gcc/ChangeLog

   1 2014-01-30  Jason Merrill  <jason@redhat.com>
   2
   3         PR c++/59633
   4         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
   5
   6         PR c++/59645
   7         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
   8
   9 2014-01-30  Richard Biener  <rguenther@suse.de>
  10
  11         PR tree-optimization/59951
  12         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug
  13         insns.
  14
  15 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
  16
  17         PR target/59784
  18         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
  19         SFmode to DFmode case.
  20
  21 2014-01-29  DJ Delorie  <dj@redhat.com>
  22
  23         * config/msp430/msp430.opt (-minrt): New.
  24         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
  25         if -minrt given.
  26         (ENDFILE_SPEC): Likewise.
  27
  28 2014-01-29  Jan Hubicka  <jh@suse.cz>
  29
  30         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
  31         (estimate_function_body_sizes): Use it.
  32
  33 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
  34
  35         PR c++/58561
  36         * dwarf2out.c (is_cxx_auto): New.
  37         (is_base_type): Use it.
  38         (gen_type_die_with_usage): Likewise.
  39
  40 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
  41
  42         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
  43         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
  44         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
  45         -maltivec=be with LE targets.
  46         (vsx_mergeh_<mode>): Likewise.
  47         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New
  48         unspecs.
  49         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
  50         (altivec_vmrghb): Replace with define_expand and new
  51         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE
  52         targets.
  53         (altivec_vmrghb_direct): New define_insn.
  54         (altivec_vmrghh): Replace with define_expand and new
  55         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE
  56         targets.
  57         (altivec_vmrghh_direct): New define_insn.
  58         (altivec_vmrghw): Replace with define_expand and new
  59         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE
  60         targets.
  61         (altivec_vmrghw_direct): New define_insn.
  62         (*altivec_vmrghsf): Adjust for endianness.
  63         (altivec_vmrglb): Replace with define_expand and new
  64         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE
  65         targets.
  66         (altivec_vmrglb_direct): New define_insn.
  67         (altivec_vmrglh): Replace with define_expand and new
  68         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE
  69         targets.
  70         (altivec_vmrglh_direct): New define_insn.
  71         (altivec_vmrglw): Replace with define_expand and new
  72         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE
  73         targets.
  74         (altivec_vmrglw_direct): New define_insn.
  75         (*altivec_vmrglsf): Adjust for endianness.
  76         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
  77         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
  78         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
  79         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
  80         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
  81         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
  82         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
  83         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
  84
  85 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
  86
  87         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
  88         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
  89         whitespace.
  90
  91 2014-01-29  Richard Biener  <rguenther@suse.de>
  92
  93         PR tree-optimization/58742
  94         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
  95         associate_pointerplus_align.
  96         (associate_pointerplus_diff): New function.
  97         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
  98         and associate_pointerplus_diff.
  99
 100 2014-01-29  Richard Biener  <rguenther@suse.de>
 101
 102         * lto-streamer.h (LTO_major_version): Bump to 3.
 103         (LTO_minor_version): Reset to 0.
 104
 105 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
 106
 107         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
 108         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
 109         (arm_file_start): Generate correct asm header for armv7ve.
 110         * config/arm/bpabi.h: Add multilib support for armv7ve.
 111         * config/arm/driver-arm.c: Change the architectures of cortex-a7
 112         and cortex-a15 to armv7ve.
 113         * config/arm/t-aprofile: Add multilib support for armv7ve.
 114         * doc/invoke.texi: Document -march=armv7ve.
 115
 116 2014-01-29  Richard Biener  <rguenther@suse.de>
 117
 118         PR tree-optimization/58742
 119         * tree-ssa-forwprop.c (associate_plusminus): Return true
 120         if we changed sth, defer EH cleanup to ...
 121         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
 122         (simplify_mult): New function.
 123
 124 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
 125
 126         PR middle-end/59917
 127         PR tree-optimization/59920
 128         * tree.c (build_common_builtin_nodes): Remove
 129         __builtin_setjmp_dispatcher initialization.
 130         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
 131         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
 132         instead of gsi_after_labels + manually skipping debug stmts.
 133         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
 134         ignore bbs with IFN_ABNORMAL_DISPATCHER.
 135         * tree-inline.c (copy_edges_for_bb): Remove
 136         can_make_abnormal_goto argument, instead add abnormal_goto_dest
 137         argument.  Ignore computed_goto_p stmts.  Don't call
 138         make_abnormal_goto_edges.  If a call might need abnormal edges
 139         for non-local gotos, see if it already has an edge to
 140         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
 141         with true argument, don't do anything then, otherwise add
 142         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
 143         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
 144         caller.
 145         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
 146         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
 147         (lower_stmt): Don't set data->calls_builtin_setjmp.
 148         (lower_builtin_setjmp): Adjust comment.
 149         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
 150         * tree-cfg.c (found_computed_goto): Remove.
 151         (factor_computed_gotos): Remove.
 152         (make_goto_expr_edges): Return bool, true for computed gotos.
 153         Don't call make_abnormal_goto_edges.
 154         (build_gimple_cfg): Don't set found_computed_goto, don't call
 155         factor_computed_gotos.
 156         (computed_goto_p): No longer static.
 157         (make_blocks): Don't set found_computed_goto.
 158         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
 159         (make_edges): If make_goto_expr_edges returns true, push bb
 160         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
 161         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
 162         vector.  Record mapping between bbs and OpenMP regions if there
 163         are any, adjust make_gimple_omp_edges caller.  Call
 164         handle_abnormal_edges.
 165         (make_abnormal_goto_edges): Remove.
 166         * tree-cfg.h (make_abnormal_goto_edges): Remove.
 167         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
 168         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
 169         * builtins.c (expand_builtin): Don't handle
 170         BUILT_IN_SETJMP_DISPATCHER.
 171         * internal-fn.def (ABNORMAL_DISPATCHER): New.
 172         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
 173         filling *region also set *region_idx to (*region)->entry->index.
 174
 175         PR other/58712
 176         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
 177         For REGs set ORIGINAL_REGNO.
 178
 179 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
 180
 181         * doc/md.texi: Mention that a target shouldn't implement
 182         vec_widen_(s|u)mul_even/odd pair if it is less efficient
 183         than hi/lo pair.
 184
 185 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
 186
 187         PR tree-optimization/59594
 188         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
 189         a copy of the datarefs vector rather than the vector itself.
 190
 191 2014-01-28  Jason Merrill  <jason@redhat.com>
 192
 193         PR c++/53756
 194         * dwarf2out.c (auto_die): New static.
 195         (gen_type_die_with_usage): Handle C++1y 'auto'.
 196         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
 197         on definition.
 198
 199 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
 200
 201         PR target/59672
 202         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
 203         (SPEC_X32): Likewise.
 204         (SPEC_64): Likewise.
 205         * config/i386/i386.c (ix86_option_override_internal): Turn off
 206         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
 207         for TARGET_16BIT.
 208         (x86_file_start): Output .code16gcc for TARGET_16BIT.
 209         * config/i386/i386.h (TARGET_16BIT): New macro.
 210         (TARGET_16BIT_P): Likewise.
 211         * config/i386/i386.opt: Add m16.
 212         * doc/invoke.texi: Document -m16.
 213
 214 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
 215
 216         PR preprocessor/59935
 217         * input.c (location_get_source_line): Bail out on when line number
 218         is zero, and test the return value of
 219         lookup_or_add_file_to_cache_tab.
 220
 221 2014-01-28  Richard Biener  <rguenther@suse.de>
 222
 223         PR tree-optimization/58742
 224         * tree-ssa-forwprop.c (associate_plusminus): Handle
 225         pointer subtraction of the form (T)(P + A) - (T)P.
 226
 227 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 228
 229         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
 230         at const_int_cost.
 231
 232 2014-01-28  Richard Biener  <rguenther@suse.de>
 233
 234         Revert
 235         2014-01-28  Richard Biener  <rguenther@suse.de>
 236
 237         PR rtl-optimization/45364
 238         PR rtl-optimization/59890
 239         * var-tracking.c (local_get_addr_clear_given_value): Handle
 240         already cleared slot.
 241         (val_reset): Handle not allocated local_get_addr_cache.
 242         (vt_find_locations): Use post-order on the inverted CFG.
 243
 244 2014-01-28  Richard Biener  <rguenther@suse.de>
 245
 246         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps):
 247         Remove.
 248
 249 2014-01-28  Richard Biener  <rguenther@suse.de>
 250
 251         PR rtl-optimization/45364
 252         PR rtl-optimization/59890
 253         * var-tracking.c (local_get_addr_clear_given_value): Handle
 254         already cleared slot.
 255         (val_reset): Handle not allocated local_get_addr_cache.
 256         (vt_find_locations): Use post-order on the inverted CFG.
 257
 258 2014-01-28  Alan Modra  <amodra@gmail.com>
 259
 260         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
 261         * configure.ac <recursive call for build != host>: Define
 262         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
 263         and LD_FOR_BUILD too.
