gcc/ChangeLog

   1 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
   2
   3         PR target/85657
   4         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
   5         define __ibm128 as long double.
   6         * config/rs6000/rs6000.c (rs6000_init_builtins): Always create
   7         __ibm128 as a distinct type.
   8         (init_float128_ieee): Fix up conversions between IFmode and IEEE
   9         128-bit types to use the correct functions.
  10         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
  11         convert between 128-bit floating point types that have different
  12         modes but the same representation, instead of using gen_lowpart to
  13         makean alias.
  14         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
  15         KFmode.
  16         (IFKF_reg): New attributes to give the register constraints for
  17         IFmode and KFmode.
  18         (extend<mode>tf2_internal): New insns to mark an explicit
  19         conversion between 128-bit floating point types that have a
  20         different mode but share the same representation.
  21
  22 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
  23
  24         PR tree-optimization/85814
  25         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
  26         a null return from get_strinfo when unsharing the next
  27         strinfo in the chain.
  28
  29 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
  30
  31         PR gcc/84923
  32         * varasm.c (weak_finish): Clean up weak_decls.
  33
  34 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
  35
  36         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
  37         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
  38         UNSPEC_UADALP values.
  39         * config/aarch64/iterators.md (ABAL): New int iterator.
  40         (ABDL2): Likewise.
  41         (ADALP): Likewise.
  42         (sur): Add mappings for the above.
  43         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
  44         New define_insn.
  45         (aarch64_<sur>abal<mode>_4): Likewise.
  46         (aarch64_<sur>adalp<mode>_3): Likewise.
  47         (<sur>sadv16qi): New define_expand.
  48
  49 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
  50
  51         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
  52         (*movdf_internal): Ditto.
  53         (*rcpsf2_sse): Ditto.
  54         (*rsqrtsf2_sse): Ditto.
  55         (*sqrt<mode>2_sse): Ditto.
  56
  57 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
  58
  59         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
  60         eor3q<mode>4.
  61         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
  62         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
  63         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
  64         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
  65         vbcaxq_s64): New.
  66         * config/aarch64/arm_neon.h: Likewise.
  67         * config/aarch64/iterators.md (VQ_I): New.
  68
  69 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
  70
  71         * config.gcc: Add arc/t-multilib-linux to tmake_file for
  72         arc*-*-linux*.
  73         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
  74         MULTILIB_DIRNAMES
  75
  76 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
  77
  78         * config/nds32/constraints.md (S): New constraint.
  79         * config/nds32/nds32.md (call_internal): Use constraint S.
  80         (call_value_internal): Likewise.
  81         (sibcall_internal): Likewise.
  82         (sibcall_value_internal): Likewise.
  83
  84 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
  85             Chung-Ju Wu  <jasonwucj@gmail.com>
  86
  87         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
  88         into consideration.
  89
  90 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
  91             Chung-Ju Wu  <jasonwucj@gmail.com>
  92
  93         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
  94         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
  95         (nds32_rtx_costs_impl): Simplify.
  96         (nds32_address_cost_impl): Simplify.
  97         (nds32_init_rtx_costs): New function.
  98         (nds32_rtx_costs_speed_prefer): Likewise.
  99         (nds32_rtx_costs_size_prefer): Likewise.
 100         (nds32_address_cost_speed_prefer): Likewise.
 101         (nds32_address_cost_speed_fwprop): Likewise.
 102         (nds32_address_cost_size_prefer): Likewise.
 103         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
 104         * config/nds32/nds32.c (nds32_option_override): Use
 105         nds32_init_rtx_costs function.
 106
 107 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
 108
 109         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
 110         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
 111         (TARGET_PIPELINE_N8): Likewise.
 112         (TARGET_PIPELINE_N10): Likewise.
 113         (TARGET_PIPELINE_N13): Likewise.
 114         (TARGET_PIPELINE_GRAYWOLF): Likewise.
 115
 116 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
 117
 118         * config/nds32/nds32-fpu.md: Update copyright year.
 119
 120 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
 121
 122         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
 123
 124 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
 125
 126         * config/nds32/nds32.c
 127         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
 128         * config/nds32/nds32.opt (minline-asm-r15): New option.
 129
 130 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
 131
 132         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
 133         MASK_HW_ABS.
 134         * config/nds32/nds32.md (abssi2): New pattern.
 135
 136 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
 137
 138         * config/i386/i386.md (rex64namesuffix): New mode attribute.
 139         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
 140         Merge insn pattern from sse_cvtsi2ss<round_name> and
 141         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
 142         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
 143         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
 144         using SWI48 mode iterator.
 145         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
 146         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
 147         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
 148         pattern from sse_cvttss2si<round_saeonly_name>
 149         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
 150         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
 151         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
 152         using SWI48 mode iterator.
 153         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
 154         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
 155         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
 156         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
 157         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
 158         using SWI48 mode iterator.
 159         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
 160         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
 161         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
 162         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
 163         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
 164         SWI48 mode iterator.
 165         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
 166         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
 167         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
 168         pattern from sse_cvttsd2si<round_saeonly_name>
 169         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
 170
 171 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
 172
 173         * config/nds32/nds32-md-auxiliary.c
 174         (nds32_valid_smw_lwm_base_p): Refine.
