gcc/ChangeLog

   1 2014-01-27  Jeff Law  <law@redhat.com>
   2
   3         * ipa-inline.c (inline_small_functions): Fix typo.
   4
   5 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
   6
   7         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
   8         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
   9         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
  10         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
  11         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
  12         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
  13         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
  14         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
  15         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
  16         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
  17         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
  18         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
  19         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
  20         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
  21         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
  22         (_mm512_storeu_epi64): Ditto.
  23         (_mm512_cmpge_epi32_mask): Ditto.
  24         (_mm512_cmpge_epu32_mask): Ditto.
  25         (_mm512_cmpge_epi64_mask): Ditto.
  26         (_mm512_cmpge_epu64_mask): Ditto.
  27         (_mm512_cmple_epi32_mask): Ditto.
  28         (_mm512_cmple_epu32_mask): Ditto.
  29         (_mm512_cmple_epi64_mask): Ditto.
  30         (_mm512_cmple_epu64_mask): Ditto.
  31         (_mm512_cmplt_epi32_mask): Ditto.
  32         (_mm512_cmplt_epu32_mask): Ditto.
  33         (_mm512_cmplt_epi64_mask): Ditto.
  34         (_mm512_cmplt_epu64_mask): Ditto.
  35         (_mm512_cmpneq_epi32_mask): Ditto.
  36         (_mm512_cmpneq_epu32_mask): Ditto.
  37         (_mm512_cmpneq_epi64_mask): Ditto.
  38         (_mm512_cmpneq_epu64_mask): Ditto.
  39         (_mm512_expand_pd): Ditto.
  40         (_mm512_expand_ps): Ditto.
  41         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
  42         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
  43         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
  44         * config/i386/i386.c (ix86_builtins): Add
  45         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
  46         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
  47         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
  48         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
  49         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
  50         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
  51         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
  52         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
  53         IX86_BUILTIN_PMOVUSQW512_MEM.
  54         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
  55         __builtin_ia32_pmovsqd512mem_mask,
  56         __builtin_ia32_pmovqd512mem_mask,
  57         __builtin_ia32_pmovusqw512mem_mask,
  58         __builtin_ia32_pmovsqw512mem_mask,
  59         __builtin_ia32_pmovqw512mem_mask,
  60         __builtin_ia32_pmovusdw512mem_mask,
  61         __builtin_ia32_pmovsdw512mem_mask,
  62         __builtin_ia32_pmovdw512mem_mask,
  63         __builtin_ia32_pmovqb512mem_mask,
  64         __builtin_ia32_pmovusqb512mem_mask,
  65         __builtin_ia32_pmovsqb512mem_mask,
  66         __builtin_ia32_pmovusdb512mem_mask,
  67         __builtin_ia32_pmovsdb512mem_mask,
  68         __builtin_ia32_pmovdb512mem_mask.
  69         (bdesc_args): Add __builtin_ia32_expanddf512,
  70         __builtin_ia32_expandsf512.
  71         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
  72         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
  73         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
  74         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
  75         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
  76         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
  77         (avx512f_<code>v8div16qi2_mask_store): This.
  78         (avx512f_expand<mode>): New.
  79
  80 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
  81
  82         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd): New.
  83         (_mm512_mask_prefetch_i64gather_pd): Ditto.
  84         (_mm512_prefetch_i32scatter_pd): Ditto.
  85         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
  86         (_mm512_prefetch_i64scatter_pd): Ditto.
  87         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
  88         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
  89         (_mm512_mask_prefetch_i64gather_ps): Ditto.
  90         (_mm512_prefetch_i32scatter_ps): Ditto.
  91         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
  92         (_mm512_prefetch_i64scatter_ps): Ditto.
  93         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
  94         * config/i386/i386-builtin-types.def: Define
  95         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
  96         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
  97         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
  98         IX86_BUILTIN_SCATTERPFQPD.
  99         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
 100         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
 101         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
 102         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
 103         __builtin_ia32_scatterpfqps.
 104         (ix86_expand_builtin): Expand new built-ins.
 105         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
 106         fix memory access data type.
 107         (*avx512pf_gatherpf<mode>_mask): Ditto.
 108         (*avx512pf_gatherpf<mode>): Ditto.
 109         (avx512pf_scatterpf<mode>): Ditto.
 110         (*avx512pf_scatterpf<mode>_mask): Ditto.
 111         (*avx512pf_scatterpf<mode>): Ditto.
 112         (GATHER_SCATTER_SF_MEM_MODE): New.
 113         (avx512pf_gatherpf<mode>df): Ditto.
 114         (*avx512pf_gatherpf<mode>df_mask): Ditto.
 115         (*avx512pf_scatterpf<mode>df): Ditto.
 116
 117 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
 118
 119         PR bootstrap/59934
 120         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
 121         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
 122         reached.
 123
 124 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
 125
 126         * common/config/arm/arm-common.c
 127         (arm_rewrite_mcpu): Handle multiple names.
 128         * config/arm/arm.h
 129         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
 130
 131 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
 132
 133         * gimple-builder.h (create_gimple_tmp): Delete.
 134
 135 2014-01-27  Christian Bruel  <christian.bruel@st.com>
 136
 137         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
 138         words comparisons.
 139
 140 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
 141
 142         * config/pa/pa.md (call): Generate indirect long calls to non-local
 143         functions when outputing 32-bit code.
 144         (call_value): Likewise except for special call to buggy powf function.
 145
 146         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
 147         portable runtime and PIC indirect calls.
 148         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
 149         and PIC call sequences.  Use ldo instead of blr to set return register
 150         in PIC call sequence.
 151
 152 2014-01-25  Walter Lee  <walt@tilera.com>
 153
 154         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
 155         avoid clobbering a live register.
 156
 157 2014-01-25  Walter Lee  <walt@tilera.com>
 158
 159         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
 160         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
 161         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
 162         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
 163
 164 2014-01-25  Walter Lee  <walt@tilera.com>
 165
 166         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
 167         arguments on even registers.
 168         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
 169         STACK_BOUNDARY.
 170         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
 171         (BIGGEST_ALIGNMENT): Ditto.
 172         (BIGGEST_FIELD_ALIGNMENT): Ditto.
 173
 174 2014-01-25  Walter Lee  <walt@tilera.com>
 175
 176         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
 177         insns before bundling.
 178         * config/tilegx/tilegx.md (tile_network_barrier): Update
 179         comment.
 180
 181 2014-01-25  Walter Lee  <walt@tilera.com>
 182
 183         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
 184         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
 185         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
 186
 187 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
 188
 189         * config/mips/constraints.md (kl): Delete.
 190         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
 191         define expands, using...
 192         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
 193         instructions for MIPS16.
 194         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
 195         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
 196
 197 2014-01-25  Walter Lee  <walt@tilera.com>
 198
 199         * config/tilepro/tilepro.md (ctzdi2): Use register_operand
 200         predicate.
 201         (clzdi2): Ditto.
 202         (ffsdi2): Ditto.
 203
 204 2014-01-25  Walter Lee  <walt@tilera.com>
 205
 206         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
 207         (TARGET_EXPAND_TO_RTL_HOOK): Define.
 208
 209 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
 210
 211         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
 212         Handle XOR.
 213
 214 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
 215
 216         * print-rtl.c (in_call_function_usage): New var.
 217         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
 218         EXPR_LIST mode as mode and not as reg note name.
 219
 220         PR middle-end/59561
 221         * cfgloopmanip.c (copy_loop_info): If
 222         loop->warned_aggressive_loop_optimizations, make sure
 223         the flag is set in target loop too.
 224
 225 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
 226
 227         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
 228         flag_cilkplus.
 229         * builtins.def: Likewise.
 230         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
 231         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
 232         * ira.c (ira_setup_eliminable_regset): Likewise.
 233         * omp-low.c (gate_expand_omp): Likewise.