 264         * configure: Regenerate.
 265
 266 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
 267
 268         * config/i386/i386.c (get_builtin_code_for_version): Separate
 269         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
 270         Broadwell from Haswell.
 271
 272 2014-01-27  Steve Ellcey  <sellcey@mips.com>
 273
 274         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
 275         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
 276         * config/mips/mips.c (mips_option_override): Change setting
 277         of TARGET_DSP.
 278         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
 279         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
 280         Change from Mask to Var.
 281
 282 2014-01-27  Jeff Law  <law@redhat.com>
 283
 284         * ipa-inline.c (inline_small_functions): Fix typo.
 285
 286 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
 287
 288         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
 289         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
 290         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
 291         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
 292         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
 293         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
 294         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
 295         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
 296         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
 297         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
 298         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
 299         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
 300         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
 301         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
 302         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
 303         (_mm512_storeu_epi64): Ditto.
 304         (_mm512_cmpge_epi32_mask): Ditto.
 305         (_mm512_cmpge_epu32_mask): Ditto.
 306         (_mm512_cmpge_epi64_mask): Ditto.
 307         (_mm512_cmpge_epu64_mask): Ditto.
 308         (_mm512_cmple_epi32_mask): Ditto.
 309         (_mm512_cmple_epu32_mask): Ditto.
 310         (_mm512_cmple_epi64_mask): Ditto.
 311         (_mm512_cmple_epu64_mask): Ditto.
 312         (_mm512_cmplt_epi32_mask): Ditto.
 313         (_mm512_cmplt_epu32_mask): Ditto.
 314         (_mm512_cmplt_epi64_mask): Ditto.
 315         (_mm512_cmplt_epu64_mask): Ditto.
 316         (_mm512_cmpneq_epi32_mask): Ditto.
 317         (_mm512_cmpneq_epu32_mask): Ditto.
 318         (_mm512_cmpneq_epi64_mask): Ditto.
 319         (_mm512_cmpneq_epu64_mask): Ditto.
 320         (_mm512_expand_pd): Ditto.
 321         (_mm512_expand_ps): Ditto.
 322         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
 323         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
 324         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
 325         * config/i386/i386.c (ix86_builtins): Add
 326         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
 327         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
 328         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
 329         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
 330         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
 331         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
 332         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
 333         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
 334         IX86_BUILTIN_PMOVUSQW512_MEM.
 335         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
 336         __builtin_ia32_pmovsqd512mem_mask,
 337         __builtin_ia32_pmovqd512mem_mask,
 338         __builtin_ia32_pmovusqw512mem_mask,
 339         __builtin_ia32_pmovsqw512mem_mask,
 340         __builtin_ia32_pmovqw512mem_mask,
 341         __builtin_ia32_pmovusdw512mem_mask,
 342         __builtin_ia32_pmovsdw512mem_mask,
 343         __builtin_ia32_pmovdw512mem_mask,
 344         __builtin_ia32_pmovqb512mem_mask,
 345         __builtin_ia32_pmovusqb512mem_mask,
 346         __builtin_ia32_pmovsqb512mem_mask,
 347         __builtin_ia32_pmovusdb512mem_mask,
 348         __builtin_ia32_pmovsdb512mem_mask,
 349         __builtin_ia32_pmovdb512mem_mask.
 350         (bdesc_args): Add __builtin_ia32_expanddf512,
 351         __builtin_ia32_expandsf512.
 352         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
 353         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
 354         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
 355         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
 356         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
 357         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
 358         (avx512f_<code>v8div16qi2_mask_store): This.
 359         (avx512f_expand<mode>): New.
 360
 361 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
 362
 363         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
 364         New.
 365         (_mm512_mask_prefetch_i64gather_pd): Ditto.
 366         (_mm512_prefetch_i32scatter_pd): Ditto.
 367         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
 368         (_mm512_prefetch_i64scatter_pd): Ditto.
 369         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
 370         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
 371         (_mm512_mask_prefetch_i64gather_ps): Ditto.
 372         (_mm512_prefetch_i32scatter_ps): Ditto.
 373         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
 374         (_mm512_prefetch_i64scatter_ps): Ditto.
 375         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
 376         * config/i386/i386-builtin-types.def: Define
 377         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
 378         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
 379         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
 380         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
 381         IX86_BUILTIN_SCATTERPFQPD.
 382         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
 383         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
 384         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
 385         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
 386         __builtin_ia32_scatterpfqps.
 387         (ix86_expand_builtin): Expand new built-ins.
 388         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
 389         fix memory access data type.
 390         (*avx512pf_gatherpf<mode>_mask): Ditto.
 391         (*avx512pf_gatherpf<mode>): Ditto.
 392         (avx512pf_scatterpf<mode>): Ditto.
 393         (*avx512pf_scatterpf<mode>_mask): Ditto.
 394         (*avx512pf_scatterpf<mode>): Ditto.
 395         (GATHER_SCATTER_SF_MEM_MODE): New.
 396         (avx512pf_gatherpf<mode>df): Ditto.
 397         (*avx512pf_gatherpf<mode>df_mask): Ditto.
 398         (*avx512pf_scatterpf<mode>df): Ditto.
 399
 400 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
 401
 402         PR bootstrap/59934
 403         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
 404         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
 405         reached.
 406
 407 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
 408
 409         * common/config/arm/arm-common.c
 410         (arm_rewrite_mcpu): Handle multiple names.
 411         * config/arm/arm.h
 412         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
 413
 414 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
 415
 416         * gimple-builder.h (create_gimple_tmp): Delete.
 417
 418 2014-01-27  Christian Bruel  <christian.bruel@st.com>
 419
 420         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
 421         words comparisons.
 422
 423 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
 424
 425         * config/pa/pa.md (call): Generate indirect long calls to non-local
 426         functions when outputing 32-bit code.
 427         (call_value): Likewise except for special call to buggy powf function.
 428
 429         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
 430         portable runtime and PIC indirect calls.
 431         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
 432         and PIC call sequences.  Use ldo instead of blr to set return register
 433         in PIC call sequence.
 434
 435 2014-01-25  Walter Lee  <walt@tilera.com>
 436
 437         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
 438         avoid clobbering a live register.
 439
 440 2014-01-25  Walter Lee  <walt@tilera.com>
 441
 442         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
 443         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
 444         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
 445         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
 446
 447 2014-01-25  Walter Lee  <walt@tilera.com>
 448
 449         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
 450         arguments on even registers.
 451         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
 452         STACK_BOUNDARY.
 453         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
 454         (BIGGEST_ALIGNMENT): Ditto.
 455         (BIGGEST_FIELD_ALIGNMENT): Ditto.
 456
 457 2014-01-25  Walter Lee  <walt@tilera.com>
 458
 459         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
 460         insns before bundling.
 461         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
 462
 463 2014-01-25  Walter Lee  <walt@tilera.com>
 464
 465         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
 466         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
 467         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
 468
 469 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
 470
 471         * config/mips/constraints.md (kl): Delete.
 472         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
 473         define expands, using...
 474         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
 475         instructions for MIPS16.
 476         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
 477         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
 478
 479 2014-01-25  Walter Lee  <walt@tilera.com>
 480
 481         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
 482         (clzdi2): Ditto.
 483         (ffsdi2): Ditto.
 484
 485 2014-01-25  Walter Lee  <walt@tilera.com>
 486
 487         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
 488         (TARGET_EXPAND_TO_RTL_HOOK): Define.
 489
 490 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
 491
 492         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
 493         Handle XOR.
 494
 495 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
 496
 497         * print-rtl.c (in_call_function_usage): New var.
 498         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
 499         EXPR_LIST mode as mode and not as reg note name.
 500
 501         PR middle-end/59561
 502         * cfgloopmanip.c (copy_loop_info): If
 503         loop->warned_aggressive_loop_optimizations, make sure
 504         the flag is set in target loop too.
 505
 506 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
 507
 508         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
 509         flag_cilkplus.
 510         * builtins.def: Likewise.
 511         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
 512         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
 513         * ira.c (ira_setup_eliminable_regset): Likewise.
 514         * omp-low.c (gate_expand_omp): Likewise.
 515         (execute_lower_omp): Likewise.
 516         (diagnose_sb_0): Likewise.
 517         (gate_diagnose_omp_blocks): Likewise.
 518         (simd_clone_clauses_extract): Likewise.
 519         (gate): Likewise.
 520
 521 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
 522
 523         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
 524         correction for little endian...
 525         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
 526         here.
 527
 528 2014-01-24  Jeff Law  <law@redhat.com>
 529
 530         PR tree-optimization/59919
 531         * tree-vrp.c (find_assert_locations_1): Do not register asserts
 532         for non-returning calls.
 533
 534 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
 535
 536         * common/config/aarch64/aarch64-common.c
 537         (aarch64_rewrite_mcpu): Handle multiple names.
 538         * config/aarch64/aarch64.h
 539         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
 540
 541 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
 542
 543         * input.c (add_file_to_cache_tab): Handle the case where fopen
 544         returns NULL.
 545
 546 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
 547
 548         PR target/59929
 549         * config/i386/i386.md (pushsf splitter): Get stack adjustment
 550         from push operand if code of push isn't PRE_DEC.
 551
 552 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
 553
 554         PR target/59909
 555         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
 556         -mquad-memory-atomic.  Update -mquad-memory documentation to say
 557         it is only used for non-atomic loads/stores.
 558
 559         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
 560         -mquad-memory or -mquad-memory-atomic switches.
 561
 562         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
 563         -mquad-memory-atomic to ISA 2.07 support.
 564
 565         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
 566         to separate support of normal quad word memory operations (ldq, stq)
 567         from the atomic quad word memory operations.
 568
 569         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
 570         support to separate non-atomic quad word operations from atomic
 571         quad word operations.  Disable non-atomic quad word operations in
 572         little endian mode so that we don't have to swap words after the
 573         load and before the store.
 574         (quad_load_store_p): Add comment about atomic quad word support.
 575         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
 576         options printed with -mdebug=reg.
 577
 578         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
 579         -mquad-memory-atomic as the test for whether we have quad word
 580         atomic instructions.
 581         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
 582         or -mp8-vector are used, allow byte/half-word atomic operations.
 583
 584         * config/rs6000/sync.md (load_lockedti): Insure that the address
 585         is a proper indexed or indirect address for the lqarx instruction.
 586         On little endian systems, swap the hi/lo registers after the lqarx
 587         instruction.
 588         (load_lockedpti): Use indexed_or_indirect_operand predicate to
 589         insure the address is valid for the lqarx instruction.
 590         (store_conditionalti): Insure that the address is a proper indexed
 591         or indirect address for the stqcrx. instruction.  On little endian
 592         systems, swap the hi/lo registers before doing the stqcrx.
 593         instruction.
 594         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
 595         insure the address is valid for the stqcrx. instruction.
 596
 597         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
 598         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
 599         type of quad memory support is available.
 600
 601 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
 602
 603         PR regression/59915
 604         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
 605         there is a danger of looping.
 606
 607 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
 608
 609         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
 610         force flag_ira_loop_pressure if set via command line.
 611
 612 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
 613
 614         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
 615         (ashr_simd): New builtin handling DI mode.
 616         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
 617         (aarch64_sshr_simddi): New match pattern.
 618         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
 619         (vshrd_n_s64): Likewise.
 620         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
 621
 622 2014-01-23  Nick Clifton  <nickc@redhat.com>
 623
 624         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
 625         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
 626         favour of mcu specific scripts.
 627         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
 628         430x multilibs.
 629
 630 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
 631             Alex Velenko  <Alex.Velenko@arm.com>
 632
 633         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
 634         (vaddv_s16): Likewise.
 635         (vaddv_s32): Likewise.
 636         (vaddv_u8): Likewise.
 637         (vaddv_u16): Likewise.
 638         (vaddv_u32): Likewise.
 639         (vaddvq_s8): Likewise.
 640         (vaddvq_s16): Likewise.
 641         (vaddvq_s32): Likewise.
 642         (vaddvq_s64): Likewise.
 643         (vaddvq_u8): Likewise.
 644         (vaddvq_u16): Likewise.
 645         (vaddvq_u32): Likewise.
 646         (vaddvq_u64): Likewise.
 647         (vaddv_f32): Likewise.
 648         (vaddvq_f32): Likewise.
 649         (vaddvq_f64): Likewise.
 650         (vmaxv_f32): Likewise.
 651         (vmaxv_s8): Likewise.
 652         (vmaxv_s16): Likewise.
 653         (vmaxv_s32): Likewise.
 654         (vmaxv_u8): Likewise.
 655         (vmaxv_u16): Likewise.
 656         (vmaxv_u32): Likewise.
 657         (vmaxvq_f32): Likewise.
 658         (vmaxvq_f64): Likewise.
 659         (vmaxvq_s8): Likewise.
 660         (vmaxvq_s16): Likewise.
 661         (vmaxvq_s32): Likewise.
 662         (vmaxvq_u8): Likewise.
 663         (vmaxvq_u16): Likewise.
 664         (vmaxvq_u32): Likewise.
 665         (vmaxnmv_f32): Likewise.
 666         (vmaxnmvq_f32): Likewise.
 667         (vmaxnmvq_f64): Likewise.
 668         (vminv_f32): Likewise.
 669         (vminv_s8): Likewise.
 670         (vminv_s16): Likewise.
 671         (vminv_s32): Likewise.
 672         (vminv_u8): Likewise.
 673         (vminv_u16): Likewise.
 674         (vminv_u32): Likewise.
 675         (vminvq_f32): Likewise.
 676         (vminvq_f64): Likewise.
 677         (vminvq_s8): Likewise.
 678         (vminvq_s16): Likewise.
 679         (vminvq_s32): Likewise.
 680         (vminvq_u8): Likewise.
 681         (vminvq_u16): Likewise.
 682         (vminvq_u32): Likewise.
 683         (vminnmv_f32): Likewise.
 684         (vminnmvq_f32): Likewise.
 685         (vminnmvq_f64): Likewise.
 686
 687 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
 688
 689         * config/aarch64/aarch64-simd.md
 690         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
 691         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
 692         (*aarch64_mul3_elt<mode>): Likewise.
 693         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
 694         (*aarch64_mul3_elt_to_64v2df): Likewise.
 695         (*aarch64_mla_elt<mode>): Likewise.
 696         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
 697         (*aarch64_mls_elt<mode>): Likewise.
 698         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
 699         (*aarch64_fma4_elt<mode>): Likewise.
 700         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
 701         (*aarch64_fma4_elt_to_64v2df): Likewise.
 702         (*aarch64_fnma4_elt<mode>): Likewise.
 703         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
 704         (*aarch64_fnma4_elt_to_64v2df): Likewise.
 705         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
 706         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
 707         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
 708         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
 709         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
 710         (aarch64_sqdmull_lane<mode>_internal): Likewise.
 711         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
 712
 713 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
 714
 715         * config/aarch64/aarch64-simd.md
 716         (aarch64_be_checked_get_lane<mode>): New define_expand.
 717         * config/aarch64/aarch64-simd-builtins.def
 718         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
 719         New builtin definition.
 720         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
 721         Use new safe be builtin.
 722
 723 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
 724
 725         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
 726         New define_insn.
 727         (aarch64_be_st1<mode>): Likewise.
 728         (aarch_ld1<VALL:mode>): Define_expand modified.
 729         (aarch_st1<VALL:mode>): Likewise.
 730         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
 731         (UNSPEC_ST1): Likewise.
 732
 733 2014-01-23  David Holsgrove <david.holsgrove@xilinx.com>
 734
 735         * config/microblaze/microblaze.md: Add trap insn and attribute
 736
 737 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
 738
 739         PR preprocessor/58580
 740         * input.h (location_get_source_line): Take an additional line_size
 741         parameter.
 742         (void diagnostics_file_cache_fini): Declare new function.
 743         * input.c (struct fcache): New type.
 744         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
 745         New static constants.
 746         (diagnostic_file_cache_init, total_lines_num)
 747         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
 748         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
 749         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
 750         (get_next_line, read_next_line, goto_next_line, read_line_num):
 751         New static function definitions.
 752         (diagnostic_file_cache_fini): New function.
 753         (location_get_source_line): Take an additional output line_len
 754         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
 755         read_line_num.
 756         * diagnostic.c (diagnostic_finish): Call
 757         diagnostic_file_cache_fini.
 758         (adjust_line): Take an additional input parameter for the length
 759         of the line, rather than calculating it with strlen.
 760         (diagnostic_show_locus): Adjust the use of
 761         location_get_source_line and adjust_line with respect to their new
 762         signature.  While displaying a line now, do not stop at the first
 763         null byte.  Rather, display the zero byte as a space and keep
 764         going until we reach the size of the line.
 765         * Makefile.in: Add vec.o to OBJS-libcommon
 766
 767 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
 768             Ilya Tocar     <ilya.tocar@intel.com>
 769
 770         * config/i386/avx512fintrin.h (_mm512_kmov): New.
 771         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
 772         (__builtin_ia32_kmov16): Ditto.
 773         * config/i386/i386.md (UNSPEC_KMOV): New.
 774         (kmovw): Ditto.
 775
 776 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
 777
 778         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
 779         (_mm512_storeu_si512): Ditto.
 780
 781 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
 782
 783         PR target/52125
 784         * rtl.h (get_referenced_operands): Declare.
 785         * recog.c (get_referenced_operands): New function.
 786         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
 787         operands have been referenced when recording LO_SUM references.
 788
 789 2014-01-22  David Holsgrove <david.holsgrove@xilinx.com>
 790
 791         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
 792
 793 2014-01-22  Jan Hubicka  <jh@suse.cz>
 794
 795         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
 796         Enable for generic and recent AMD targets.
 797
 798 2014-01-22  Jan Hubicka  <jh@suse.cz>
 799
 800         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
 801         ARG_SIZE note when adjustment was eliminated.
 802
 803 2014-01-22  Jeff Law  <law@redhat.com>
 804
 805         PR tree-optimization/59597
 806         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
 807         in file.  Accept new argument REGISTERING and use it to modify
 808         dump output appropriately.
 809         (register_jump_thread): Corresponding changes.
 810         (mark_threaded_blocks): Reinstate code to cancel unprofitable
 811         thread paths involving joiner blocks.  Add code to dump cancelled
 812         jump threading paths.
 813
 814 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
 815
 816         PR rtl-optimization/59477
 817         * lra-constraints.c (inherit_in_ebb): Process call for living hard
 818         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
 819
 820 2014-01-22  Tom Tromey  <tromey@redhat.com>
 821
 822         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
 823         PARAMS.
 824         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
 825
 826 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
 827
 828         PR rtl-optimization/59896
 829         * lra-constraints.c (process_alt_operands): Check unused note for
 830         matched operands of insn with no output reloads.
 831
 832 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
 833
 834         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
 835         (mips_move_from_gpr_cost): Likewise.
 836
 837 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
 838
 839         PR rtl-optimization/59858
 840         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
 841         ira_class_hard_regs_num.
 842         (process_alt_operands): Increase reject for dying matched operand.
 843
 844 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
 845
 846         PR target/59003
 847         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
 848         smaller than size, perform several stores or loads and stores
 849         at dst + count - size to store or copy all of size bytes, rather
 850         than just last modesize bytes.
 851
 852 2014-01-20  DJ Delorie  <dj@redhat.com>
 853
 854         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
 855         that CLOBBERs are REGs before propogating their values.
 856
 857 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
 858
 859         PR middle-end/59789
 860         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
 861         (cgraph_inline_failed_type): New function.
 862         * cgraph.h (DEFCIFCODE): Add type.
 863         (cgraph_inline_failed_type_t): New enum.
 864         (cgraph_inline_failed_type): New prototype.
 865         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
 866         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
 867         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
 868         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
 869         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
 870         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
 871         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
 872         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
 873         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
 874         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
 875         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
 876         OPTIMIZATION_MISMATCH.
 877         * tree-inline.c (expand_call_inline): Emit errors during
 878         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
 879
 880 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
 881
 882         PR target/59685
 883         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
 884         mode attribute in insn output.
 885
 886 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
 887
 888         * output.h (output_constant): Delete.
 889         * varasm.c (output_constant): Make private.
 890
 891 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
 892
 893         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
 894
 895 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
 896
 897         PR middle-end/59860
 898         * tree.h (fold_builtin_strcat): New prototype.
 899         * builtins.c (fold_builtin_strcat): No longer static.  Add len
 900         argument, if non-NULL, don't call c_strlen.  Optimize
 901         directly into __builtin_memcpy instead of __builtin_strcpy.
 902         (fold_builtin_2): Adjust fold_builtin_strcat caller.
 903         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
 904
 905 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
 906
 907         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
 908         for SImode_address_operand operands, having only a REG argument.
 909
 910 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
 911
 912         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
 913         loader name using mbig-endian.
 914         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
 915
 916 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
 917
 918         * doc/invoke.texi (-march): Clarify documentation for AArch64.
 919         (-mtune): Likewise.
 920         (-mcpu): Likewise.
 921
 922 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
 923
 924         * config/aarch64/aarch64-protos.h
 925         (aarch64_cannot_change_mode_class_ptr): Declare.
 926         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
 927         aarch64_cannot_change_mode_class_ptr): New.
 928         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
 929         backend hook aarch64_cannot_change_mode_class.
 930
 931 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
 932
 933         * common/config/aarch64/aarch64-common.c
 934         (aarch64_handle_option): Don't handle any option order logic here.
 935         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
 936         selected_cpu, warn on architecture version mismatch.
 937         (aarch64_override_options): Fix parsing order for option strings.
 938
 939 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
 940             Iain Sandoe  <iain@codesourcery.com>
 941
 942         PR bootstrap/59496
 943         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
 944         warning.  Amend comment to reflect current functionality.
 945
 946 2014-01-20  Richard Biener  <rguenther@suse.de>
 947
 948         PR middle-end/59860
 949         * builtins.c (fold_builtin_strcat): Remove case better handled
 950         by tree-ssa-strlen.c.
 951
 952 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
 953
 954         * config/aarch64/aarch64.opt
 955         (mcpu, march, mtune): Make case-insensitive.
 956
 957 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
 958
 959         PR target/59880
 960         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
 961         if operands[1] is a REG or ZERO_EXTEND of a REG.
 962
 963 2014-01-19  Jan Hubicka  <jh@suse.cz>
 964
 965         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
 966
 967 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
 968
 969         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
 970         long non-pic millicode calls.
 971
 972 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
 973
 974         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
 975
 976 2014-01-19  Kito Cheng  <kito@0xlab.org>
 977
 978         * builtins.c (expand_movstr): Check movstr expand done or fail.
 979
 980 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
 981             H.J. Lu  <hongjiu.lu@intel.com>
 982
 983         PR target/59379
 984         * config/i386/i386.md (*lea<mode>): Zero-extend return register
 985         to DImode for zero-extended addresses.
 986
 987 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
 988
 989         PR rtl-optimization/57763
 990         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
 991         on the new indirect jump_insn and increment LABEL_NUSES (label).
 992
 993 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
 994
 995         PR bootstrap/59580
 996         PR bootstrap/59583
 997         * config.gcc (x86_archs): New variable.
 998         (x86_64_archs): Likewise.
 999         (x86_cpus): Likewise.
1000         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
1001         --with-arch/--with-cpu= options.
1002         Support --with-arch=/--with-cpu={nehalem,westmere,
1003         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
1004
1005 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
1006
1007         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
1008         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
1009
1010 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
1011
1012         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
1013
1014 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
1015
1016         PR target/58944
1017         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
1018         clear cpp_get_options (parse_in)->warn_unused_macros for
1019         ix86_target_macros_internal with cpp_define.
1020
1021 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
1022
1023         * jump.c (delete_related_insns): Keep (use (insn))s.
1024         * reorg.c (redundant_insn): Check for barriers too.
1025
1026 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
1027
1028         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
1029
1030 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
1031
1032         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
1033         call to $$dyncall when TARGET_LONG_CALLS is true.
1034
1035 2014-01-17  Jeff Law  <law@redhat.com>
1036
1037         * ree.c (combine_set_extension): Temporarily disable test for
1038         changing number of hard registers.
1039
1040 2014-01-17  Jan Hubicka  <jh@suse.cz>
1041
1042         PR middle-end/58125
1043         * ipa-inline-analysis.c (inline_free_summary):
1044         Do not free summary of aliases.
1045
1046 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
1047
1048         PR middle-end/59706
1049         * gimplify.c (gimplify_expr): Use create_tmp_var
1050         instead of create_tmp_var_raw.  If cond doesn't have
1051         integral type, don't add the IFN_ANNOTATE builtin at all.
1052
1053 2014-01-17  Martin Jambor  <mjambor@suse.cz>
1054
1055         PR ipa/59736
1056         * ipa-cp.c (prev_edge_clone): New variable.
1057         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
1058         Also resize prev_edge_clone vector.
1059         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
1060         (ipcp_edge_removal_hook): New function.
1061         (ipcp_driver): Register ipcp_edge_removal_hook.
1062
1063 2014-01-17  Andrew Pinski <apinski@cavium.com>
1064             Steve Ellcey  <sellcey@mips.com>
1065
1066         PR target/59462
1067         * config/mips/mips.c (mips_print_operand): Check operand mode instead
1068         of operator mode.
1069
1070 2014-01-17  Jeff Law  <law@redhat.com>
1071
1072         PR middle-end/57904
1073         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
1074         so that pass_ccp runs first.
1075
1076 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
1077
1078         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
1079         (ix86_adjust_cost): Use !TARGET_XXX.
1080         (do_reorder_for_imul): Likewise.
1081         (swap_top_of_ready_list): Likewise.
1082         (ix86_sched_reorder): Likewise.
1083
1084 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
1085
1086         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1087         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
1088         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
1089         (intel_memset): New.  Duplicate slm_memset.
1090         (intel_cost): New.  Duplicate slm_cost.
1091         (m_INTEL): New macro.
1092         (processor_target_table): Add "intel".
1093         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
1094         with PROCESSOR_INTEL for "intel".
1095         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
1096         PROCESSOR_SILVERMONT.
1097         (ix86_issue_rate): Likewise.
1098         (ix86_adjust_cost): Likewise.
1099         (ia32_multipass_dfa_lookahead): Likewise.
1100         (swap_top_of_ready_list): Likewise.
1101         (ix86_sched_reorder): Likewise.
1102         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
1103         instead of TARGET_OPT_AGU.
1104         * config/i386/i386.h (TARGET_INTEL): New.
1105         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
1106         (processor_type): Add PROCESSOR_INTEL.
1107         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
1108         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
1109
1110 2014-01-17  Marek Polacek  <polacek@redhat.com>
1111
1112         PR c/58346
1113         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
1114         size is zero.
1115
1116 2014-01-17  Richard Biener  <rguenther@suse.de>
1117
1118         PR tree-optimization/46590
1119         * opts.c (default_options_table): Add entries for
1120         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
1121         all enabled at -O1 but not for -Og.
1122         * common.opt (fbranch-count-reg): Remove Init(1).
1123         (fmove-loop-invariants): Likewise.
1124         (ftree-pta): Likewise.
1125
1126 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
1127
1128         * config/i386/i386.c (ix86_data_alignment): For compatibility with
1129         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
1130         decls to at least the GCC 4.8 used alignments.
1131
1132         PR fortran/59440
1133         * tree-nested.c (convert_nonlocal_reference_stmt,
1134         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
1135         of GIMPLE_BIND stmts, adjust associated decls.
1136
1137 2014-01-17  Richard Biener  <rguenther@suse.de>
1138
1139         PR tree-optimization/46590
1140         * vec.h (vec<>::bseach): New member function implementing
1141         binary search according to C89 bsearch.
1142         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
1143         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
1144         bitmap pointer again.  Make accesses_in_loop a flat array.
1145         (mem_ref_obstack): New global.
1146         (outermost_indep_loop): Adjust for mem_ref->stored changes.
1147         (mark_ref_stored): Likewise.
1148         (ref_indep_loop_p_2): Likewise.
1149         (set_ref_stored_in_loop): New helper function.
1150         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
1151         (memref_free): Adjust.
1152         (record_mem_ref_loc): Simplify.
1153         (gather_mem_refs_stmt): Adjust.
1154         (sort_locs_in_loop_postorder_cmp): New function.
1155         (analyze_memory_references): Sort accesses_in_loop after
1156         loop postorder number.
1157         (find_ref_loc_in_loop_cmp): New function.
1158         (for_all_locs_in_loop): Find relevant cluster of locs in
1159         accesses_in_loop and iterate without recursion.
1160         (execute_sm): Avoid uninit warning.
1161         (struct ref_always_accessed): Simplify.
1162         (ref_always_accessed::operator ()): Likewise.
1163         (ref_always_accessed_p): Likewise.
1164         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
1165         loop postorder numbers here.
1166         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
1167         numbers.
1168
1169 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
1170
1171         PR c++/57945
1172         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
1173         on decls for which assemble_alias has been called.
1174
1175 2014-01-17  Nick Clifton  <nickc@redhat.com>
1176
1177         * config/msp430/msp430.opt: (mcpu): New option.
1178         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
1179         (msp430_option_override): Parse target_cpu.  If the MCU name
1180         matches a generic string, clear target_mcu.
1181         (msp430_attr): Allow numeric interrupt values up to 63.
1182         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
1183         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
1184         option.
1185         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
1186         Add mcpu matches.
1187         * config/msp430/msp430.md (popm): Use %J rather than %I.
1188         (addsi3): Use msp430_nonimmediate_operand for operand 2.
1189         (addhi_cy_i): Use immediate_operand for operand 2.
1190         * doc/invoke.texi: Document -mcpu option.
1191
1192 2014-01-17  Richard Biener  <rguenther@suse.de>
1193
1194         PR rtl-optimization/38518
1195         * df.h (df_analyze_loop): Declare.
1196         * df-core.c: Include cfgloop.h.
1197         (df_analyze_1): Split out main part of df_analyze.
1198         (df_analyze): Adjust.
1199         (loop_inverted_post_order_compute): New function.
1200         (loop_post_order_compute): Likewise.
1201         (df_analyze_loop): New function avoiding whole-function
1202         postorder computes.
1203         * loop-invariant.c (find_defs): Use df_analyze_loop.
1204         (find_invariants): Adjust.
1205         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
1206
1207 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
1208
1209         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
1210         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
1211
1212 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
1213
1214         * ipa-ref.c (ipa_remove_stmt_references): Fix references
1215         traversal when removing references.
1216
1217 2014-01-16  Jan Hubicka  <jh@suse.cz>
1218
1219         PR ipa/59775
1220         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
1221
1222 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
1223
1224         PR middle-end/56791
1225         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
1226         pushing a reload for an autoinc when we had previously reloaded an
1227         inner part of the address.
1228
1229 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
1230
1231         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
1232         field.
1233         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
1234         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
1235         when not giving up or versioning for alias only because of
1236         loop->safelen.
1237         (vect_analyze_data_ref_dependences): Set to true.
1238         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
1239         is a GIMPLE_PHI.
1240         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
1241         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
1242         to the condition.
1243
1244         PR middle-end/58344
1245         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
1246
1247         PR target/59839
1248         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
1249         operand 0 predicate for gathers, use a new pseudo as subtarget.
1250
1251 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
1252
1253         PR middle-end/59609
1254         * lra-constraints.c (process_alt_operands): Add printing debug info.
1255         Check absence of input/output reloads for matched operands too.
1256
1257 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
1258
1259         PR rtl-optimization/59835
1260         * ira.c (ira_init_register_move_cost): Increase cost for
1261         impossible modes.
1262
1263 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
1264
1265         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
1266
1267 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
1268
1269         PR target/59780
1270         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
1271         non-register objects.  Use gen_(high/low)part more consistently.
1272         Fix assertions.
1273
1274 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
1275
1276         PR target/59844
1277         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
1278         endian support, remove tests for WORDS_BIG_ENDIAN.
1279         (p8_mfvsrd_3_<mode>): Likewise.
1280         (reload_gpr_from_vsx<mode>): Likewise.
1281         (reload_gpr_from_vsxsf): Likewise.
1282         (p8_mfvsrd_4_disf): Likewise.
1283
1284 2014-01-16  Richard Biener  <rguenther@suse.de>
1285
1286         PR rtl-optimization/46590
1287         * lcm.c (compute_antinout_edge): Use postorder iteration.
1288         (compute_laterin): Use inverted postorder iteration.
1289
1290 2014-01-16  Nick Clifton  <nickc@redhat.com>
1291
1292         PR middle-end/28865
1293         * varasm.c (output_constant): Return the number of bytes actually
1294         emitted.
1295         (output_constructor_array_range): Update the field size with the
1296         number of bytes emitted by output_constant.
1297         (output_constructor_regular_field): Likewise.  Also do not
1298         complain if the total number of bytes emitted is now greater
1299         than the expected fieldpos.
1300         * output.h (output_constant): Update prototype and descriptive comment.
1301
1302 2014-01-16  Marek Polacek  <polacek@redhat.com>
1303
1304         PR middle-end/59827
1305         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
1306         it is error_mark_node.
1307
1308 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
1309
1310         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
1311         VALID_AVX256_REG_OR_OI_MODE.
1312
1313 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
1314
1315         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
1316         current procedure should be profiled.
1317
1318 2014-01-15  Andrew Pinski  <apinski@cavium.com>
1319
1320         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
1321         of moving from/to the STACK_REG register class.
1322
1323 2014-01-15  Richard Henderson <rth@redhat.com>
1324
1325         PR debug/54694
1326         * reginfo.c (global_regs_decl): Globalize.
1327         * rtl.h (global_regs_decl): Declare.
1328         * ira.c (do_reload): Diagnose frame_pointer_needed and it
1329         reserved via global_regs.
1330
1331 2014-01-15  Teresa Johnson  <tejohnson@google.com>
1332
1333         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
1334
1335 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
1336
1337         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
1338         and vmulosh rather than call gen_vec_widen_smult_*.
1339         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
1340         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
1341         (vec_widen_smult_even_v16qi): Likewise.
1342         (vec_widen_umult_even_v8hi): Likewise.
1343         (vec_widen_smult_even_v8hi): Likewise.
1344         (vec_widen_umult_odd_v16qi): Likewise.
1345         (vec_widen_smult_odd_v16qi): Likewise.
1346         (vec_widen_umult_odd_v8hi): Likewise.
1347         (vec_widen_smult_odd_v8hi): Likewise.
1348         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
1349         vmuloub rather than call gen_vec_widen_umult_*.
1350         (vec_widen_umult_lo_v16qi): Likewise.
1351         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
1352         vmulosb rather than call gen_vec_widen_smult_*.
1353         (vec_widen_smult_lo_v16qi): Likewise.
1354         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
1355         rather than call gen_vec_widen_umult_*.
1356         (vec_widen_umult_lo_v8hi): Likewise.
1357         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
1358         rather than call gen_vec_widen_smult_*.
1359         (vec_widen_smult_lo_v8hi): Likewise.
1360
1361 2014-01-15  Jeff Law  <law@redhat.com>
1362
1363         PR tree-optimization/59747
1364         * ree.c (find_and_remove_re): Properly handle case where a second
1365         eliminated extension requires widening a copy created for elimination
1366         of a prior extension.
1367         (combine_set_extension): Ensure that the number of hard regs needed
1368         for a destination register does not change when we widen it.
1369
1370 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
1371
1372         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
1373         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
1374         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
1375         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
1376         (avr-*-rtems*): Likewise.
1377         (bfin*-rtems*): Likewise.
1378         (moxie-*-rtems*): Likewise.
1379         (h8300-*-rtems*): Likewise.
1380         (i[34567]86-*-rtems*): Likewise.
1381         (lm32-*-rtems*): Likewise.
1382         (m32r-*-rtems*): Likewise.
1383         (m68k-*-rtems*): Likewise.
1384         (microblaze*-*-rtems*): Likewise.
1385         (mips*-*-rtems*): Likewise.
1386         (powerpc-*-rtems*): Likewise.
1387         (sh-*-rtems*): Likewise.
1388         (sparc-*-rtems*): Likewise.
1389         (sparc64-*-rtems*): Likewise.
1390         (v850-*-rtems*): Likewise.
1391         (m32c-*-rtems*): Likewise.
1392
1393 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
1394
1395         PR rtl-optimization/59511
1396         * ira.c (ira_init_register_move_cost): Use memory costs for some
1397         cases of register move cost calculations.
1398         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
1399         instead of BB frequency.
1400         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
1401         * lra-assigns.c (find_hard_regno_for): Ditto.
1402
1403 2014-01-15  Richard Biener  <rguenther@suse.de>
1404
1405         PR tree-optimization/59822
1406         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
1407         (vectorizable_load): Use it to hoist defs of uses of invariant
1408         loads out of the loop.
1409
1410 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
1411             Kugan Vivekanandarajah  <kuganv@linaro.org>
1412
1413         PR target/59695
1414         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
1415         truncation.
1416
1417 2014-01-15  Richard Biener  <rguenther@suse.de>
1418
1419         PR rtl-optimization/59802
1420         * lcm.c (compute_available): Use inverted postorder to seed
1421         the initial worklist.
1422
1423 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
1424
1425         PR target/59803
1426         * config/s390/s390.c (s390_preferred_reload_class): Don't return
1427         ADDR_REGS for invalid symrefs in non-PIC code.
1428
1429 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
1430
1431         PR other/58712
1432         * builtins.c (determine_block_size): Initialize *probable_max_size
1433         even if len_rtx is CONST_INT.
1434
1435 2014-01-14  Andrew Pinski  <apinski@cavium.com>
1436
1437         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
1438         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
1439         (cortexa53_tunings): Likewise.
1440         (aarch64_sched_issue_rate): New function.
1441         (TARGET_SCHED_ISSUE_RATE): Define.
1442
1443 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
1444
1445         * ira-costs.c (find_costs_and_classes): Add missed
1446         ira_init_register_move_cost_if_necessary.
1447
1448 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
1449
1450         PR target/59787
1451         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
1452
1453 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
1454
1455         PR target/59794
1456         * config/i386/i386.c (type_natural_mode): Add a bool parameter
1457         to indicate if type is used for function return value.  Warn ABI
1458         change if the vector mode isn't available for function return value.
1459         (ix86_function_arg_advance): Pass false to type_natural_mode.
1460         (ix86_function_arg): Likewise.
1461         (ix86_gimplify_va_arg): Likewise.
1462         (function_arg_32): Don't warn ABI change.
1463         (ix86_function_value): Pass true to type_natural_mode.
1464         (ix86_return_in_memory): Likewise.
1465         (ix86_struct_value_rtx): Removed.
1466         (TARGET_STRUCT_VALUE_RTX): Likewise.
1467
1468 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
1469
1470         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
1471         converting a conditional jump into a conditional return.
1472
1473 2014-01-14  Richard Biener  <rguenther@suse.de>
1474
1475         PR tree-optimization/58921
1476         PR tree-optimization/59006
1477         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
1478         hoisting invariant stmts.
1479         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
1480         invariant loads on the preheader edge if possible.
1481
1482 2014-01-14  Joey Ye  <joey.ye@arm.com>
1483
1484         * doc/plugin.texi (Building GCC plugins): Update to C++.
1485
1486 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
1487
1488         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
1489         (_mm_rcp28_round_ss): Ditto.
1490         (_mm_rsqrt28_round_sd): Ditto.
1491         (_mm_rsqrt28_round_ss): Ditto.
1492         (_mm_rcp28_sd): Ditto.
1493         (_mm_rcp28_ss): Ditto.
1494         (_mm_rsqrt28_sd): Ditto.
1495         (_mm_rsqrt28_ss): Ditto.
1496         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
1497         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
1498         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
1499         (IX86_BUILTIN_RCP28SD): Ditto.
1500         (IX86_BUILTIN_RCP28SS): Ditto.
1501         (IX86_BUILTIN_RSQRT28SD): Ditto.
1502         (IX86_BUILTIN_RSQRT28SS): Ditto.
1503         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
1504         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
1505         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
1506         (ix86_expand_special_args_builtin): Expand new FTYPE.
1507         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
1508         (srcp14<mode>): Make insn unary.
1509         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
1510         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
1511         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
1512         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
1513         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
1514         Fix rounding: make it SAE only.
1515         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
1516         Ditto.
1517         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
1518         Ditto.
1519         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
1520         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
1521         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
1522         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
1523         (round_saeonly_mask_scalar_operand4): Ditto.
1524         (round_saeonly_mask_scalar_op3): Ditto.
1525         (round_saeonly_mask_scalar_op4): Ditto.
1526
1527 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1528
1529         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1530         Implement -maltivec=be for vec_insert and vec_extract.
1531
1532 2014-01-10  DJ Delorie  <dj@redhat.com>
1533
1534         * config/msp430/msp430.md (call_internal): Don't allow memory
1535         references with SP as the base register.
1536         (call_value_internal): Likewise.
1537         * config/msp430/constraints.md (Yc): New.  For memory references
1538         that don't use SP as a base register.
1539
1540         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
1541         "an integer without a # prefix"
1542         * config/msp430/msp430.md (epilogue_helper): Use it.
1543
1544 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
1545
1546         PR target/59617
1547         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
1548         AVX512F gather builtins.
1549         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
1550         on gather decls with INTEGER_TYPE masktype.
1551         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
1552         directly into the builtin rather than hoisting it before loop.
1553
1554         PR tree-optimization/59387
1555         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
1556         (scev_const_prop): If folded_casts and type has undefined overflow,
1557         use force_gimple_operand instead of force_gimple_operand_gsi and
1558         for each added stmt if it is assign with
1559         arith_code_with_undefined_signed_overflow, call
1560         rewrite_to_defined_overflow.
1561         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
1562         gimple-fold.h instead.
1563         (arith_code_with_undefined_signed_overflow,
1564         rewrite_to_defined_overflow): Moved to ...
1565         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
1566         rewrite_to_defined_overflow): ... here.  No longer static.
1567         Include gimplify-me.h.
1568         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
1569         rewrite_to_defined_overflow): New prototypes.
1570
1571 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1572
1573         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
1574
1575 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
1576
1577         * builtins.c (get_object_alignment_2): Minor tweak.
1578         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
1579
1580 2014-01-13  Christian Bruel  <christian.bruel@st.com>
1581
1582         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
1583         optimized non constant lengths.
1584
1585 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
1586
1587         PR libgomp/59194
1588         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
1589         load as __atomic_load_N if possible.
1590
1591 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
1592
1593         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
1594         target parameter.
1595         (rs6000_expand_builtin): Adjust call.
1596
1597 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
1598
1599         PR target/58115
1600         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
1601         * config/rs6000/rs6000.c: Include target-globals.h.
1602         (rs6000_set_current_function): Instead of doing target_reinit
1603         unconditionally, use save_target_globals_default_opts and
1604         restore_target_globals.
1605
1606         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
1607         FPSCR.
1608         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
1609         (rs6000_expand_builtin): Handle mffs and mtfsf.
1610         (rs6000_init_builtins): Define mffs and mtfsf.
1611         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
1612         (rs6000_mffs): New pattern.
1613         (rs6000_mtfsf): New pattern.
1614
1615 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
1616
1617         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
1618         Start narrowing with START.  Apply candidate-use pair
1619         and check overall cost in narrowing.
1620         (iv_ca_prune): Pass new argument.
1621
1622 2014-01-10  Jeff Law  <law@redhat.com>
1623
1624         PR middle-end/59743
1625         * ree.c (combine_reaching_defs): Ensure the defining statement
1626         occurs before the extension when optimizing extensions with
1627         different source and destination hard registers.
1628
1629 2014-01-10  Jan Hubicka  <jh@suse.cz>
1630
1631         PR ipa/58585
1632         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
1633         vtables into the type inheritance graph.
1634
1635 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
1636
1637         PR rtl-optimization/59754
1638         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
1639         modes in the REGNO != REGNO case.
1640
1641 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1642
1643         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
1644
1645 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
1646
1647         PR tree-optimization/59745
1648         * tree-predcom.c (tree_predictive_commoning_loop): Call
1649         free_affine_expand_cache if giving up because components is NULL.
1650
1651         * target-globals.c (save_target_globals): Allocate < 4KB structs using
1652         GC in payload of target_globals struct instead of allocating them on
1653         the heap and the larger structs separately using GC.
1654         * target-globals.h (struct target_globals): Make regs, hard_regs,
1655         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
1656         of GTY((skip)) and change type to void *.
1657         (reset_target_globals): Cast loads from those fields to corresponding
1658         types.
1659
1660 2014-01-10  Steve Ellcey  <sellcey@mips.com>
1661
1662         PR plugins/59335
1663         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
1664         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
1665         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
1666
1667 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
1668
1669         PR target/59744
1670         * aarch64-modes.def (CC_Zmode): New flags mode.
1671         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
1672         represents an equality.
1673         (aarch64_get_condition_code): Handle CC_Zmode.
1674         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
1675
1676 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
1677
1678         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
1679         extraction in good case.
1680
1681 2014-01-10  Richard Biener  <rguenther@suse.de>
1682
1683         PR tree-optimization/59374
1684         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
1685         checking after SLP discovery.  Mark stmts not participating
1686         in any SLP instance properly.
1687
1688 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1689
1690         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
1691         when handling a SET rtx.
1692
1693 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1694
1695         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
1696         (cortex-a57): Likewise.
1697         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
1698
1699 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1700
1701         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
1702         non-iwmmxt builtins.
1703
1704 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
1705
1706         PR ipa/58252
1707         PR ipa/59226
1708         * ipa-devirt.c record_target_from_binfo): Take as argument
1709         stack of binfos and lookup matching one for virtual inheritance.
1710         (possible_polymorphic_call_targets_1): Update.
1711
1712 2014-01-10  Huacai Chen  <chenhc@lemote.com>
1713
1714         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
1715         kernel strings for Loongson-2E/2F/3A.
1716
1717 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
1718
1719         PR middle-end/59670
1720         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
1721         is_gimple_call before calling gimple_call_internal_p.
1722
1723 2014-01-09  Steve Ellcey  <sellcey@mips.com>
1724
1725         * Makefile.in (TREE_FLOW_H): Remove.
1726         (TREE_SSA_H): Add file names from tree-flow.h.
1727         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
1728         * tree.h: Remove tree-flow.h reference.
1729         * hash-table.h: Remove tree-flow.h reference.
1730         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
1731         reference with tree-ssa-loop.h.
1732
1733 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1734
1735         * doc/invoke.texi: Add -maltivec={be,le} options, and document
1736         default element-order behavior for -maltivec.
1737         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
1738         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
1739         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
1740         when targeting big endian, at least for now.
1741         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
1742
1743 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
1744
1745         PR middle-end/47735
1746         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
1747         var satisfies use_register_for_decl, just take into account type
1748         alignment, rather than decl alignment.
1749
1750         PR tree-optimization/59622
1751         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
1752         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
1753         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
1754         Don't devirtualize for inplace at all.  For targets.length () == 1,
1755         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
1756
1757 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
1758
1759         * config/i386/i386.md (cpu): Remove the unused btver1.
1760
1761 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
1762
1763         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
1764
1765 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
1766
1767         PR target/58115
1768         * tree-core.h (struct target_globals): New forward declaration.
1769         (struct tree_target_option): Add globals field.
1770         * tree.h (TREE_TARGET_GLOBALS): Define.
1771         (prepare_target_option_nodes_for_pch): New prototype.
1772         * target-globals.h (struct target_globals): Define even if
1773         !SWITCHABLE_TARGET.
1774         * tree.c (prepare_target_option_node_for_pch,
1775         prepare_target_option_nodes_for_pch): New functions.
1776         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
1777         * config/i386/i386.c: Include target-globals.h.
1778         (ix86_set_current_function): Instead of doing target_reinit
1779         unconditionally, use save_target_globals_default_opts and
1780         restore_target_globals.
1781
1782 2014-01-09  Richard Biener  <rguenther@suse.de>
1783
1784         PR tree-optimization/59715
1785         * tree-cfg.h (split_critical_edges): Declare.
1786         * tree-cfg.c (split_critical_edges): Export.
1787         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
1788
1789 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
1790
1791         * cfgexpand.c (expand_stack_vars): Optionally disable
1792         asan stack protection.
1793         (expand_used_vars): Likewise.
1794         (partition_stack_vars): Likewise.
1795         * asan.c (asan_emit_stack_protection): Optionally disable
1796         after return stack usage.
1797         (instrument_derefs): Optionally disable memory access instrumentation.
1798         (instrument_builtin_call): Likewise.
1799         (instrument_strlen_call): Likewise.
1800         (asan_protect_global): Optionally disable global variables protection.
1801         * doc/invoke.texi: Added doc for new options.
1802         * params.def: Added new options.
1803         * params.h: Likewise.
1804
1805 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
1806
1807         PR rtl-optimization/59724
1808         * ifcvt.c (cond_exec_process_if_block): Don't call
1809         flow_find_head_matching_sequence with 0 longest_match.
1810         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
1811         non-active insns if !stop_after.
1812         (try_head_merge_bb): Revert 2014-01-07 changes.
1813
1814 2014-01-08  Jeff Law  <law@redhat.com>
1815
1816         * ree.c (get_sub_rtx): New function, extracted from...
1817         (merge_def_and_ext): Here.
1818         (combine_reaching_defs): Use get_sub_rtx.
1819
1820 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
1821
1822         * cgraph.h (varpool_variable_node): Do not choke on null node.
1823
1824 2014-01-08  Catherine Moore  <clm@codesourcery.com>
1825
1826         * config/mips/mips.md (simple_return): Attempt to use JRC
1827         for microMIPS.
1828         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
1829
1830 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
1831
1832         PR rtl-optimization/59137
1833         * reorg.c (steal_delay_list_from_target): Call update_block for
1834         elided insns.
1835         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
1836
1837 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1838
1839         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1840         two duplicate entries.
1841
1842 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
1843
1844         Revert:
1845         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
1846
1847         * config/mips/mips.c (mips_truncated_op_cost): New function.
1848         (mips_rtx_costs): Adjust test for BADDU.
1849         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
1850
1851         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
1852
1853         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
1854         (*baddu_si): ...this new pattern.
1855
1856 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
1857
1858         PR ipa/59722
1859         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
1860
1861 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1862
1863         PR middle-end/57748
1864         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
1865         inner_reference_p.
1866         (expand_expr, expand_normal): Adjust.
1867         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
1868         inner_reference_p. Use inner_reference_p to expand inner references.
1869         (store_expr): Adjust.
1870         * cfgexpand.c (expand_call_stmt): Adjust.
1871
1872 2014-01-08  Rong Xu  <xur@google.com>
1873
1874         * gcov-io.c (gcov_var): Move from gcov-io.h.
1875         (gcov_position): Ditto.
1876         (gcov_is_error): Ditto.
1877         (gcov_rewrite): Ditto.
1878         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
1879         only part to libgcc/libgcov.h.
1880
1881 2014-01-08  Marek Polacek  <polacek@redhat.com>
1882
1883         PR middle-end/59669
1884         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
1885
1886 2014-01-08  Marek Polacek  <polacek@redhat.com>
1887
1888         PR sanitizer/59667
1889         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
1890
1891 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
1892
1893         PR rtl-optimization/59649
1894         * stor-layout.c (get_mode_bounds): For BImode return
1895         0 and STORE_FLAG_VALUE.
1896
1897 2014-01-08  Richard Biener  <rguenther@suse.de>
1898
1899         PR middle-end/59630
1900         * gimple.h (is_gimple_builtin_call): Remove.
1901         (gimple_builtin_call_types_compatible_p): New.
1902         (gimple_call_builtin_p): New overload.
1903         * gimple.c (is_gimple_builtin_call): Remove.
1904         (validate_call): Rename to ...
1905         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
1906         check return types.
1907         (validate_type): New static function.
1908         (gimple_call_builtin_p): New overload and adjust.
1909         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
1910         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
1911         (gimple_fold_stmt_to_constant_1): Likewise.
1912         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
1913
1914 2014-01-08  Richard Biener  <rguenther@suse.de>
1915
1916         PR middle-end/59471
1917         * gimplify.c (gimplify_expr): Gimplify register-register type
1918         VIEW_CONVERT_EXPRs to separate stmts.
1919
1920 2014-01-07  Jeff Law  <law@redhat.com>
1921
1922         PR middle-end/53623
1923         * ree.c (combine_set_extension): Handle case where source
1924         and destination registers in an extension insn are different.
1925         (combine_reaching_defs): Allow source and destination registers
1926         in extension to be different under limited circumstances.
1927         (add_removable_extension): Remove restriction that the
1928         source and destination registers in the extension are the same.
1929         (find_and_remove_re): Emit a copy from the extension's
1930         destination to its source after the defining insn if
1931         the source and destination registers are different.
1932
1933         PR middle-end/59285
1934         * ifcvt.c (merge_if_block): If we are merging a block with more than
1935         one successor with a block with no successors, remove any BARRIER
1936         after the second block.
1937
1938 2014-01-07  Dan Xio Qiang <ziyan01@163.com>
1939
1940         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
1941
1942 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
1943
1944         PR target/59652
1945         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
1946         for 14-bit register offsets when INT14_OK_STRICT is false.
1947
1948 2014-01-07  Roland Stigge  <stigge@antcom.de>
1949             Michael Meissner  <meissner@linux.vnet.ibm.com>
1950
1951         PR 57386/target
1952         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
1953         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
1954
1955 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
1956
1957         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
1958         -mcpu.
1959
1960 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
1961
1962         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
1963         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
1964         rtx is const0_rtx or not.
1965
1966 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
1967
1968         PR target/58115
1969         * target-globals.c (save_target_globals): Remove this_fn_optab
1970         handling.
1971         * toplev.c: Include optabs.h.
1972         (target_reinit): Temporarily restore the global options if another
1973         set of options are in force.
1974
1975 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
1976
1977         PR rtl-optimization/58668
1978         * cfgcleanup.c (flow_find_cross_jump): Don't count
1979         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
1980         to determine what is counted.
1981         (flow_find_head_matching_sequence): Use active_insn_p to determine
1982         what is counted.
1983         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
1984         counting change.
1985         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
1986         determine what is counted.
1987
1988         PR tree-optimization/59643
1989         * tree-predcom.c (split_data_refs_to_components): If one dr is
1990         read and one write, determine_offset fails and the write isn't
1991         in the bad component, just put the read into the bad component.
1992
1993 2014-01-07  Mike Stump  <mikestump@comcast.net>
1994             Jakub Jelinek  <jakub@redhat.com>
1995
1996         PR pch/59436
1997         * tree-core.h (struct tree_optimization_option): Change optabs
1998         type from unsigned char * to void *.
1999         * optabs.c (init_tree_optimization_optabs): Adjust
2000         TREE_OPTIMIZATION_OPTABS initialization.
2001
2002 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
2003
2004         PR target/59644
2005         * config/i386/i386.h (struct machine_function): Add
2006         no_drap_save_restore field.
2007         * config/i386/i386.c (ix86_save_reg): Use
2008         !cfun->machine->no_drap_save_restore instead of
2009         crtl->stack_realign_needed.
2010         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
2011         this function clears frame_pointer_needed.  Set
2012         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
2013         and DRAP reg is needed.
2014
2015 2014-01-06  Marek Polacek  <polacek@redhat.com>
2016
2017         PR c/57773
2018         * doc/implement-c.texi: Mention that other integer types are
2019         permitted as bit-field types in strictly conforming mode.
2020
2021 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
2022
2023         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
2024         is newly allocated.
2025
2026 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
2027
2028         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
2029
2030 2014-01-06  Martin Jambor  <mjambor@suse.cz>
2031
2032         PR ipa/59008
2033         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
2034         to int.
2035         * ipa-prop.c (ipa_print_node_params): Fix indentation.
2036
2037 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
2038
2039         PR debug/59350
2040         PR debug/59510
2041         * var-tracking.c (add_stores): Preserve the value of the source even if
2042         we don't record the store.
2043
2044 2014-01-06  Terry Guo  <terry.guo@arm.com>
2045
2046         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
2047
2048 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
2049
2050         PR bootstrap/59541
2051         * config/darwin.c (darwin_function_section): Adjust return values to
2052         correspond to optimisation changes made in r206070.
2053
2054 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
2055
2056         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
2057         from prefetch_block tune setting.
2058         (nocona_cost): Correct size of prefetch block to 64.
2059
2060 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
2061
2062         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
2063         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
2064         used to save the static chain register in the computation of the offset
2065         from which the FP registers need to be restored.
2066
2067 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
2068
2069         PR tree-optimization/59519
2070         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
2071         ICE if get_current_def (current_new_name) is already non-NULL, as long
2072         as it is a phi result of some other phi in *new_exit_bb that has
2073         the same argument.
2074
2075         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
2076         or vmovdqu* for misaligned_operand.
2077         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
2078         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
2079         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
2080         aligned_mem for AVX512F masked aligned load and store builtins and for
2081         non-temporal moves.
2082
2083 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
2084
2085         PR tree-optimization/59651
2086         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
2087         Address range for negative step should be added by TYPE_SIZE_UNIT.
2088
2089 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
2090
2091         * config/m68k/m68k.c (handle_move_double): Handle pushes with
2092         overlapping registers also for registers other than the stack pointer.
2093
2094 2014-01-03  Marek Polacek  <polacek@redhat.com>
2095
2096         PR other/59661
2097         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
2098         __builtin_FILE.
2099
2100 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
2101
2102         PR target/59625
2103         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
2104         asm goto as jump.
2105
2106         * config/i386/i386.md (MODE_SIZE): New mode attribute.
2107         (push splitter): Use <P:MODE_SIZE> instead of
2108         GET_MODE_SIZE (<P:MODE>mode).
2109         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
2110         (mov -1, reg peephole2): Likewise.
2111         * config/i386/sse.md (*mov<mode>_internal,
2112         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
2113         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
2114         *<code><mode>3, *andnot<mode>3<mask_name>,
2115         <mask_codefor><code><mode>3<mask_name>): Likewise.
2116         * config/i386/subst.md (mask_mode512bit_condition,
2117         sd_mask_mode512bit_condition): Likewise.
2118
2119 2014-01-02  Xinliang David Li  <davidxl@google.com>
2120
2121         PR tree-optimization/59303
2122         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
2123         (dump_predicates): Better output format.
2124         (pred_equal_p): New function.
2125         (is_neq_relop_p): Ditto.
2126         (is_neq_zero_form_p): Ditto.
2127         (pred_expr_equal_p): Ditto.
2128         (pred_neg_p): Ditto.
2129         (simplify_pred): Ditto.
2130         (simplify_preds_2): Ditto.
2131         (simplify_preds_3): Ditto.
2132         (simplify_preds_4): Ditto.
2133         (simplify_preds): Ditto.
2134         (push_pred): Ditto.
2135         (push_to_worklist): Ditto.
2136         (get_pred_info_from_cmp): Ditto.
2137         (is_degenerated_phi): Ditto.
2138         (normalize_one_pred_1): Ditto.
2139         (normalize_one_pred): Ditto.
2140         (normalize_one_pred_chain): Ditto.
2141         (normalize_preds): Ditto.
2142         (normalize_cond_1): Remove function.
2143         (normalize_cond): Ditto.
2144         (is_gcond_subset_of): Ditto.
2145         (is_subset_of_any): Ditto.
2146         (is_or_set_subset_of): Ditto.
2147         (is_and_set_subset_of): Ditto.
2148         (is_norm_cond_subset_of): Ditto.
2149         (pred_chain_length_cmp): Ditto.
2150         (convert_control_dep_chain_into_preds): Type change.
2151         (find_predicates): Ditto.
2152         (find_def_preds): Ditto.
2153         (destroy_predicates_vecs): Ditto.
2154         (find_matching_predicates_in_rest_chains): Ditto.
2155         (use_pred_not_overlap_with_undef_path_pred): Ditto.
2156         (is_pred_expr_subset): Ditto.
2157         (is_pred_chain_subset_of): Ditto.
2158         (is_included_in): Ditto.
2159         (is_superset_of): Ditto.
2160
2161 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
2162
2163         Update copyright years.
2164
2165 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
2166
2167         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
2168         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
2169         config/arc/arc.md, config/arc/arc.opt,
2170         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
2171         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
2172         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
2173         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
2174         config/linux-protos.h, config/linux.c, config/winnt-c.c,
2175         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
2176         vtable-verify.c, vtable-verify.h: Use the standard form for the
2177         copyright notice.
2178
2179 2014-01-02  Tobias Burnus  <burnus@net-b.de>
2180
2181         * gcc.c (process_command): Update copyright notice dates.
2182         * gcov-dump.c: Ditto.
2183         * gcov.c: Ditto.
2184         * doc/cpp.texi: Bump @copying's copyright year.
2185         * doc/cppinternals.texi: Ditto.
2186         * doc/gcc.texi: Ditto.
2187         * doc/gccint.texi: Ditto.
2188         * doc/gcov.texi: Ditto.
2189         * doc/install.texi: Ditto.
2190         * doc/invoke.texi: Ditto.
2191
2192 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
2193
2194         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
2195
2196 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
2197
2198         * config/i386/sse.md (*mov<mode>_internal): Guard
2199         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
2200
2201         PR rtl-optimization/59647
2202         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
2203         new_rtx into UNSIGNED_FLOAT rtxes.
2204 \f
2205 Copyright (C) 2014 Free Software Foundation, Inc.
2206
2207 Copying and distribution of this file, with or without modification,
2208 are permitted in any medium without royalty provided the copyright
2209 notice and this notice are preserved.