 175         (nds32_output_smw_single_word): Refine.
 176         (nds32_output_smw_double_word): New.
 177         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
 178
 179 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
 180
 181         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
 182         (nds32_output_stack_pop): Refine.
 183         (nds32_expand_unaligned_load): Refine.
 184         (nds32_expand_unaligned_store): Refine.
 185
 186 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
 187             Chung-Ju Wu  <jasonwucj@gmail.com>
 188
 189         * config/nds32/constants.md: Add TP_REGNUM constant.
 190         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
 191         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
 192         UNSPEC_ADD32.
 193         * config/nds32/nds32-doubleword.md: Consider flag_pic.
 194         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
 195         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
 196         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
 197         and PIC code generation.
 198         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
 199         code generation.
 200         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
 201         optimization.
 202         * config/nds32/nds32.md: Support TLS and PIC.
 203         * config/nds32/nds32.c: Support TLS and PIC.
 204         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
 205         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
 206         predicate.
 207
 208 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
 209
 210         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
 211         mode with E_ prefix.
 212
 213 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
 214             Chung-Ju Wu  <jasonwucj@gmail.com>
 215
 216         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
 217         * config/nds32/nds32-md-auxiliary.c
 218         (symbolic_reference_mentioned_p): New.
 219         (nds32_legitimize_ict_address): New.
 220         (nds32_expand_ict_move): New.
 221         (nds32_indirect_call_referenced_p): New.
 222         (nds32_symbol_binds_local_p): Delete.
 223         (nds32_long_call_p): Modify.
 224         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
 225         * config/nds32/nds32-protos.h
 226         (symbolic_reference_mentioned_p): Declare.
 227         (nds32_legitimize_ict_address): Declare.
 228         (nds32_expand_ict_move): Declare.
 229         (nds32_indirect_call_referenced_p): Declare.
 230         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
 231         (nds32_relax_group): Use nds32_ict_const_p as condition.
 232         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
 233         (nds32_asm_file_start): Output ict_model directive in asm code.
 234         (nds32_legitimate_address_p): Consider indirect call.
 235         (nds32_print_operand): Consider indirect call.
 236         (nds32_print_operand_address): Consider indirect call.
 237         (nds32_insert_attributes): Handle "indirect_call" attribute.
 238         (TARGET_LEGITIMATE_ADDRESS_P): Define.
 239         (TARGET_LEGITIMATE_CONSTANT_P): Define.
 240         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
 241         (TARGET_DELEGITIMIZE_ADDRESS): Define.
 242         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
 243         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
 244         (TARGET_ICT_MODEL_SMALL): Define.
 245         (TARGET_ICT_MODEL_LARGE): Define.
 246         * config/nds32/nds32.md (movsi): Consider ict model.
 247         (call, call_value): Consider ict model.
 248         (sibcall, sibcall_value): Consider ict model.
 249         * config/nds32/nds32.opt (mict-model): New option.
 250         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
 251         model.
 252
 253 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
 254             Monk Chiang  <sh.chiang04@gmail.com>
 255             Jim Wilson <jimw@sifive.com>
 256
 257         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
 258         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
 259         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
 260         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
 261         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
 262         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
 263         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
 264         compute save_libcall_adjustment properly.
 265         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
 266         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
 267         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
 268         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
 269         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
 270         (ABI_SPEC): Handle mabi=ilp32e.
 271         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
 272         (RVE): Add RVE mask.
 273         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
 274         <-march>: Add rv32e as an example.
 275
 276 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
 277
 278         PR c++/82899
 279         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
 280         (intra_create_variable_infos): Handle C++ constructors.
 281
 282 2018-05-18  Martin Liska  <mliska@suse.cz>
 283
 284         * passes.def: Remove a redundant pass.
 285
 286 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
 287
 288         PR bootstrap/85838
 289         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
 290
 291 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 292
 293         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
 294         (ARMv4): Update.
 295         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
 296         (ARMv6m): Update.
 297         (armv2, armv2a, armv3, armv3m): Delete architectures.
 298         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
 299         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
 300         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
 301         Delete cpus.
 302         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
 303         (*mulsidi3adddi): Likewise.
 304         (mulsidi3): Likewise.
 305         (*mulsidi3_nov6): Likewise.
 306         (umulsidi3): Likewise.
 307         (umulsidi3_nov6): Likewise.
 308         (umaddsidi4): Likewise.
 309         (*umulsidi3adddi): Likewise.
 310         (smulsi3_highpart): Likewise.
 311         (*smulsi3_highpart_nov6): Likewise.
 312         (umulsi3_highpart): Likewise.
 313         (*umulsi3_highpart_nov6): Likewise.
 314         * config/arm/arm.h (arm_arch3m): Delete.
 315         * config/arm/arm.c (arm_arch3m): Delete.
 316         (arm_option_override_internal): Update armv3-related comment.
 317         (arm_configure_build_target): Delete use of isa_bit_mode32.
 318         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
 319         (arm_rtx_costs_internal): Delete check of arm_arch3m.
 320         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
 321         (mulsa3): Likewise.
 322         (mulusa3): Likewise.
 323         * config/arm/arm-protos.h (arm_arch3m): Delete.
 324         * config/arm/arm-tables.opt: Regenerate.
 325         * config/arm/arm-tune.md: Likewise.
 326         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
 327         deleted architectures.
 328
 329 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 330
 331         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
 332         (armv5t, armv5te): New features.
 333         (ARMv5, ARMv5e): Delete fgroups.
 334         (ARMv5t, ARMv5te): Adjust for above changes.
 335         (ARMv6m): Likewise.
 336         (armv5, armv5e): Delete arches.
 337         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
 338         arm_arch5.
 339         (*call_reg_arm): Likewise.
 340         (*call_value_reg_armv5): Likewise.
 341         (*call_value_reg_arm): Likewise.
 342         (*call_symbol): Likewise.
 343         (*call_value_symbol): Likewise.
 344         (*sibcall_insn): Likewise.
 345         (*sibcall_value_insn): Likewise.
 346         (clzsi2): Likewise.
 347         (prefetch): Likewise.
 348         (define_split and define_peephole2 dependent on arm_arch5):
 349         Likewise.
 350         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
 351         arm_arch5e.
 352         (TARGET_ARM_QBIT): Likewise.
 353         (TARGET_DSP_MULTIPLY): Likewise.
 354         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
 355         (arm_arch5, arm_arch5e): Delete.
 356         (arm_arch5t, arm_arch5te): Declare.
 357         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
 358         (arm_arch5t): Declare.
 359         (arm_option_reconfigure_globals): Update for the above.
 360         (arm_options_perform_arch_sanity_checks): Update comment, replace
 361         use of arm_arch5 with arm_arch5t.
 362         (use_return_insn): Likewise.
 363         (arm_emit_call_insn): Likewise.
 364         (output_return_instruction): Likewise.
 365         (arm_final_prescan_insn): Likewise.
 366         (arm_coproc_builtin_available): Likewise.
 367         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
 368         arm_arch5e with arm_arch5t and arm_arch5te.
 369         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
 370         (arm_arch5t, arm_arch5te): Declare.
 371         * config/arm/arm-tables.opt: Regenerate.
 372         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
 373         * config/arm/t-multilib: Likewise.
 374         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
 375         instead of arm_arch5.
 376         (*call_reg_thumb1): Likewise.
 377         (*call_value_reg_thumb1_v5): Likewise.
 378         (*call_value_reg_thumb1): Likewise.
 379         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
 380         unreachable path.
 381         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
 382
 383 2018-05-18  Martin Liska  <mliska@suse.cz>
 384
 385         PR gcov-profile/84846
 386         * doc/gcov.texi: Document -t option of gcov tool.
 387
 388 2018-05-18  Martin Liska  <mliska@suse.cz>
 389
 390         PR gcov-profile/84846
 391         * gcov.c (print_usage): Add new -t option.
 392         (process_args): Handle the option.
 393         (generate_results): Use stdout as output when requested by
 394         the option.
 395
 396 2018-05-18  Martin Liska  <mliska@suse.cz>
 397
 398         PR gcov-profile/84846
 399         * coverage.c (coverage_init): Write PWD to .gcno file.
 400         * doc/gcov.texi: Document how working directory is printed.
 401         * gcov-dump.c (dump_gcov_file): Print PWD.
 402         * gcov.c (output_intermediate_file): Likewise.
 403         (read_graph_file): Read PWD string.
 404         (output_lines): Print PWD.
 405
 406 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
 407
 408         PR middle-end/85817
 409         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
 410         for retval and return false if all args to phi are zero.
 411
 412 2018-05-18  Richard Biener  <rguenther@suse.de>
 413
 414         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
 415         method.
 416         (evrp_dom_walker::before_dom_children): Call it.
 417
 418 2018-05-18  Richard Biener  <rguenther@suse.de>
 419
 420         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
 421         results when processing array refs with variable index.
 422
 423 2018-05-18  Toon Moene  <toon@moene.org>
 424
 425         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
 426         directly after that of -floop-interchange. Indicate that both
 427         options are enabled by default when specifying -O3.
 428
 429 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 430
 431         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
 432         iterator.  Delete separate integer-mode vec_set<mode> expander.
 433         (aarch64_simd_vec_setv2di): Delete.
 434         (vec_setv2di): Delete.
 435         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
 436         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
 437         the "w, r" alternative.
 438
 439 2018-05-18  Martin Liska  <mliska@suse.cz>
 440
 441         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
 442         * tree-pass.h (make_pass_lower_switch_O0): New function.
 443         * tree-switch-conversion.c (node_has_low_bound): Remove.
 444         (node_has_high_bound): Likewise.
 445         (node_is_bounded): Likewise.
 446         (class pass_lower_switch): Make it a template type and create
 447         two instances.
 448         (pass_lower_switch::execute): Add template argument.
 449         (make_pass_lower_switch): New function.
 450         (make_pass_lower_switch_O0): New function.
 451         (do_jump_if_equal): Remove.
 452         (emit_case_nodes): Simplify to just handle all 3 cases and leave
 453         all the hard work to tree optimization passes.
 454
 455 2018-05-18  Martin Liska  <mliska@suse.cz>
 456
 457         * dbgcnt.c (limit_low): Renamed from limit.
 458         (limit_high): New variable.
 459         (dbg_cnt_is_enabled): Check for upper limit.
 460         (dbg_cnt): Adjust dumping.
 461         (dbg_cnt_set_limit_by_index): Add new argument for high
 462         value.
 463         (dbg_cnt_set_limit_by_name): Likewise.
 464         (dbg_cnt_process_single_pair): Parse new format.
 465         (dbg_cnt_process_opt): Use strtok.
 466         (dbg_cnt_list_all_counters): Remove 'value' and add
 467         'limit_high'.
 468         * doc/invoke.texi: Document changes.
 469
 470 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
 471
 472         * doc/sourcebuild.texi (scalar_all_fma): Document.
 473         * tree.def (FMA_EXPR): Delete.
 474         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
 475         * internal-fn.c (ternary_direct): New macro.
 476         (expand_ternary_optab_fn): Likewise.
 477         (direct_ternary_optab_supported_p): Likewise.
 478         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
 479         * builtins.c (fold_builtin_fma): Delete.
 480         (fold_builtin_3): Don't call it.
 481         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
 482         * expr.c (expand_expr_real_2): Likewise.
 483         * fold-const.c (operand_equal_p): Likewise.
 484         (fold_ternary_loc): Likewise.
 485         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
 486         * gimple.c (DEFTREECODE): Likewise.
 487         * gimplify.c (gimplify_expr): Likewise.
 488         * optabs-tree.c (optab_for_tree_code): Likewise.
 489         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
 490         * tree-eh.c (operation_could_trap_p): Likewise.
 491         (stmt_could_throw_1_p): Likewise.
 492         * tree-inline.c (estimate_operator_cost): Likewise.
 493         * tree-pretty-print.c (dump_generic_node): Likewise.
 494         (op_code_prio): Likewise.
 495         * tree-ssa-loop-im.c (stmt_cost): Likewise.
 496         * tree-ssa-operands.c (get_expr_operands): Likewise.
 497         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
 498         * fold-const-call.h (fold_fma): Delete.
 499         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
 500         CFN_FNMA and CFN_FNMS.
 501         (fold_fma): Delete.
 502         * genmatch.c (combined_fn): New enum.
 503         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
 504         (commutative_op): New function.
 505         (commutate): Use it.  Handle more than 2 operands.
 506         (dt_operand::gen_gimple_expr): Use commutative_op.
 507         (parser::parse_expr): Allow :c to be used with non-binary
 508         operators if the commutative operand is known.
 509         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
 510         CFN_FMS, CFN_FNMA and CFN_FNMS.
 511         (backprop::process_assign_use): Remove FMA_EXPR handling.
 512         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
 513         (gen_hsa_fma): New function.
 514         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
 515         IFN_FNMA and IFN_FNMS.
 516         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
 517         * gimple-fold.h (follow_all_ssa_edges): Declare.
 518         * gimple-fold.c (follow_all_ssa_edges): New function.
 519         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
 520         gimple_build interface and use follow_all_ssa_edges to fold the result.
 521         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
 522         instead of checking for optabs directly.
 523         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
 524         rather than FMA_EXPRs.
 525         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
 526         call to IFN_FMA instead of an FMA_EXPR.
 527
 528 2018-05-17  Jim Wilson  <jimw@sifive.com>
 529
 530         * expr.c (do_tablejump): When converting index to Pmode, if we have a
 531         sign extended promoted subreg, and the range does not have the sign bit
 532         set, then do a sign extend.
 533
 534         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
 535         test, check for sign extended subreg and/or constant operands, and
 536         do a sign extend in that case.
 537
 538 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
 539
 540         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
 541         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
 542         Add untyped.
 543         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
 544         Change logics_shift_reg to logics_shift_imm.
 545         (thunderx2t99_fp_loadpair_basic): Delete.
 546         (thunderx2t99_fp_storepair_basic): Delete.
 547         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
 548         (thunderx2t99_asimd_polynomial): Delete.
 549         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
 550         and neon_fp_mul_d_scalar_q.
 551         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
 552         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
 553         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
 554         (thunderx2t99_asimd_lut): Add missing tbl types.
 555         (thunderx2t99_asimd_ext): Delete.
 556         (thunderx2t99_asimd_load1_1_mult): Delete.
 557         (thunderx2t99_asimd_load1_2_mult): Delete.
 558         (thunderx2t99_asimd_load1_ldp): New.
 559         (thunderx2t99_asimd_load1): New.
 560         (thunderx2t99_asimd_load2): Add missing *load2* types.
 561         (thunderx2t99_asimd_load3): New.
 562         (thunderx2t99_asimd_load4): New.
 563         (thunderx2t99_asimd_store1_1_mult): Delete.
 564         (thunderx2t99_asimd_store1_2_mult): Delete.
 565         (thunderx2t99_asimd_store2_mult): Delete.
 566         (thunderx2t99_asimd_store2_onelane): Delete.
 567         (thunderx2t99_asimd_store_stp): New.
 568         (thunderx2t99_asimd_store1): New.
 569         (thunderx2t99_asimd_store2): New.
 570         (thunderx2t99_asimd_store3): New.
 571         (thunderx2t99_asimd_store4): New.
 572
 573 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
 574
 575         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
 576         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
 577
 578 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
 579             Segher Boessenkool  <segher@kernel.crashing.org>
 580
 581         PR target/85698
 582         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
 583         operand.
 584
 585 2018-05-17  Richard Biener  <rguenther@suse.de>
 586
 587         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
 588         for pruning loop and prune defs feeding only already visited PHIs.
 589
 590 2018-05-17  Richard Biener  <rguenther@suse.de>
 591
 592         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
 593
 594 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
 595             Richard Biener  <rguenther@suse.de>
 596
 597         PR tree-optimization/85793
 598         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
 599         for VMAT_ELEMENTWISE.
 600
 601 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
 602
 603         * internal-fn.h (lookup_internal_fn): Declare
 604         * internal-fn.c (lookup_internal_fn): New function.
 605         * gimple.c (gimple_build_call_from_tree): Handle calls to
 606         internal functions.
 607         * gimple-pretty-print.c (dump_gimple_call): Print "." before
 608         internal function names.
 609         * tree-pretty-print.c (dump_generic_node): Likewise.
 610         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
 611
 612 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
 613
 614         * gimple-fold.h (gimple_build): Make the function forms take
 615         combined_fn rather than built_in_function.
 616         (gimple_simplify): Likewise.
 617         * gimple-match-head.c (gimple_simplify): Likewise.
 618         * gimple-fold.c (gimple_build): Likewise.
 619         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
 620         rather than gimple_build_call_internal.
 621         (get_initial_defs_for_reduction): Likewise.
 622         (vect_create_epilog_for_reduction): Likewise.
 623         (vectorizable_live_operation): Likewise.
 624
 625 2018-05-17  Martin Liska  <mliska@suse.cz>
 626
 627         * gimple-ssa-sprintf.c (format_directive): Do not use
 628         space in between 'G_' and '('.
 629
 630 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
 631
 632         PR target/85323
 633         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
 634         even if the mask is not all ones.
 635
 636         PR target/85323
 637         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
 638         vector.
 639         (ix86_gimple_fold_builtin): Likewise.
 640
 641         PR target/85323
 642         * config/i386/i386.c: Include tree-vector-builder.h.
 643         (ix86_vector_shift_count): New function.
 644         (ix86_fold_builtin): Fold shift builtins by scalar count.
 645         (ix86_gimple_fold_builtin): Likewise.
 646
 647         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
 648         _mm512_setzero): New intrinsics.
 649
 650 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
 651             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 652
 653         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
 654         code generation for cases where splatting a value is not useful.
 655         * simplify-rtx.c (simplify_ternary_operation): Simplify
 656         vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
 657         mode to a vec_concat.
 658
 659 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
 660
 661         * config.gcc: Support "goldmont-plus".
 662         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
 663         "goldmont-plus".
 664         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
 665         PROCESSOR_GOLDMONT_PLUS.
 666         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
 667         (processor_target_table): Add "goldmont-plus".
 668         (PTA_GOLDMONT_PLUS): Define.
 669         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
 670         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
 671         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
 672         (fold_builtin_cpu): Add "goldmont-plus".
 673         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
 674         (ix86_option_override_internal): Add "goldmont-plus".
 675         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
 676         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
 677         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
 678         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
 679
 680 2018-05-17  Richard Biener  <rguenther@suse.de>
 681
 682         PR tree-optimization/85757
 683         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
 684         remove defs that only feed that PHI from further processing.
 685
 686 2018-05-16  Jim Wilson  <jimw@sifive.com>
 687
 688         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
 689         asterisk to name.
 690         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
 691
 692 2018-05-16  Mark Wielaard  <mark@klomp.org>
 693
 694         * dwarf2out.c (count_index_strings): New function.
 695         (output_indirect_strings): Call count_index_strings and generate
 696         header for dwarf_version >= 5.
 697
 698 2018-05-16  Mark Wielaard  <mark@klomp.org>
 699
 700         * dwarf2out.c (dwarf_FORM): New function.
 701         (set_indirect_string): Use dwarf_FORM.
 702         (reset_indirect_string): Likewise.
 703         (size_of_die): Likewise.
 704         (value_format): Likewise.
 705         (output_die): Likewise.
 706         (add_skeleton_AT_string): Likewise.
 707         (output_macinfo_op): Likewise.
 708         (index_string): Likewise.
 709         (output_index_string_offset): Likewise.
 710         (output_index_string): Likewise.
 711         (count_index_strings): Likewise.
 712
 713 2018-05-16  Carl Love  <cel@us.ibm.com>
 714
 715         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
 716         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
 717
 718 2018-05-16  Martin Jambor  <mjambor@suse.cz>
 719
 720         * ipa-prop.c (ipa_free_all_edge_args): Remove.
 721         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
 722
 723 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
 724
 725         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
 726         (fnma<mode>4): Likewise.
 727         (fms<mode>4): Likewise.
 728         (fnms<mode>4): Likewise.
 729         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
 730         (aarch64_fnma<mode>4): Likewise.
 731         (aarch64_fms<mode>4): Likewise.
 732         (aarch64_fnms<mode>4): Likewise.
 733         (aarch64_fnmadd<mode>4): Likewise.
 734
 735 2018-05-16  Jason Merrill  <jason@redhat.com>
 736
 737         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
 738
 739 2018-05-16  Richard Biener  <rguenther@suse.de>
 740
 741         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
 742         (dump_stmt_cost): Declare.
 743         (add_stmt_cost): Dump cost we add.
 744         (add_stmt_costs): New function.
 745         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
 746         No longer exported.
 747         (vect_analyze_stmt): Adjust prototype.
 748         (vectorizable_condition): Likewise.
 749         (vectorizable_live_operation): Likewise.
 750         (vectorizable_reduction): Likewise.
 751         (vectorizable_induction): Likewise.
 752         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
 753         cost vector to pass to vectorizable_ and record afterwards.
 754         (vect_model_reduction_cost): Take cost vector argument and adjust.
 755         (vect_model_induction_cost): Likewise.
 756         (vectorizable_reduction): Likewise.
 757         (vectorizable_induction): Likewise.
 758         (vectorizable_live_operation): Likewise.
 759         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
 760         SLP_TREE_NUMBER_OF_VEC_STMTS.
 761         (vect_analyze_slp_cost_1): Remove.
 762         (vect_analyze_slp_cost): Likewise.
 763         (vect_slp_analyze_node_operations): Take visited args and
 764         a target cost vector.  Avoid processing already visited stmt sets.
 765         (vect_slp_analyze_operations): Use a local cost vector to gather
 766         costs and register those of non-discarded instances.
 767         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
 768         (vect_schedule_slp_instance): Remove copying of
 769         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
 770         zero.
 771         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
 772         adding cost.  Record cost entry location.
 773         (vect_prologue_cost_for_slp_op): Function to compute cost of
 774         a constant or invariant generated for SLP vect in the prologue,
 775         split out from vect_analyze_slp_cost_1.
 776         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
 777         (vect_model_promotion_demotion_cost): Likewise.
 778         (vect_model_store_cost): Likewise, make static.
 779         (vect_model_load_cost): Likewise.
 780         (vectorizable_bswap): Add cost vector arg and adjust.
 781         (vectorizable_call): Likewise.
 782         (vectorizable_simd_clone_call): Likewise.
 783         (vectorizable_conversion): Likewise.
 784         (vectorizable_assignment): Likewise.
 785         (vectorizable_shift): Likewise.
 786         (vectorizable_operation): Likewise.
 787         (vectorizable_store): Likewise.
 788         (vectorizable_load): Likewise.
 789         (vectorizable_condition): Likewise.
 790         (vectorizable_comparison): Likewise.
 791         (can_vectorize_live_stmts): Likewise.
 792         (vect_analyze_stmt): Likewise.
 793         (vect_transform_stmt): Adjust calls to vectorizable_*.
 794         * tree-vectorizer.c: Include gimple-pretty-print.h.
 795         (dump_stmt_cost): New function.
 796
 797 2018-05-16  Richard Biener  <rguenther@suse.de>
 798
 799         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
 800         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
 801         * tree-ssa-dse.c: Include tree-ssa-loop.h.
 802         (check_name): New callback.
 803         (dse_classify_store): Track cycles via a visited bitmap of PHI
 804         defs and simplify handling of in-loop and across loop dead stores
 805         and properly fail for loop-variant refs.  Handle byte-tracking with
 806         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
 807         limiting the walk.
 808
 809 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
 810
 811         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
 812         (vect_get_mask_type_for_stmt): Likewise.
 813         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
 814         split out from...
 815         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
 816         to determine the statement's vector type and the vector type that
 817         should be used for calculating nunits.  Deal with cases in which
 818         the type has to be deferred.
 819         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
 820         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
 821         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
 822         (vect_determine_vf_for_stmt): New functions, split out from...
 823         (vect_determine_vectorization_factor): ...here.
 824         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
 825         (vect_get_mask_type_for_stmt): New functions, split out from
 826         vect_determine_vectorization_factor.
 827
 828 2018-05-16  Richard Biener  <rguenther@suse.de>
 829
 830         * tree-cfg.c (verify_gimple_assign_ternary): Properly
 831         verify the [VEC_]COND_EXPR embedded comparison.
 832
 833 2018-05-15  Martin Sebor  <msebor@redhat.com>
 834
 835         PR tree-optimization/85753
 836         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
 837         RECORD_TYPE in addition to ARRAY_TYPE.
 838
 839 2018-05-15  Martin Sebor  <msebor@redhat.com>
 840
 841         PR middle-end/85643
 842         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
 843
 844 2018-05-15  Richard Biener  <rguenther@suse.de>
 845
 846         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
 847         add by_clobber_p one.  Change algorithm to collect all defs
 848         representing uses we need to walk and try reducing them to
 849         a single one before failing.
 850         (dse_dom_walker::dse_optimize_stmt): Adjust.
 851
 852 2018-05-13  Mark Wielaard  <mark@klomp.org>
 853
 854         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
 855         (size_of_loc_descr): Likewise.
 856         (output_loc_operands): Likewise.
 857         (output_loc_operands_raw): Likewise.
 858         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
 859         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
 860         (hash_loc_operands): Likewise.
 861         (compare_loc_operands): Likewise.
 862
 863 2018-05-14  Mark Wielaard  <mark@klomp.org>
 864
 865         * dwarf2out.c (count_index_addrs): New function.
 866         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
 867
 868 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
 869
 870         PR tree-optimization/83648
 871         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
 872         return value as malloc candidate.
 873
 874 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
 875
 876         PR ipa/85734
 877         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
 878         as true in call to suggest_attribute.
 879
 880 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
 881
 882         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
 883         -mreadonly-in-sdata.
 884
 885 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 886
 887         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
 888         New pattern.
 889         (aarch64_crypto_aesd_fused): Likewise.
 890
 891 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
 892
 893         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
 894         (movsi_aarch64): Likewise.
 895         (load_pairsi): Likewise.
 896         (load_pairdi): Likewise.
 897         (store_pairsi): Likewise.
 898         (store_pairdi): Likewise.
 899         (load_pairsf): Likewise.
 900         (load_pairdf): Likewise.
 901         (store_pairsf): Likewise.
 902         (store_pairdf): Likewise.
 903         (zero_extend): Likewise.
 904         (trunc): Swap alternatives.
 905         (fcvt_target): Add '?' to prefer w over r.
 906
 907 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
 908
 909         PR target/85756
 910         * config/i386/i386.md: Disallow non-commutative arithmetics in
 911         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
 912         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
 913         in the peephole2 before it.
 914
 915 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
 916
 917         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
 918         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
 919         (ix86_handle_option): Handle -mcldemote.
 920         * config.gcc: New header.
 921         * config/i386/cldemoteintrin.h: New file.
 922         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
 923         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
 924         -mcldemote.
 925         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
 926         OPTION_MASK_ISA_CLDEMOTE.
 927         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
 928         (ix86_valid_target_attribute_inner_p): Ditto.
 929         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
 930         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
 931         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
 932         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
 933         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
 934         (cldemote): New.
 935         * config/i386/i386.opt: Add -mcldemote.
 936         * config/i386/x86intrin.h: New header.
 937         * doc/invoke.texi: Add -mcldemote.
 938
 939 2018-05-14  Richard Biener  <rguenther@suse.de>
 940
 941         * doc/match-and-simplify.texi: Adjust :s documentation.
 942
 943 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
 944
 945         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
 946         intended memcpy size.
 947         (REORDER_45): Likewise.
 948
 949 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
 950
 951         * sort.cc: New file.
 952         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
 953         * vec.c (qsort_chk): Use gcc_qsort.
 954         * Makefile.in (OBJS-libcommon): Add sort.o.
 955         (build/sort.o): New target.  Use it...
 956         (BUILD_RTL): ... here, and...
 957         (build/gencfn-macros): ... here, and...
 958         (build/genmatch): ... here.
 959
 960 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
 961             Chung-Ju Wu  <jasonwucj@gmail.com>
 962
 963         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
 964         * config/nds32/nds32-graywolf.md: New file.
 965         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
 966         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
 967         pipeline.
 968         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
 969         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
 970         * config/nds32/nds32.md (pipeline_model): Add graywolf.
 971         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
 972         * config/nds32/pipelines.md: Include n15 settings.
 973
 974 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
 975             Chung-Ju Wu  <jasonwucj@gmail.com>
 976
 977         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
 978         * config/nds32/nds32-n13.md: New file.
 979         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
 980         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
 981         pipeline.
 982         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
 983         * config/nds32/nds32.md (pipeline_model): Add n13.
 984         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
 985         * config/nds32/pipelines.md: Include n13 settings.
 986
 987 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
 988             Chung-Ju Wu  <jasonwucj@gmail.com>
 989
 990         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
 991         * config/nds32/nds32-n10.md: New file.
 992         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
 993         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
 994         pipeline.
 995         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
 996         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
 997         * config/nds32/nds32.md (pipeline_model): Add n10.
 998         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
 999         * config/nds32/pipelines.md: Include n10 settings.
1000
1001 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
1002             Kito Cheng  <kito.cheng@gmail.com>
1003             Chung-Ju Wu  <jasonwucj@gmail.com>
1004
1005         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1006         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1007         Add enum values for DSP extension instructions.
1008         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1009         New constraints.
1010         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1011         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1012         New code iterators.
1013         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1014         * config/nds32/nds32-dspext.md: New file for DSP implementation.
1015         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1016         * config/nds32/nds32-intrinsic.md: Likewise.
1017         * config/nds32/nds32_intrinsic.h: Likewise.
1018         * config/nds32/nds32-md-auxiliary.c: Likewise.
1019         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1020         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1021         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1022         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1023         * config/nds32/nds32-protos.h: New declarations for DSP extension.
1024         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1025         TYPE_DMAC in switch statement.
1026         * config/nds32/nds32.c: New checking and implementation for DSP
1027         extension instructions.
1028         * config/nds32/nds32.h: Likewise.
1029         * config/nds32/nds32.md: Likewise.
1030         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1031         * config/nds32/predicates.md: Implement new predicates for DSP
1032         extension.
1033
1034 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
1035
1036         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1037         Reformat alternatives and attributes so it is easier to identify
1038         which constraints/attributes go with which instruction.
1039         (mov<mode>_hardfloat32, FMOVE64): Likewise.
1040         (mov<mode>_softfloat32, FMOVE64): Likewise.
1041         (mov<mode>_hardfloat64, FMOVE64): Likewise.
1042         (mov<mode>_softfloat64, FMOVE64): Likewise.
1043
1044 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
1045
1046         * doc/extend.texi (PowerPC Built-in Functions): Rename this
1047         subsection.
1048         (Basic PowerPC Built-in Functions): The new name of the
1049         subsection previously known as "PowerPC Built-in Functions".
1050         (Basic PowerPC Built-in Functions Available on all Configurations):
1051         New subsubsection.
1052         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1053         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1054         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1055         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1056
1057 2018-05-11  Martin Jambor  <mjambor@suse.cz>
1058
1059         PR ipa/85655
1060         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1061         single const.
1062
1063 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
1064
1065         PR target/85733
1066         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1067
1068 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
1069
1070         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1071         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1072         (ix86_handle_option): Handle -mwaitpkg.
1073         * config.gcc: New header.
1074         * config/i386/cpuid.h (bit_WAITPKG): New bit.
1075         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1076         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1077         function type.
1078         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1079         OPTION_MASK_ISA_WAITPKG.
1080         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1081         (ix86_option_override_internal): Add PTA_WAITPKG.
1082         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1083         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1084         IX86_BUILTIN_TPAUSE.
1085         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1086         __builtin_ia32_umwait and __builtin_ia32_tpause.
1087         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1088         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1089         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1090         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1091         UNSPECV_TPAUSE): New.
1092         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1093         * config/i386/i386.opt: Add -mwaitpkg.
1094         * config/i386/waitpkgintrin.h: New file.
1095         * config/i386/x86intrin.h: New header.
1096         * doc/invoke.texi: Add -mwaitpkg.
1097
1098 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
1099
1100         PR target/85606
1101         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1102         equivalent.
1103         (cortex-m0): Use armv6s-m isa.
1104         (cortex-m0plus): Likewise.
1105         (cortex-m1): Likewise.
1106         (cortex-m0.small-multiply): Likewise.
1107         (cortex-m0plus.small-multiply): Likewise.
1108         (cortex-m1.small-multiply): Likewise.
1109
1110 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
1111             Jakub Jelinek  <jakub@redhat.com>
1112
1113         PR tree-optimization/85692
1114         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1115         source permute as well.
1116
1117 2018-05-11  Martin Liska  <mliska@suse.cz>
1118
1119         PR sanitizer/85556
1120         * doc/extend.texi: Document LLVM style format for no_sanitize
1121         attribute.
1122
1123 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
1124
1125         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1126         mode_supports_vsx_dform_quad to mode_supports_dq_form.
1127         (mode_supports_vsx_dform_quad): Likewise.
1128         (mode_supports_vmx_dform): Move these functions to be next to the
1129         other mode_supports functions.
1130         (mode_supports_dq_form): Likewise.
1131         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1132         mode_supports_dq_form.
1133         (reg_offset_addressing_ok_p): Likewise.
1134         (offsettable_ok_by_alignment): Likewise.
1135         (rs6000_legitimate_offset_address_p): Likewise.
1136         (legitimate_lo_sum_address_p): Likewise.
1137         (rs6000_legitimize_address): Likewise.
1138         (rs6000_legitimize_reload_address): Likewise.
1139         (rs6000_secondary_reload_inner): Likewise.
1140         (rs6000_preferred_reload_class): Likewise.
1141         (rs6000_output_move_128bit): Likewise.
1142
1143 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
1144
1145         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1146         Generate SImode target register for null target.
1147         <case IX86_BUILTIN_XGETBV>: Ditto.
1148         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1149         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1150
1151 2018-05-10  Carl Love  <cel@us.ibm.com>
1152
1153         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1154         dcbtt and dcbtstt if operands[2] is 0.
1155
1156 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
1157
1158         PR target/85693
1159         * config/i386/sse.md (usadv64qi): New expander.
1160
1161 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
1162
1163         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1164         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1165         -maltivec=be support.
1166         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1167         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1168         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1169         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1170         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1171         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1172         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1173         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1174         altivec_vsumsws): Adjust.
1175         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1176         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1177         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1178         support.
1179         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1180         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1181         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1182         (altivec_lve<VI_char>x): Delete expand.
1183         (*altivec_lve<VI_char>x_internal): Rename to...
1184         (altivec_lve<VI_char>x): ... this.
1185         (altivec_lvxl_<mode>): Delete expand.
1186         (*altivec_lvxl_<mode>_internal): Rename to ...
1187         (altivec_lvxl_<mode>): ... this.
1188         (altivec_stvxl_<mode>): Delete expand.
1189         (*altivec_stvxl_<mode>_internal): Rename to ...
1190         (altivec_stvxl_<mode>): ... this.
1191         (altivec_stve<VI_char>x): Delete expand.
1192         (*altivec_stve<VI_char>x_internal): Rename to ...
1193         (altivec_stve<VI_char>x): ... this.
1194         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1195         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1196         reduc_plus_scal_<mode>): