optimization/59374
 604         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
 605         checking after SLP discovery.  Mark stmts not participating
 606         in any SLP instance properly.
 607
 608 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 609
 610         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
 611         when handling a SET rtx.
 612
 613 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 614
 615         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
 616         (cortex-a57): Likewise.
 617         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
 618
 619 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 620
 621         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
 622         non-iwmmxt builtins.
 623
 624 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
 625
 626         PR ipa/58252
 627         PR ipa/59226
 628         * ipa-devirt.c record_target_from_binfo): Take as argument
 629         stack of binfos and lookup matching one for virtual inheritance.
 630         (possible_polymorphic_call_targets_1): Update.
 631
 632 2014-01-10  Huacai Chen  <chenhc@lemote.com>
 633
 634         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
 635         kernel strings for Loongson-2E/2F/3A.
 636
 637 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
 638
 639         PR middle-end/59670
 640         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
 641         is_gimple_call before calling gimple_call_internal_p.
 642
 643 2014-01-09  Steve Ellcey  <sellcey@mips.com>
 644
 645         * Makefile.in (TREE_FLOW_H): Remove.
 646         (TREE_SSA_H): Add file names from tree-flow.h.
 647         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
 648         * tree.h: Remove tree-flow.h reference.
 649         * hash-table.h: Remove tree-flow.h reference.
 650         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
 651         reference with tree-ssa-loop.h.
 652
 653 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
 654
 655         * doc/invoke.texi: Add -maltivec={be,le} options, and document
 656         default element-order behavior for -maltivec.
 657         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
 658         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
 659         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
 660         when targeting big endian, at least for now.
 661         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
 662
 663 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
 664
 665         PR middle-end/47735
 666         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
 667         var satisfies use_register_for_decl, just take into account type
 668         alignment, rather than decl alignment.
 669
 670         PR tree-optimization/59622
 671         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
 672         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
 673         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
 674         Don't devirtualize for inplace at all.  For targets.length () == 1,
 675         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
 676
 677 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
 678
 679         * config/i386/i386.md (cpu): Remove the unused btver1.
 680
 681 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
 682
 683         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
 684
 685 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
 686
 687         PR target/58115
 688         * tree-core.h (struct target_globals): New forward declaration.
 689         (struct tree_target_option): Add globals field.
 690         * tree.h (TREE_TARGET_GLOBALS): Define.
 691         (prepare_target_option_nodes_for_pch): New prototype.
 692         * target-globals.h (struct target_globals): Define even if
 693         !SWITCHABLE_TARGET.
 694         * tree.c (prepare_target_option_node_for_pch,
 695         prepare_target_option_nodes_for_pch): New functions.
 696         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
 697         * config/i386/i386.c: Include target-globals.h.
 698         (ix86_set_current_function): Instead of doing target_reinit
 699         unconditionally, use save_target_globals_default_opts and
 700         restore_target_globals.
 701
 702 2014-01-09  Richard Biener  <rguenther@suse.de>
 703
 704         PR tree-optimization/59715
 705         * tree-cfg.h (split_critical_edges): Declare.
 706         * tree-cfg.c (split_critical_edges): Export.
 707         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
 708
 709 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
 710
 711         * cfgexpand.c (expand_stack_vars): Optionally disable
 712         asan stack protection.
 713         (expand_used_vars): Likewise.
 714         (partition_stack_vars): Likewise.
 715         * asan.c (asan_emit_stack_protection): Optionally disable
 716         after return stack usage.
 717         (instrument_derefs): Optionally disable memory
 718         access instrumentation.
 719         (instrument_builtin_call): Likewise.
 720         (instrument_strlen_call): Likewise.
 721         (asan_protect_global): Optionally disable
 722         global variables protection.
 723         * doc/invoke.texi: Added doc for new options.
 724         * params.def: Added new options.
 725         * params.h: Likewise.
 726
 727 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
 728
 729         PR rtl-optimization/59724
 730         * ifcvt.c (cond_exec_process_if_block): Don't call
 731         flow_find_head_matching_sequence with 0 longest_match.
 732         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
 733         non-active insns if !stop_after.
 734         (try_head_merge_bb): Revert 2014-01-07 changes.
 735
 736 2014-01-08  Jeff Law  <law@redhat.com>
 737
 738         * ree.c (get_sub_rtx): New function, extracted from...
 739         (merge_def_and_ext): Here.
 740         (combine_reaching_defs): Use get_sub_rtx.
 741
 742 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
 743
 744         * cgraph.h (varpool_variable_node): Do not choke on null node.
 745
 746 2014-01-08  Catherine Moore  <clm@codesourcery.com>
 747
 748         * config/mips/mips.md (simple_return): Attempt to use JRC
 749         for microMIPS.
 750         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
 751
 752 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
 753
 754         PR rtl-optimization/59137
 755         * reorg.c (steal_delay_list_from_target): Call update_block for
 756         elided insns.
 757         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
 758
 759 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
 760
 761         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
 762         two duplicate entries.
 763
 764 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
 765
 766         Revert:
 767         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
 768
 769         * config/mips/mips.c (mips_truncated_op_cost): New function.
 770         (mips_rtx_costs): Adjust test for BADDU.
 771         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
 772
 773         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
 774
 775         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
 776         (*baddu_si): ...this new pattern.
 777
 778 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
 779
 780         PR ipa/59722
 781         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
 782
 783 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
 784
 785         PR middle-end/57748
 786         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
 787         inner_reference_p.
 788         (expand_expr, expand_normal): Adjust.
 789         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
 790         inner_reference_p. Use inner_reference_p to expand inner references.
 791         (store_expr): Adjust.
 792         * cfgexpand.c (expand_call_stmt): Adjust.
 793
 794 2014-01-08  Rong Xu  <xur@google.com>
 795
 796         * gcov-io.c (gcov_var): Move from gcov-io.h.
 797         (gcov_position): Ditto.
 798         (gcov_is_error): Ditto.
 799         (gcov_rewrite): Ditto.
 800         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
 801         only part to libgcc/libgcov.h.
 802
 803 2014-01-08  Marek Polacek  <polacek@redhat.com>
 804
 805         PR middle-end/59669
 806         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
 807
 808 2014-01-08  Marek Polacek  <polacek@redhat.com>
 809
 810         PR sanitizer/59667
 811         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
 812
 813 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
 814
 815         PR rtl-optimization/59649
 816         * stor-layout.c (get_mode_bounds): For BImode return
 817         0 and STORE_FLAG_VALUE.
 818
 819 2014-01-08  Richard Biener  <rguenther@suse.de>
 820
 821         PR middle-end/59630
 822         * gimple.h (is_gimple_builtin_call): Remove.
 823         (gimple_builtin_call_types_compatible_p): New.
 824         (gimple_call_builtin_p): New overload.
 825         * gimple.c (is_gimple_builtin_call): Remove.
 826         (validate_call): Rename to ...
 827         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
 828         check return types.
 829         (validate_type): New static function.
 830         (gimple_call_builtin_p): New overload and adjust.
 831         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
 832         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
 833         (gimple_fold_stmt_to_constant_1): Likewise.
 834         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
 835
 836 2014-01-08  Richard Biener  <rguenther@suse.de>
 837
 838         PR middle-end/59471
 839         * gimplify.c (gimplify_expr): Gimplify register-register type
 840         VIEW_CONVERT_EXPRs to separate stmts.
 841
 842 2014-01-07  Jeff Law  <law@redhat.com>
 843
 844         PR middle-end/53623
 845         * ree.c (combine_set_extension): Handle case where source
 846         and destination registers in an extension insn are different.
 847         (combine_reaching_defs): Allow source and destination registers
 848         in extension to be different under limited circumstances.
 849         (add_removable_extension): Remove restriction that the
 850         source and destination registers in the extension are the same.
 851         (find_and_remove_re): Emit a copy from the extension's
 852         destination to its source after the defining insn if
 853         the source and destination registers are different.
 854
 855         PR middle-end/59285
 856         * ifcvt.c (merge_if_block): If we are merging a block with more than
 857         one successor with a block with no successors, remove any BARRIER
 858         after the second block.
 859
 860 2014-01-07  Dan Xio Qiang <ziyan01@163.com>
 861
 862         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
 863
 864 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
 865
 866         PR target/59652
 867         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
 868         for 14-bit register offsets when INT14_OK_STRICT is false.
 869
 870 2014-01-07  Roland Stigge  <stigge@antcom.de>
 871             Michael Meissner  <meissner@linux.vnet.ibm.com>
 872
 873         PR 57386/target
 874         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
 875         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
 876
 877 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
 878
 879         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
 880         -mcpu.
 881
 882 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
 883
 884         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
 885         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
 886         rtx is const0_rtx or not.
 887
 888 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
 889
 890         PR target/58115
 891         * target-globals.c (save_target_globals): Remove this_fn_optab
 892         handling.
 893         * toplev.c: Include optabs.h.
 894         (target_reinit): Temporarily restore the global options if another
 895         set of options are in force.
 896
 897 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
 898
 899         PR rtl-optimization/58668
 900         * cfgcleanup.c (flow_find_cross_jump): Don't count
 901         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
 902         to determine what is counted.
 903         (flow_find_head_matching_sequence): Use active_insn_p to determine
 904         what is counted.
 905         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
 906         counting change.
 907         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
 908         determine what is counted.
 909
 910         PR tree-optimization/59643
 911         * tree-predcom.c (split_data_refs_to_components): If one dr is
 912         read and one write, determine_offset fails and the write isn't
 913         in the bad component, just put the read into the bad component.
 914
 915 2014-01-07  Mike Stump  <mikestump@comcast.net>
 916             Jakub Jelinek  <jakub@redhat.com>
 917
 918         PR pch/59436
 919         * tree-core.h (struct tree_optimization_option): Change optabs
 920         type from unsigned char * to void *.
 921         * optabs.c (init_tree_optimization_optabs): Adjust
 922         TREE_OPTIMIZATION_OPTABS initialization.
 923
 924 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
 925
 926         PR target/59644
 927         * config/i386/i386.h (struct machine_function): Add
 928         no_drap_save_restore field.
 929         * config/i386/i386.c (ix86_save_reg): Use
 930         !cfun->machine->no_drap_save_restore instead of
 931         crtl->stack_realign_needed.
 932         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
 933         this function clears frame_pointer_needed.  Set
 934         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
 935         and DRAP reg is needed.
 936
 937 2014-01-06  Marek Polacek  <polacek@redhat.com>
 938
 939         PR c/57773
 940         * doc/implement-c.texi: Mention that other integer types are
 941         permitted as bit-field types in strictly conforming mode.
 942
 943 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
 944
 945         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
 946         is newly allocated.
 947
 948 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
 949
 950         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
 951
 952 2014-01-06  Martin Jambor  <mjambor@suse.cz>
 953
 954         PR ipa/59008
 955         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
 956         to int.
 957         * ipa-prop.c (ipa_print_node_params): Fix indentation.
 958
 959 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
 960
 961         PR debug/59350
 962         PR debug/59510
 963         * var-tracking.c (add_stores): Preserve the value of the source even if
 964         we don't record the store.
 965
 966 2014-01-06  Terry Guo  <terry.guo@arm.com>
 967
 968         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
 969
 970 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
 971
 972         PR bootstrap/59541
 973         * config/darwin.c (darwin_function_section): Adjust return values to
 974         correspond to optimisation changes made in r206070.
 975
 976 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
 977
 978         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
 979         from prefetch_block tune setting.
 980         (nocona_cost): Correct size of prefetch block to 64.
 981
 982 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
 983
 984         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
 985         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
 986         used to save the static chain register in the computation of the offset
 987         from which the FP registers need to be restored.
 988
 989 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
 990
 991         PR tree-optimization/59519
 992         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
 993         ICE if get_current_def (current_new_name) is already non-NULL, as long
 994         as it is a phi result of some other phi in *new_exit_bb that has
 995         the same argument.
 996
 997         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
 998         or vmovdqu* for misaligned_operand.
 999         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
1000         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
1001         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
1002         aligned_mem for AVX512F masked aligned load and store builtins and for
1003         non-temporal moves.
1004
1005 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
1006
1007         PR tree-optimization/59651
1008         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
1009         Address range for negative step should be added by TYPE_SIZE_UNIT.
1010
1011 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
1012
1013         * config/m68k/m68k.c (handle_move_double): Handle pushes with
1014         overlapping registers also for registers other than the stack
1015         pointer.
1016
1017 2014-01-03  Marek Polacek  <polacek@redhat.com>
1018
1019         PR other/59661
1020         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
1021         __builtin_FILE.
1022
1023 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
1024
1025         PR target/59625
1026         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
1027         asm goto as jump.
1028
1029         * config/i386/i386.md (MODE_SIZE): New mode attribute.
1030         (push splitter): Use <P:MODE_SIZE> instead of
1031         GET_MODE_SIZE (<P:MODE>mode).
1032         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
1033         (mov -1, reg peephole2): Likewise.
1034         * config/i386/sse.md (*mov<mode>_internal,
1035         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
1036         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
1037         *<code><mode>3, *andnot<mode>3<mask_name>,
1038         <mask_codefor><code><mode>3<mask_name>): Likewise.
1039         * config/i386/subst.md (mask_mode512bit_condition,
1040         sd_mask_mode512bit_condition): Likewise.
1041
1042 2014-01-02  Xinliang David Li  <davidxl@google.com>
1043
1044         PR tree-optimization/59303
1045         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
1046         (dump_predicates): Better output format.
1047         (pred_equal_p): New function.
1048         (is_neq_relop_p): Ditto.
1049         (is_neq_zero_form_p): Ditto.
1050         (pred_expr_equal_p): Ditto.
1051         (pred_neg_p): Ditto.
1052         (simplify_pred): Ditto.
1053         (simplify_preds_2): Ditto.
1054         (simplify_preds_3): Ditto.
1055         (simplify_preds_4): Ditto.
1056         (simplify_preds): Ditto.
1057         (push_pred): Ditto.
1058         (push_to_worklist): Ditto.
1059         (get_pred_info_from_cmp): Ditto.
1060         (is_degenerated_phi): Ditto.
1061         (normalize_one_pred_1): Ditto.
1062         (normalize_one_pred): Ditto.
1063         (normalize_one_pred_chain): Ditto.
1064         (normalize_preds): Ditto.
1065         (normalize_cond_1): Remove function.
1066         (normalize_cond): Ditto.
1067         (is_gcond_subset_of): Ditto.
1068         (is_subset_of_any): Ditto.
1069         (is_or_set_subset_of): Ditto.
1070         (is_and_set_subset_of): Ditto.
1071         (is_norm_cond_subset_of): Ditto.
1072         (pred_chain_length_cmp): Ditto.
1073         (convert_control_dep_chain_into_preds): Type change.
1074         (find_predicates): Ditto.
1075         (find_def_preds): Ditto.
1076         (destroy_predicates_vecs): Ditto.
1077         (find_matching_predicates_in_rest_chains): Ditto.
1078         (use_pred_not_overlap_with_undef_path_pred): Ditto.
1079         (is_pred_expr_subset): Ditto.
1080         (is_pred_chain_subset_of): Ditto.
1081         (is_included_in): Ditto.
1082         (is_superset_of): Ditto.
1083
1084 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
1085
1086         Update copyright years.
1087
1088 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
1089
1090         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
1091         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
1092         config/arc/arc.md, config/arc/arc.opt,
1093         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
1094         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
1095         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
1096         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
1097         config/linux-protos.h, config/linux.c, config/winnt-c.c,
1098         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
1099         vtable-verify.c, vtable-verify.h: Use the standard form for the
1100         copyright notice.
1101
1102 2014-01-02  Tobias Burnus  <burnus@net-b.de>
1103
1104         * gcc.c (process_command): Update copyright notice dates.
1105         * gcov-dump.c: Ditto.
1106         * gcov.c: Ditto.
1107         * doc/cpp.texi: Bump @copying's copyright year.
1108         * doc/cppinternals.texi: Ditto.
1109         * doc/gcc.texi: Ditto.
1110         * doc/gccint.texi: Ditto.
1111         * doc/gcov.texi: Ditto.
1112         * doc/install.texi: Ditto.
1113         * doc/invoke.texi: Ditto.
1114
1115 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
1116
1117         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
1118
1119 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
1120
1121         * config/i386/sse.md (*mov<mode>_internal): Guard
1122         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
1123
1124         PR rtl-optimization/59647
1125         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
1126         new_rtx into UNSIGNED_FLOAT rtxes.
1127 \f
1128 Copyright (C) 2014 Free Software Foundation, Inc.
1129
1130 Copying and distribution of this file, with or without modification,
1131 are permitted in any medium without royalty provided the copyright
1132 notice and this notice are preserved.