 234         (execute_lower_omp): Likewise.
 235         (diagnose_sb_0): Likewise.
 236         (gate_diagnose_omp_blocks): Likewise.
 237         (simd_clone_clauses_extract): Likewise.
 238         (gate): Likewise.
 239
 240 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
 241
 242         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
 243         correction for little endian...
 244         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
 245         here.
 246
 247 2014-01-24  Jeff Law  <law@redhat.com>
 248
 249         PR tree-optimization/59919
 250         * tree-vrp.c (find_assert_locations_1): Do not register asserts
 251         for non-returning calls.
 252
 253 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
 254
 255         * common/config/aarch64/aarch64-common.c
 256         (aarch64_rewrite_mcpu): Handle multiple names.
 257         * config/aarch64/aarch64.h
 258         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
 259
 260 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
 261
 262         * input.c (add_file_to_cache_tab): Handle the case where fopen
 263         returns NULL.
 264
 265 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
 266
 267         PR target/59929
 268         * config/i386/i386.md (pushsf splitter): Get stack adjustment
 269         from push operand if code of push isn't PRE_DEC.
 270
 271 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
 272
 273         PR target/59909
 274         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
 275         -mquad-memory-atomic.  Update -mquad-memory documentation to say
 276         it is only used for non-atomic loads/stores.
 277
 278         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
 279         -mquad-memory or -mquad-memory-atomic switches.
 280
 281         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
 282         -mquad-memory-atomic to ISA 2.07 support.
 283
 284         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
 285         to separate support of normal quad word memory operations (ldq,
 286         stq) from the atomic quad word memory operations.
 287
 288         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
 289         support to separate non-atomic quad word operations from atomic
 290         quad word operations.  Disable non-atomic quad word operations in
 291         little endian mode so that we don't have to swap words after the
 292         load and before the store.
 293         (quad_load_store_p): Add comment about atomic quad word support.
 294         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
 295         options printed with -mdebug=reg.
 296
 297         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
 298         -mquad-memory-atomic as the test for whether we have quad word
 299         atomic instructions.
 300         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic,
 301         -mquad-memory, or -mp8-vector are used, allow byte/half-word
 302         atomic operations.
 303
 304         * config/rs6000/sync.md (load_lockedti): Insure that the address
 305         is a proper indexed or indirect address for the lqarx instruction.
 306         On little endian systems, swap the hi/lo registers after the lqarx
 307         instruction.
 308         (load_lockedpti): Use indexed_or_indirect_operand predicate to
 309         insure the address is valid for the lqarx instruction.
 310         (store_conditionalti): Insure that the address is a proper indexed
 311         or indirect address for the stqcrx. instruction.  On little endian
 312         systems, swap the hi/lo registers before doing the stqcrx.
 313         instruction.
 314         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
 315         insure the address is valid for the stqcrx. instruction.
 316
 317         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
 318         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
 319         type of quad memory support is available.
 320
 321 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
 322
 323         PR regression/59915
 324         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
 325         there is a danger of looping.
 326
 327 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
 328
 329         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
 330         force flag_ira_loop_pressure if set via command line.
 331
 332 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
 333
 334         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
 335         (ashr_simd): New builtin handling DI mode.
 336         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
 337         (aarch64_sshr_simddi): New match pattern.
 338         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
 339         (vshrd_n_s64): Likewise.
 340         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
 341
 342 2014-01-23  Nick Clifton  <nickc@redhat.com>
 343
 344         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
 345         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
 346         favour of mcu specific scripts.
 347         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
 348         430x multilibs.
 349
 350 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
 351             Alex Velenko  <Alex.Velenko@arm.com>
 352
 353         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
 354         (vaddv_s16): Likewise.
 355         (vaddv_s32): Likewise.
 356         (vaddv_u8): Likewise.
 357         (vaddv_u16): Likewise.
 358         (vaddv_u32): Likewise.
 359         (vaddvq_s8): Likewise.
 360         (vaddvq_s16): Likewise.
 361         (vaddvq_s32): Likewise.
 362         (vaddvq_s64): Likewise.
 363         (vaddvq_u8): Likewise.
 364         (vaddvq_u16): Likewise.
 365         (vaddvq_u32): Likewise.
 366         (vaddvq_u64): Likewise.
 367         (vaddv_f32): Likewise.
 368         (vaddvq_f32): Likewise.
 369         (vaddvq_f64): Likewise.
 370         (vmaxv_f32): Likewise.
 371         (vmaxv_s8): Likewise.
 372         (vmaxv_s16): Likewise.
 373         (vmaxv_s32): Likewise.
 374         (vmaxv_u8): Likewise.
 375         (vmaxv_u16): Likewise.
 376         (vmaxv_u32): Likewise.
 377         (vmaxvq_f32): Likewise.
 378         (vmaxvq_f64): Likewise.
 379         (vmaxvq_s8): Likewise.
 380         (vmaxvq_s16): Likewise.
 381         (vmaxvq_s32): Likewise.
 382         (vmaxvq_u8): Likewise.
 383         (vmaxvq_u16): Likewise.
 384         (vmaxvq_u32): Likewise.
 385         (vmaxnmv_f32): Likewise.
 386         (vmaxnmvq_f32): Likewise.
 387         (vmaxnmvq_f64): Likewise.
 388         (vminv_f32): Likewise.
 389         (vminv_s8): Likewise.
 390         (vminv_s16): Likewise.
 391         (vminv_s32): Likewise.
 392         (vminv_u8): Likewise.
 393         (vminv_u16): Likewise.
 394         (vminv_u32): Likewise.
 395         (vminvq_f32): Likewise.
 396         (vminvq_f64): Likewise.
 397         (vminvq_s8): Likewise.
 398         (vminvq_s16): Likewise.
 399         (vminvq_s32): Likewise.
 400         (vminvq_u8): Likewise.
 401         (vminvq_u16): Likewise.
 402         (vminvq_u32): Likewise.
 403         (vminnmv_f32): Likewise.
 404         (vminnmvq_f32): Likewise.
 405         (vminnmvq_f64): Likewise.
 406
 407 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
 408
 409         * config/aarch64/aarch64-simd.md
 410         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
 411         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
 412         (*aarch64_mul3_elt<mode>): Likewise.
 413         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
 414         (*aarch64_mul3_elt_to_64v2df): Likewise.
 415         (*aarch64_mla_elt<mode>): Likewise.
 416         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
 417         (*aarch64_mls_elt<mode>): Likewise.
 418         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
 419         (*aarch64_fma4_elt<mode>): Likewise.
 420         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
 421         (*aarch64_fma4_elt_to_64v2df): Likewise.
 422         (*aarch64_fnma4_elt<mode>): Likewise.
 423         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
 424         (*aarch64_fnma4_elt_to_64v2df): Likewise.
 425         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
 426         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
 427         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
 428         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
 429         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
 430         (aarch64_sqdmull_lane<mode>_internal): Likewise.
 431         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
 432
 433 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
 434
 435         * config/aarch64/aarch64-simd.md
 436         (aarch64_be_checked_get_lane<mode>): New define_expand.
 437         * config/aarch64/aarch64-simd-builtins.def
 438         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0):
 439         New builtin definition.
 440         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
 441         Use new safe be builtin.
 442
 443 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
 444
 445         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
 446         New define_insn.
 447         (aarch64_be_st1<mode>): Likewise.
 448         (aarch_ld1<VALL:mode>): Define_expand modified.
 449         (aarch_st1<VALL:mode>): Likewise.
 450         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
 451         (UNSPEC_ST1): Likewise.
 452
 453 2014-01-23  David Holsgrove <david.holsgrove@xilinx.com>
 454
 455         * config/microblaze/microblaze.md: Add trap insn and attribute
 456
 457 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
 458
 459         PR preprocessor/58580
 460         * input.h (location_get_source_line): Take an additional line_size
 461         parameter.
 462         (void diagnostics_file_cache_fini): Declare new function.
 463         * input.c (struct fcache): New type.
 464         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
 465         New static constants.
 466         (diagnostic_file_cache_init, total_lines_num)
 467         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
 468         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
 469         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
 470         (get_next_line, read_next_line, goto_next_line, read_line_num):
 471         New static function definitions.
 472         (diagnostic_file_cache_fini): New function.
 473         (location_get_source_line): Take an additional output line_len
 474         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
 475         read_line_num.
 476         * diagnostic.c (diagnostic_finish): Call
 477         diagnostic_file_cache_fini.
 478         (adjust_line): Take an additional input parameter for the length
 479         of the line, rather than calculating it with strlen.
 480         (diagnostic_show_locus): Adjust the use of
 481         location_get_source_line and adjust_line with respect to their new
 482         signature.  While displaying a line now, do not stop at the first
 483         null byte.  Rather, display the zero byte as a space and keep
 484         going until we reach the size of the line.
 485         * Makefile.in: Add vec.o to OBJS-libcommon
 486
 487 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
 488             Ilya Tocar     <ilya.tocar@intel.com>
 489
 490         * config/i386/avx512fintrin.h (_mm512_kmov): New.
 491         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
 492         (__builtin_ia32_kmov16): Ditto.
 493         * config/i386/i386.md (UNSPEC_KMOV): New.
 494         (kmovw): Ditto.
 495
 496 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
 497
 498         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
 499         (_mm512_storeu_si512): Ditto.
 500
 501 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
 502
 503         PR target/52125
 504         * rtl.h (get_referenced_operands): Declare.
 505         * recog.c (get_referenced_operands): New function.
 506         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
 507         operands have been referenced when recording LO_SUM references.
 508
 509 2014-01-22  David Holsgrove <david.holsgrove@xilinx.com>
 510
 511         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
 512
 513 2014-01-22  Jan Hubicka  <jh@suse.cz>
 514
 515         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
 516         Enable for generic and recent AMD targets.
 517
 518 2014-01-22  Jan Hubicka  <jh@suse.cz>
 519
 520         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
 521         ARG_SIZE note when adjustment was eliminated.
 522
 523 2014-01-22  Jeff Law  <law@redhat.com>
 524
 525         PR tree-optimization/59597
 526         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
 527         in file.  Accept new argument REGISTERING and use it to modify
 528         dump output appropriately.
 529         (register_jump_thread): Corresponding changes.
 530         (mark_threaded_blocks): Reinstate code to cancel unprofitable
 531         thread paths involving joiner blocks.  Add code to dump cancelled
 532         jump threading paths.
 533
 534 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
 535
 536         PR rtl-optimization/59477
 537         * lra-constraints.c (inherit_in_ebb): Process call for living hard
 538         regs.  Update reloads_num and potential_reload_hard_regs for all
 539         insns.
 540
 541 2014-01-22  Tom Tromey  <tromey@redhat.com>
 542
 543         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
 544         PARAMS.
 545         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
 546
 547 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
 548
 549         PR rtl-optimization/59896
 550         * lra-constraints.c (process_alt_operands): Check unused note for
 551         matched operands of insn with no output reloads.
 552
 553 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
 554
 555         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
 556         (mips_move_from_gpr_cost): Likewise.
 557
 558 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
 559
 560         PR rtl-optimization/59858
 561         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
 562         ira_class_hard_regs_num.
 563         (process_alt_operands): Increase reject for dying matched operand.
 564
 565 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
 566
 567         PR target/59003
 568         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
 569         smaller than size, perform several stores or loads and stores
 570         at dst + count - size to store or copy all of size bytes, rather
 571         than just last modesize bytes.
 572
 573 2014-01-20  DJ Delorie  <dj@redhat.com>
 574
 575         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
 576         that CLOBBERs are REGs before propogating their values.
 577
 578 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
 579
 580         PR middle-end/59789
 581         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
 582         (cgraph_inline_failed_type): New function.
 583         * cgraph.h (DEFCIFCODE): Add type.
 584         (cgraph_inline_failed_type_t): New enum.
 585         (cgraph_inline_failed_type): New prototype.
 586         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
 587         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
 588         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
 589         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
 590         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
 591         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
 592         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
 593         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
 594         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
 595         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
 596         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
 597         OPTIMIZATION_MISMATCH.
 598         * tree-inline.c (expand_call_inline): Emit errors during
 599         early_inlining if cgraph_inline_failed_type returns
 600         CIF_FINAL_ERROR.
 601
 602 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
 603
 604         PR target/59685
 605         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
 606         mode attribute in insn output.
 607
 608 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
 609
 610         * output.h (output_constant): Delete.
 611         * varasm.c (output_constant): Make private.
 612
 613 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
 614
 615         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
 616
 617 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
 618
 619         PR middle-end/59860
 620         * tree.h (fold_builtin_strcat): New prototype.
 621         * builtins.c (fold_builtin_strcat): No longer static.  Add len
 622         argument, if non-NULL, don't call c_strlen.  Optimize
 623         directly into __builtin_memcpy instead of __builtin_strcpy.
 624         (fold_builtin_2): Adjust fold_builtin_strcat caller.
 625         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
 626
 627 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
 628
 629         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
 630         for SImode_address_operand operands, having only a REG argument.
 631
 632 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
 633
 634         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
 635         loader name using mbig-endian.
 636         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
 637
 638 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
 639
 640         * doc/invoke.texi (-march): Clarify documentation for AArch64.
 641         (-mtune): Likewise.
 642         (-mcpu): Likewise.
 643
 644 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
 645
 646         * config/aarch64/aarch64-protos.h
 647         (aarch64_cannot_change_mode_class_ptr): Declare.
 648         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
 649         aarch64_cannot_change_mode_class_ptr): New.
 650         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
 651         backend hook aarch64_cannot_change_mode_class.
 652
 653 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
 654
 655         * common/config/aarch64/aarch64-common.c
 656         (aarch64_handle_option): Don't handle any option order logic here.
 657         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
 658         selected_cpu, warn on architecture version mismatch.
 659         (aarch64_override_options): Fix parsing order for option strings.
 660
 661 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
 662             Iain Sandoe  <iain@codesourcery.com>
 663
 664         PR bootstrap/59496
 665         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
 666         warning.  Amend comment to reflect current functionality.
 667
 668 2014-01-20  Richard Biener  <rguenther@suse.de>
 669
 670         PR middle-end/59860
 671         * builtins.c (fold_builtin_strcat): Remove case better handled
 672         by tree-ssa-strlen.c.
 673
 674 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
 675
 676         * config/aarch64/aarch64.opt
 677         (mcpu, march, mtune): Make case-insensitive.
 678
 679 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
 680
 681         PR target/59880
 682         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
 683         if operands[1] is a REG or ZERO_EXTEND of a REG.
 684
 685 2014-01-19  Jan Hubicka  <jh@suse.cz>
 686
 687         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
 688
 689 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
 690
 691         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
 692         long non-pic millicode calls.
 693
 694 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
 695
 696         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
 697
 698 2014-01-19  Kito Cheng  <kito@0xlab.org>
 699
 700         * builtins.c (expand_movstr): Check movstr expand done or fail.
 701
 702 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
 703             H.J. Lu  <hongjiu.lu@intel.com>
 704
 705         PR target/59379
 706         * config/i386/i386.md (*lea<mode>): Zero-extend return register
 707         to DImode for zero-extended addresses.
 708
 709 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
 710
 711         PR rtl-optimization/57763
 712         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
 713         on the new indirect jump_insn and increment LABEL_NUSES (label).
 714
 715 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
 716
 717         PR bootstrap/59580
 718         PR bootstrap/59583
 719         * config.gcc (x86_archs): New variable.
 720         (x86_64_archs): Likewise.
 721         (x86_cpus): Likewise.
 722         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
 723         --with-arch/--with-cpu= options.
 724         Support --with-arch=/--with-cpu={nehalem,westmere,
 725         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
 726
 727 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
 728
 729         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
 730         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
 731
 732 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
 733
 734         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
 735
 736 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
 737
 738         PR target/58944
 739         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
 740         clear cpp_get_options (parse_in)->warn_unused_macros for
 741         ix86_target_macros_internal with cpp_define.
 742
 743 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
 744
 745         * jump.c (delete_related_insns): Keep (use (insn))s.
 746         * reorg.c (redundant_insn): Check for barriers too.
 747
 748 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
 749
 750         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
 751
 752 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
 753
 754         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
 755         call to $$dyncall when TARGET_LONG_CALLS is true.
 756
 757 2014-01-17  Jeff Law  <law@redhat.com>
 758
 759         * ree.c (combine_set_extension): Temporarily disable test for
 760         changing number of hard registers.
 761
 762 2014-01-17  Jan Hubicka  <jh@suse.cz>
 763
 764         PR middle-end/58125
 765         * ipa-inline-analysis.c (inline_free_summary):
 766         Do not free summary of aliases.
 767
 768 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
 769
 770         PR middle-end/59706
 771         * gimplify.c (gimplify_expr): Use create_tmp_var
 772         instead of create_tmp_var_raw.  If cond doesn't have
 773         integral type, don't add the IFN_ANNOTATE builtin at all.
 774
 775 2014-01-17  Martin Jambor  <mjambor@suse.cz>
 776
 777         PR ipa/59736
 778         * ipa-cp.c (prev_edge_clone): New variable.
 779         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
 780         Also resize prev_edge_clone vector.
 781         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
 782         (ipcp_edge_removal_hook): New function.
 783         (ipcp_driver): Register ipcp_edge_removal_hook.
 784
 785 2014-01-17  Andrew Pinski <apinski@cavium.com>
 786             Steve Ellcey  <sellcey@mips.com>
 787
 788         PR target/59462
 789         * config/mips/mips.c (mips_print_operand): Check operand mode instead
 790         of operator mode.
 791
 792 2014-01-17  Jeff Law  <law@redhat.com>
 793
 794         PR middle-end/57904
 795         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
 796         so that pass_ccp runs first.
 797
 798 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
 799
 800         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
 801         (ix86_adjust_cost): Use !TARGET_XXX.
 802         (do_reorder_for_imul): Likewise.
 803         (swap_top_of_ready_list): Likewise.
 804         (ix86_sched_reorder): Likewise.
 805
 806 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
 807
 808         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
 809         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
 810         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
 811         (intel_memset): New.  Duplicate slm_memset.
 812         (intel_cost): New.  Duplicate slm_cost.
 813         (m_INTEL): New macro.
 814         (processor_target_table): Add "intel".
 815         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
 816         with PROCESSOR_INTEL for "intel".
 817         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
 818         PROCESSOR_SILVERMONT.
 819         (ix86_issue_rate): Likewise.
 820         (ix86_adjust_cost): Likewise.
 821         (ia32_multipass_dfa_lookahead): Likewise.
 822         (swap_top_of_ready_list): Likewise.
 823         (ix86_sched_reorder): Likewise.
 824         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
 825         instead of TARGET_OPT_AGU.
 826         * config/i386/i386.h (TARGET_INTEL): New.
 827         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
 828         (processor_type): Add PROCESSOR_INTEL.
 829         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
 830         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
 831
 832 2014-01-17  Marek Polacek  <polacek@redhat.com>
 833
 834         PR c/58346
 835         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
 836         size is zero.
 837
 838 2014-01-17  Richard Biener  <rguenther@suse.de>
 839
 840         PR tree-optimization/46590
 841         * opts.c (default_options_table): Add entries for
 842         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
 843         all enabled at -O1 but not for -Og.
 844         * common.opt (fbranch-count-reg): Remove Init(1).
 845         (fmove-loop-invariants): Likewise.
 846         (ftree-pta): Likewise.
 847
 848 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
 849
 850         * config/i386/i386.c (ix86_data_alignment): For compatibility with
 851         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
 852         decls to at least the GCC 4.8 used alignments.
 853
 854         PR fortran/59440
 855         * tree-nested.c (convert_nonlocal_reference_stmt,
 856         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
 857         of GIMPLE_BIND stmts, adjust associated decls.
 858
 859 2014-01-17  Richard Biener  <rguenther@suse.de>
 860
 861         PR tree-optimization/46590
 862         * vec.h (vec<>::bseach): New member function implementing
 863         binary search according to C89 bsearch.
 864         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
 865         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
 866         bitmap pointer again.  Make accesses_in_loop a flat array.
 867         (mem_ref_obstack): New global.
 868         (outermost_indep_loop): Adjust for mem_ref->stored changes.
 869         (mark_ref_stored): Likewise.
 870         (ref_indep_loop_p_2): Likewise.
 871         (set_ref_stored_in_loop): New helper function.
 872         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
 873         (memref_free): Adjust.
 874         (record_mem_ref_loc): Simplify.
 875         (gather_mem_refs_stmt): Adjust.
 876         (sort_locs_in_loop_postorder_cmp): New function.
 877         (analyze_memory_references): Sort accesses_in_loop after
 878         loop postorder number.
 879         (find_ref_loc_in_loop_cmp): New function.
 880         (for_all_locs_in_loop): Find relevant cluster of locs in
 881         accesses_in_loop and iterate without recursion.
 882         (execute_sm): Avoid uninit warning.
 883         (struct ref_always_accessed): Simplify.
 884         (ref_always_accessed::operator ()): Likewise.
 885         (ref_always_accessed_p): Likewise.
 886         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
 887         loop postorder numbers here.
 888         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
 889         numbers.
 890
 891 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
 892
 893         PR c++/57945
 894         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
 895         on decls for which assemble_alias has been called.
 896
 897 2014-01-17  Nick Clifton  <nickc@redhat.com>
 898
 899         * config/msp430/msp430.opt: (mcpu): New option.
 900         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
 901         (msp430_option_override): Parse target_cpu.  If the MCU name
 902         matches a generic string, clear target_mcu.
 903         (msp430_attr): Allow numeric interrupt values up to 63.
 904         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
 905         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
 906         option.
 907         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
 908         Add mcpu matches.
 909         * config/msp430/msp430.md (popm): Use %J rather than %I.
 910         (addsi3): Use msp430_nonimmediate_operand for operand 2.
 911         (addhi_cy_i): Use immediate_operand for operand 2.
 912         * doc/invoke.texi: Document -mcpu option.
 913
 914 2014-01-17  Richard Biener  <rguenther@suse.de>
 915
 916         PR rtl-optimization/38518
 917         * df.h (df_analyze_loop): Declare.
 918         * df-core.c: Include cfgloop.h.
 919         (df_analyze_1): Split out main part of df_analyze.
 920         (df_analyze): Adjust.
 921         (loop_inverted_post_order_compute): New function.
 922         (loop_post_order_compute): Likewise.
 923         (df_analyze_loop): New function avoiding whole-function
 924         postorder computes.
 925         * loop-invariant.c (find_defs): Use df_analyze_loop.
 926         (find_invariants): Adjust.
 927         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
 928
 929 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
 930
 931         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
 932         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
 933
 934 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
 935
 936         * ipa-ref.c (ipa_remove_stmt_references): Fix references
 937         traversal when removing references.
 938
 939 2014-01-16  Jan Hubicka  <jh@suse.cz>
 940
 941         PR ipa/59775
 942         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
 943
 944 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
 945
 946         PR middle-end/56791
 947         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
 948         pushing a reload for an autoinc when we had previously reloaded an
 949         inner part of the address.
 950
 951 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
 952
 953         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
 954         field.
 955         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
 956         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
 957         when not giving up or versioning for alias only because of
 958         loop->safelen.
 959         (vect_analyze_data_ref_dependences): Set to true.
 960         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
 961         is a GIMPLE_PHI.
 962         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
 963         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
 964         to the condition.
 965
 966         PR middle-end/58344
 967         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
 968
 969         PR target/59839
 970         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
 971         operand 0 predicate for gathers, use a new pseudo as subtarget.
 972
 973 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
 974
 975         PR middle-end/59609
 976         * lra-constraints.c (process_alt_operands): Add printing debug info.
 977         Check absence of input/output reloads for matched operands too.
 978
 979 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
 980
 981         PR rtl-optimization/59835
 982         * ira.c (ira_init_register_move_cost): Increase cost for
 983         impossible modes.
 984
 985 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
 986
 987         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
 988
 989 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
 990
 991         PR target/59780
 992         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
 993         non-register objects.  Use gen_(high/low)part more consistently.
 994         Fix assertions.
 995
 996 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
 997
 998         PR target/59844
 999         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
1000         endian support, remove tests for WORDS_BIG_ENDIAN.
1001         (p8_mfvsrd_3_<mode>): Likewise.
1002         (reload_gpr_from_vsx<mode>): Likewise.
1003         (reload_gpr_from_vsxsf): Likewise.
1004         (p8_mfvsrd_4_disf): Likewise.
1005
1006 2014-01-16  Richard Biener  <rguenther@suse.de>
1007
1008         PR rtl-optimization/46590
1009         * lcm.c (compute_antinout_edge): Use postorder iteration.
1010         (compute_laterin): Use inverted postorder iteration.
1011
1012 2014-01-16  Nick Clifton  <nickc@redhat.com>
1013
1014         PR middle-end/28865
1015         * varasm.c (output_constant): Return the number of bytes actually
1016         emitted.
1017         (output_constructor_array_range): Update the field size with the
1018         number of bytes emitted by output_constant.
1019         (output_constructor_regular_field): Likewise.  Also do not
1020         complain if the total number of bytes emitted is now greater
1021         than the expected fieldpos.
1022         * output.h (output_constant): Update prototype and descriptive comment.
1023
1024 2014-01-16  Marek Polacek  <polacek@redhat.com>
1025
1026         PR middle-end/59827
1027         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
1028         it is error_mark_node.
1029
1030 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
1031
1032         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
1033         VALID_AVX256_REG_OR_OI_MODE.
1034
1035 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
1036
1037         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
1038         current procedure should be profiled.
1039
1040 2014-01-15  Andrew Pinski  <apinski@cavium.com>
1041
1042         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
1043         of moving from/to the STACK_REG register class.
1044
1045 2014-01-15  Richard Henderson <rth@redhat.com>
1046
1047         PR debug/54694
1048         * reginfo.c (global_regs_decl): Globalize.
1049         * rtl.h (global_regs_decl): Declare.
1050         * ira.c (do_reload): Diagnose frame_pointer_needed and it
1051         reserved via global_regs.
1052
1053 2014-01-15  Teresa Johnson  <tejohnson@google.com>
1054
1055         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
1056
1057 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
1058
1059         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
1060         and vmulosh rather than call gen_vec_widen_smult_*.
1061         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
1062         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
1063         (vec_widen_smult_even_v16qi): Likewise.
1064         (vec_widen_umult_even_v8hi): Likewise.
1065         (vec_widen_smult_even_v8hi): Likewise.
1066         (vec_widen_umult_odd_v16qi): Likewise.
1067         (vec_widen_smult_odd_v16qi): Likewise.
1068         (vec_widen_umult_odd_v8hi): Likewise.
1069         (vec_widen_smult_odd_v8hi): Likewise.
1070         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
1071         vmuloub rather than call gen_vec_widen_umult_*.
1072         (vec_widen_umult_lo_v16qi): Likewise.
1073         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
1074         vmulosb rather than call gen_vec_widen_smult_*.
1075         (vec_widen_smult_lo_v16qi): Likewise.
1076         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
1077         rather than call gen_vec_widen_umult_*.
1078         (vec_widen_umult_lo_v8hi): Likewise.
1079         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
1080         rather than call gen_vec_widen_smult_*.
1081         (vec_widen_smult_lo_v8hi): Likewise.
1082
1083 2014-01-15  Jeff Law  <law@redhat.com>
1084
1085         PR tree-optimization/59747
1086         * ree.c (find_and_remove_re): Properly handle case where a second
1087         eliminated extension requires widening a copy created for elimination
1088         of a prior extension.
1089         (combine_set_extension): Ensure that the number of hard regs needed
1090         for a destination register does not change when we widen it.
1091
1092 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
1093
1094         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
1095         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
1096         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
1097         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
1098         (avr-*-rtems*): Likewise.
1099         (bfin*-rtems*): Likewise.
1100         (moxie-*-rtems*): Likewise.
1101         (h8300-*-rtems*): Likewise.
1102         (i[34567]86-*-rtems*): Likewise.
1103         (lm32-*-rtems*): Likewise.
1104         (m32r-*-rtems*): Likewise.
1105         (m68k-*-rtems*): Likewise.
1106         (microblaze*-*-rtems*): Likewise.
1107         (mips*-*-rtems*): Likewise.
1108         (powerpc-*-rtems*): Likewise.
1109         (sh-*-rtems*): Likewise.
1110         (sparc-*-rtems*): Likewise.
1111         (sparc64-*-rtems*): Likewise.
1112         (v850-*-rtems*): Likewise.
1113         (m32c-*-rtems*): Likewise.
1114
1115 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
1116
1117         PR rtl-optimization/59511
1118         * ira.c (ira_init_register_move_cost): Use memory costs for some
1119         cases of register move cost calculations.
1120         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
1121         instead of BB frequency.
1122         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
1123         * lra-assigns.c (find_hard_regno_for): Ditto.
1124
1125 2014-01-15  Richard Biener  <rguenther@suse.de>
1126
1127         PR tree-optimization/59822
1128         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
1129         (vectorizable_load): Use it to hoist defs of uses of invariant
1130         loads out of the loop.
1131
1132 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
1133             Kugan Vivekanandarajah  <kuganv@linaro.org>
1134
1135         PR target/59695
1136         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
1137         truncation.
1138
1139 2014-01-15  Richard Biener  <rguenther@suse.de>
1140
1141         PR rtl-optimization/59802
1142         * lcm.c (compute_available): Use inverted postorder to seed
1143         the initial worklist.
1144
1145 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
1146
1147         PR target/59803
1148         * config/s390/s390.c (s390_preferred_reload_class): Don't return
1149         ADDR_REGS for invalid symrefs in non-PIC code.
1150
1151 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
1152
1153         PR other/58712
1154         * builtins.c (determine_block_size): Initialize *probable_max_size
1155         even if len_rtx is CONST_INT.
1156
1157 2014-01-14  Andrew Pinski  <apinski@cavium.com>
1158
1159         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
1160         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
1161         (cortexa53_tunings): Likewise.
1162         (aarch64_sched_issue_rate): New function.
1163         (TARGET_SCHED_ISSUE_RATE): Define.
1164
1165 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
1166
1167         * ira-costs.c (find_costs_and_classes): Add missed
1168         ira_init_register_move_cost_if_necessary.
1169
1170 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
1171
1172         PR target/59787
1173         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
1174
1175 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
1176
1177         PR target/59794
1178         * config/i386/i386.c (type_natural_mode): Add a bool parameter
1179         to indicate if type is used for function return value.  Warn ABI
1180         change if the vector mode isn't available for function return value.
1181         (ix86_function_arg_advance): Pass false to type_natural_mode.
1182         (ix86_function_arg): Likewise.
1183         (ix86_gimplify_va_arg): Likewise.
1184         (function_arg_32): Don't warn ABI change.
1185         (ix86_function_value): Pass true to type_natural_mode.
1186         (ix86_return_in_memory): Likewise.
1187         (ix86_struct_value_rtx): Removed.
1188         (TARGET_STRUCT_VALUE_RTX): Likewise.
1189
1190 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
1191
1192         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
1193         converting a conditional jump into a conditional return.
1194
1195 2014-01-14  Richard Biener  <rguenther@suse.de>
1196
1197         PR tree-optimization/58921
1198         PR tree-optimization/59006
1199         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
1200         hoisting invariant stmts.
1201         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
1202         invariant loads on the preheader edge if possible.
1203
1204 2014-01-14  Joey Ye  <joey.ye@arm.com>
1205
1206         * doc/plugin.texi (Building GCC plugins): Update to C++.
1207
1208 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
1209
1210         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
1211         (_mm_rcp28_round_ss): Ditto.
1212         (_mm_rsqrt28_round_sd): Ditto.
1213         (_mm_rsqrt28_round_ss): Ditto.
1214         (_mm_rcp28_sd): Ditto.
1215         (_mm_rcp28_ss): Ditto.
1216         (_mm_rsqrt28_sd): Ditto.
1217         (_mm_rsqrt28_ss): Ditto.
1218         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
1219         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
1220         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
1221         (IX86_BUILTIN_RCP28SD): Ditto.
1222         (IX86_BUILTIN_RCP28SS): Ditto.
1223         (IX86_BUILTIN_RSQRT28SD): Ditto.
1224         (IX86_BUILTIN_RSQRT28SS): Ditto.
1225         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
1226         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
1227         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
1228         (ix86_expand_special_args_builtin): Expand new FTYPE.
1229         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
1230         (srcp14<mode>): Make insn unary.
1231         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
1232         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
1233         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
1234         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
1235         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
1236         Fix rounding: make it SAE only.
1237         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
1238         Ditto.
1239         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
1240         Ditto.
1241         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
1242         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
1243         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
1244         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
1245         (round_saeonly_mask_scalar_operand4): Ditto.
1246         (round_saeonly_mask_scalar_op3): Ditto.
1247         (round_saeonly_mask_scalar_op4): Ditto.
1248
1249 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1250
1251         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1252         Implement -maltivec=be for vec_insert and vec_extract.
1253
1254 2014-01-10  DJ Delorie  <dj@redhat.com>
1255
1256         * config/msp430/msp430.md (call_internal): Don't allow memory
1257         references with SP as the base register.
1258         (call_value_internal): Likewise.
1259         * config/msp430/constraints.md (Yc): New.  For memory references
1260         that don't use SP as a base register.
1261
1262         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
1263         "an integer without a # prefix"
1264         * config/msp430/msp430.md (epilogue_helper): Use it.
1265
1266 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
1267
1268         PR target/59617
1269         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
1270         AVX512F gather builtins.
1271         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
1272         on gather decls with INTEGER_TYPE masktype.
1273         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
1274         directly into the builtin rather than hoisting it before loop.
1275
1276         PR tree-optimization/59387
1277         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
1278         (scev_const_prop): If folded_casts and type has undefined overflow,
1279         use force_gimple_operand instead of force_gimple_operand_gsi and
1280         for each added stmt if it is assign with
1281         arith_code_with_undefined_signed_overflow, call
1282         rewrite_to_defined_overflow.
1283         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
1284         gimple-fold.h instead.
1285         (arith_code_with_undefined_signed_overflow,
1286         rewrite_to_defined_overflow): Moved to ...
1287         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
1288         rewrite_to_defined_overflow): ... here.  No longer static.
1289         Include gimplify-me.h.
1290         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
1291         rewrite_to_defined_overflow): New prototypes.
1292
1293 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1294
1295         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
1296
1297 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
1298
1299         * builtins.c (get_object_alignment_2): Minor tweak.
1300         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
1301
1302 2014-01-13  Christian Bruel  <christian.bruel@st.com>
1303
1304         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
1305         optimized non constant lengths.
1306
1307 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
1308
1309         PR libgomp/59194
1310         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
1311         load as __atomic_load_N if possible.
1312
1313 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
1314
1315         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
1316         target parameter.
1317         (rs6000_expand_builtin): Adjust call.
1318
1319 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
1320
1321         PR target/58115
1322         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
1323         * config/rs6000/rs6000.c: Include target-globals.h.
1324         (rs6000_set_current_function): Instead of doing target_reinit
1325         unconditionally, use save_target_globals_default_opts and
1326         restore_target_globals.
1327
1328         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
1329         FPSCR.
1330         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
1331         (rs6000_expand_builtin): Handle mffs and mtfsf.
1332         (rs6000_init_builtins): Define mffs and mtfsf.
1333         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
1334         (rs6000_mffs): New pattern.
1335         (rs6000_mtfsf): New pattern.
1336
1337 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
1338
1339         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
1340         Start narrowing with START.  Apply candidate-use pair
1341         and check overall cost in narrowing.
1342         (iv_ca_prune): Pass new argument.
1343
1344 2014-01-10  Jeff Law  <law@redhat.com>
1345
1346         PR middle-end/59743
1347         * ree.c (combine_reaching_defs): Ensure the defining statement
1348         occurs before the extension when optimizing extensions with
1349         different source and destination hard registers.
1350
1351 2014-01-10  Jan Hubicka  <jh@suse.cz>
1352
1353         PR ipa/58585
1354         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
1355         vtables into the type inheritance graph.
1356
1357 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
1358
1359         PR rtl-optimization/59754
1360         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
1361         modes in the REGNO != REGNO case.
1362
1363 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1364
1365         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
1366
1367 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
1368
1369         PR tree-optimization/59745
1370         * tree-predcom.c (tree_predictive_commoning_loop): Call
1371         free_affine_expand_cache if giving up because components is NULL.
1372
1373         * target-globals.c (save_target_globals): Allocate < 4KB structs using
1374         GC in payload of target_globals struct instead of allocating them on
1375         the heap and the larger structs separately using GC.
1376         * target-globals.h (struct target_globals): Make regs, hard_regs,
1377         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
1378         of GTY((skip)) and change type to void *.
1379         (reset_target_globals): Cast loads from those fields to corresponding
1380         types.
1381
1382 2014-01-10  Steve Ellcey  <sellcey@mips.com>
1383
1384         PR plugins/59335
1385         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
1386         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
1387         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
1388
1389 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
1390
1391         PR target/59744
1392         * aarch64-modes.def (CC_Zmode): New flags mode.
1393         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
1394         represents an equality.
1395         (aarch64_get_condition_code): Handle CC_Zmode.
1396         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
1397
1398 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
1399
1400         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
1401         extraction in good case.
1402
1403 2014-01-10  Richard Biener  <rguenther@suse.de>
1404
1405         PR tree-optimization/59374
1406         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
1407         checking after SLP discovery.  Mark stmts not participating
1408         in any SLP instance properly.
1409
1410 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1411
1412         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
1413         when handling a SET rtx.
1414
1415 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1416
1417         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
1418         (cortex-a57): Likewise.
1419         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
1420
1421 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1422
1423         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
1424         non-iwmmxt builtins.
1425
1426 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
1427
1428         PR ipa/58252
1429         PR ipa/59226
1430         * ipa-devirt.c record_target_from_binfo): Take as argument
1431         stack of binfos and lookup matching one for virtual inheritance.
1432         (possible_polymorphic_call_targets_1): Update.
1433
1434 2014-01-10  Huacai Chen  <chenhc@lemote.com>
1435
1436         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
1437         kernel strings for Loongson-2E/2F/3A.
1438
1439 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
1440
1441         PR middle-end/59670
1442         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
1443         is_gimple_call before calling gimple_call_internal_p.
1444
1445 2014-01-09  Steve Ellcey  <sellcey@mips.com>
1446
1447         * Makefile.in (TREE_FLOW_H): Remove.
1448         (TREE_SSA_H): Add file names from tree-flow.h.
1449         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
1450         * tree.h: Remove tree-flow.h reference.
1451         * hash-table.h: Remove tree-flow.h reference.
1452         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
1453         reference with tree-ssa-loop.h.
1454
1455 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1456
1457         * doc/invoke.texi: Add -maltivec={be,le} options, and document
1458         default element-order behavior for -maltivec.
1459         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
1460         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
1461         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
1462         when targeting big endian, at least for now.
1463         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
1464
1465 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
1466
1467         PR middle-end/47735
1468         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
1469         var satisfies use_register_for_decl, just take into account type
1470         alignment, rather than decl alignment.
1471
1472         PR tree-optimization/59622
1473         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
1474         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
1475         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
1476         Don't devirtualize for inplace at all.  For targets.length () == 1,
1477         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
1478
1479 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
1480
1481         * config/i386/i386.md (cpu): Remove the unused btver1.
1482
1483 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
1484
1485         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
1486
1487 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
1488
1489         PR target/58115
1490         * tree-core.h (struct target_globals): New forward declaration.
1491         (struct tree_target_option): Add globals field.
1492         * tree.h (TREE_TARGET_GLOBALS): Define.
1493         (prepare_target_option_nodes_for_pch): New prototype.
1494         * target-globals.h (struct target_globals): Define even if
1495         !SWITCHABLE_TARGET.
1496         * tree.c (prepare_target_option_node_for_pch,
1497         prepare_target_option_nodes_for_pch): New functions.
1498         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
1499         * config/i386/i386.c: Include target-globals.h.
1500         (ix86_set_current_function): Instead of doing target_reinit
1501         unconditionally, use save_target_globals_default_opts and
1502         restore_target_globals.
1503
1504 2014-01-09  Richard Biener  <rguenther@suse.de>
1505
1506         PR tree-optimization/59715
1507         * tree-cfg.h (split_critical_edges): Declare.
1508         * tree-cfg.c (split_critical_edges): Export.
1509         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
1510
1511 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
1512
1513         * cfgexpand.c (expand_stack_vars): Optionally disable
1514         asan stack protection.
1515         (expand_used_vars): Likewise.
1516         (partition_stack_vars): Likewise.
1517         * asan.c (asan_emit_stack_protection): Optionally disable
1518         after return stack usage.
1519         (instrument_derefs): Optionally disable memory access instrumentation.
1520         (instrument_builtin_call): Likewise.
1521         (instrument_strlen_call): Likewise.
1522         (asan_protect_global): Optionally disable global variables protection.
1523         * doc/invoke.texi: Added doc for new options.
1524         * params.def: Added new options.
1525         * params.h: Likewise.
1526
1527 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
1528
1529         PR rtl-optimization/59724
1530         * ifcvt.c (cond_exec_process_if_block): Don't call
1531         flow_find_head_matching_sequence with 0 longest_match.
1532         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
1533         non-active insns if !stop_after.
1534         (try_head_merge_bb): Revert 2014-01-07 changes.
1535
1536 2014-01-08  Jeff Law  <law@redhat.com>
1537
1538         * ree.c (get_sub_rtx): New function, extracted from...
1539         (merge_def_and_ext): Here.
1540         (combine_reaching_defs): Use get_sub_rtx.
1541
1542 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
1543
1544         * cgraph.h (varpool_variable_node): Do not choke on null node.
1545
1546 2014-01-08  Catherine Moore  <clm@codesourcery.com>
1547
1548         * config/mips/mips.md (simple_return): Attempt to use JRC
1549         for microMIPS.
1550         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
1551
1552 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
1553
1554         PR rtl-optimization/59137
1555         * reorg.c (steal_delay_list_from_target): Call update_block for
1556         elided insns.
1557         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
1558
1559 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1560
1561         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1562         two duplicate entries.
1563
1564 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
1565
1566         Revert:
1567         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
1568
1569         * config/mips/mips.c (mips_truncated_op_cost): New function.
1570         (mips_rtx_costs): Adjust test for BADDU.
1571         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
1572
1573         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
1574
1575         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
1576         (*baddu_si): ...this new pattern.
1577
1578 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
1579
1580         PR ipa/59722
1581         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
1582
1583 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1584
1585         PR middle-end/57748
1586         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
1587         inner_reference_p.
1588         (expand_expr, expand_normal): Adjust.
1589         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
1590         inner_reference_p. Use inner_reference_p to expand inner references.
1591         (store_expr): Adjust.
1592         * cfgexpand.c (expand_call_stmt): Adjust.
1593
1594 2014-01-08  Rong Xu  <xur@google.com>
1595
1596         * gcov-io.c (gcov_var): Move from gcov-io.h.
1597         (gcov_position): Ditto.
1598         (gcov_is_error): Ditto.
1599         (gcov_rewrite): Ditto.
1600         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
1601         only part to libgcc/libgcov.h.
1602
1603 2014-01-08  Marek Polacek  <polacek@redhat.com>
1604
1605         PR middle-end/59669
1606         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
1607
1608 2014-01-08  Marek Polacek  <polacek@redhat.com>
1609
1610         PR sanitizer/59667
1611         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
1612
1613 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
1614
1615         PR rtl-optimization/59649
1616         * stor-layout.c (get_mode_bounds): For BImode return
1617         0 and STORE_FLAG_VALUE.
1618
1619 2014-01-08  Richard Biener  <rguenther@suse.de>
1620
1621         PR middle-end/59630
1622         * gimple.h (is_gimple_builtin_call): Remove.
1623         (gimple_builtin_call_types_compatible_p): New.
1624         (gimple_call_builtin_p): New overload.
1625         * gimple.c (is_gimple_builtin_call): Remove.
1626         (validate_call): Rename to ...
1627         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
1628         check return types.
1629         (validate_type): New static function.
1630         (gimple_call_builtin_p): New overload and adjust.
1631         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
1632         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
1633         (gimple_fold_stmt_to_constant_1): Likewise.
1634         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
1635
1636 2014-01-08  Richard Biener  <rguenther@suse.de>
1637
1638         PR middle-end/59471
1639         * gimplify.c (gimplify_expr): Gimplify register-register type
1640         VIEW_CONVERT_EXPRs to separate stmts.
1641
1642 2014-01-07  Jeff Law  <law@redhat.com>
1643
1644         PR middle-end/53623
1645         * ree.c (combine_set_extension): Handle case where source
1646         and destination registers in an extension insn are different.
1647         (combine_reaching_defs): Allow source and destination registers
1648         in extension to be different under limited circumstances.
1649         (add_removable_extension): Remove restriction that the
1650         source and destination registers in the extension are the same.
1651         (find_and_remove_re): Emit a copy from the extension's
1652         destination to its source after the defining insn if
1653         the source and destination registers are different.
1654
1655         PR middle-end/59285
1656         * ifcvt.c (merge_if_block): If we are merging a block with more than
1657         one successor with a block with no successors, remove any BARRIER
1658         after the second block.
1659
1660 2014-01-07  Dan Xio Qiang <ziyan01@163.com>
1661
1662         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
1663
1664 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
1665
1666         PR target/59652
1667         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
1668         for 14-bit register offsets when INT14_OK_STRICT is false.
1669
1670 2014-01-07  Roland Stigge  <stigge@antcom.de>
1671             Michael Meissner  <meissner@linux.vnet.ibm.com>
1672
1673         PR 57386/target
1674         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
1675         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
1676
1677 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
1678
1679         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
1680         -mcpu.
1681
1682 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
1683
1684         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
1685         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
1686         rtx is const0_rtx or not.
1687
1688 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
1689
1690         PR target/58115
1691         * target-globals.c (save_target_globals): Remove this_fn_optab
1692         handling.
1693         * toplev.c: Include optabs.h.
1694         (target_reinit): Temporarily restore the global options if another
1695         set of options are in force.
1696
1697 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
1698
1699         PR rtl-optimization/58668
1700         * cfgcleanup.c (flow_find_cross_jump): Don't count
1701         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
1702         to determine what is counted.
1703         (flow_find_head_matching_sequence): Use active_insn_p to determine
1704         what is counted.
1705         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
1706         counting change.
1707         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
1708         determine what is counted.
1709
1710         PR tree-optimization/59643
1711         * tree-predcom.c (split_data_refs_to_components): If one dr is
1712         read and one write, determine_offset fails and the write isn't
1713         in the bad component, just put the read into the bad component.
1714
1715 2014-01-07  Mike Stump  <mikestump@comcast.net>
1716             Jakub Jelinek  <jakub@redhat.com>
1717
1718         PR pch/59436
1719         * tree-core.h (struct tree_optimization_option): Change optabs
1720         type from unsigned char * to void *.
1721         * optabs.c (init_tree_optimization_optabs): Adjust
1722         TREE_OPTIMIZATION_OPTABS initialization.
1723
1724 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
1725
1726         PR target/59644
1727         * config/i386/i386.h (struct machine_function): Add
1728         no_drap_save_restore field.
1729         * config/i386/i386.c (ix86_save_reg): Use
1730         !cfun->machine->no_drap_save_restore instead of
1731         crtl->stack_realign_needed.
1732         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
1733         this function clears frame_pointer_needed.  Set
1734         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
1735         and DRAP reg is needed.
1736
1737 2014-01-06  Marek Polacek  <polacek@redhat.com>
1738
1739         PR c/57773
1740         * doc/implement-c.texi: Mention that other integer types are
1741         permitted as bit-field types in strictly conforming mode.
1742
1743 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
1744
1745         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
1746         is newly allocated.
1747
1748 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
1749
1750         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
1751
1752 2014-01-06  Martin Jambor  <mjambor@suse.cz>
1753
1754         PR ipa/59008
1755         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
1756         to int.
1757         * ipa-prop.c (ipa_print_node_params): Fix indentation.
1758
1759 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
1760
1761         PR debug/59350
1762         PR debug/59510
1763         * var-tracking.c (add_stores): Preserve the value of the source even if
1764         we don't record the store.
1765
1766 2014-01-06  Terry Guo  <terry.guo@arm.com>
1767
1768         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
1769
1770 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
1771
1772         PR bootstrap/59541
1773         * config/darwin.c (darwin_function_section): Adjust return values to
1774         correspond to optimisation changes made in r206070.
1775
1776 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
1777
1778         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
1779         from prefetch_block tune setting.
1780         (nocona_cost): Correct size of prefetch block to 64.
1781
1782 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
1783
1784         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
1785         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
1786         used to save the static chain register in the computation of the offset
1787         from which the FP registers need to be restored.
1788
1789 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
1790
1791         PR tree-optimization/59519
1792         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
1793         ICE if get_current_def (current_new_name) is already non-NULL, as long
1794         as it is a phi result of some other phi in *new_exit_bb that has
1795         the same argument.
1796
1797         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
1798         or vmovdqu* for misaligned_operand.
1799         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
1800         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
1801         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
1802         aligned_mem for AVX512F masked aligned load and store builtins and for
1803         non-temporal moves.
1804
1805 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
1806
1807         PR tree-optimization/59651
1808         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
1809         Address range for negative step should be added by TYPE_SIZE_UNIT.
1810
1811 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
1812
1813         * config/m68k/m68k.c (handle_move_double): Handle pushes with
1814         overlapping registers also for registers other than the stack pointer.
1815
1816 2014-01-03  Marek Polacek  <polacek@redhat.com>
1817
1818         PR other/59661
1819         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
1820         __builtin_FILE.
1821
1822 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
1823
1824         PR target/59625
1825         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
1826         asm goto as jump.
1827
1828         * config/i386/i386.md (MODE_SIZE): New mode attribute.
1829         (push splitter): Use <P:MODE_SIZE> instead of
1830         GET_MODE_SIZE (<P:MODE>mode).
1831         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
1832         (mov -1, reg peephole2): Likewise.
1833         * config/i386/sse.md (*mov<mode>_internal,
1834         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
1835         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
1836         *<code><mode>3, *andnot<mode>3<mask_name>,
1837         <mask_codefor><code><mode>3<mask_name>): Likewise.
1838         * config/i386/subst.md (mask_mode512bit_condition,
1839         sd_mask_mode512bit_condition): Likewise.
1840
1841 2014-01-02  Xinliang David Li  <davidxl@google.com>
1842
1843         PR tree-optimization/59303
1844         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
1845         (dump_predicates): Better output format.
1846         (pred_equal_p): New function.
1847         (is_neq_relop_p): Ditto.
1848         (is_neq_zero_form_p): Ditto.
1849         (pred_expr_equal_p): Ditto.
1850         (pred_neg_p): Ditto.
1851         (simplify_pred): Ditto.
1852         (simplify_preds_2): Ditto.
1853         (simplify_preds_3): Ditto.
1854         (simplify_preds_4): Ditto.
1855         (simplify_preds): Ditto.
1856         (push_pred): Ditto.
1857         (push_to_worklist): Ditto.
1858         (get_pred_info_from_cmp): Ditto.
1859         (is_degenerated_phi): Ditto.
1860         (normalize_one_pred_1): Ditto.
1861         (normalize_one_pred): Ditto.
1862         (normalize_one_pred_chain): Ditto.
1863         (normalize_preds): Ditto.
1864         (normalize_cond_1): Remove function.
1865         (normalize_cond): Ditto.
1866         (is_gcond_subset_of): Ditto.
1867         (is_subset_of_any): Ditto.
1868         (is_or_set_subset_of): Ditto.
1869         (is_and_set_subset_of): Ditto.
1870         (is_norm_cond_subset_of): Ditto.
1871         (pred_chain_length_cmp): Ditto.
1872         (convert_control_dep_chain_into_preds): Type change.
1873         (find_predicates): Ditto.
1874         (find_def_preds): Ditto.
1875         (destroy_predicates_vecs): Ditto.
1876         (find_matching_predicates_in_rest_chains): Ditto.
1877         (use_pred_not_overlap_with_undef_path_pred): Ditto.
1878         (is_pred_expr_subset): Ditto.
1879         (is_pred_chain_subset_of): Ditto.
1880         (is_included_in): Ditto.
1881         (is_superset_of): Ditto.
1882
1883 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
1884
1885         Update copyright years.
1886
1887 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
1888
1889         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
1890         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
1891         config/arc/arc.md, config/arc/arc.opt,
1892         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
1893         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
1894         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
1895         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
1896         config/linux-protos.h, config/linux.c, config/winnt-c.c,
1897         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
1898         vtable-verify.c, vtable-verify.h: Use the standard form for the
1899         copyright notice.
1900
1901 2014-01-02  Tobias Burnus  <burnus@net-b.de>
1902
1903         * gcc.c (process_command): Update copyright notice dates.
1904         * gcov-dump.c: Ditto.
1905         * gcov.c: Ditto.
1906         * doc/cpp.texi: Bump @copying's copyright year.
1907         * doc/cppinternals.texi: Ditto.
1908         * doc/gcc.texi: Ditto.
1909         * doc/gccint.texi: Ditto.
1910         * doc/gcov.texi: Ditto.
1911         * doc/install.texi: Ditto.
1912         * doc/invoke.texi: Ditto.
1913
1914 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
1915
1916         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
1917
1918 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
1919
1920         * config/i386/sse.md (*mov<mode>_internal): Guard
1921         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
1922
1923         PR rtl-optimization/59647
1924         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
1925         new_rtx into UNSIGNED_FLOAT rtxes.
1926 \f
1927 Copyright (C) 2014 Free Software Foundation, Inc.
1928
1929 Copying and distribution of this file, with or without modification,
1930 are permitted in any medium without royalty provided the copyright
1931 notice and this notice are preserved.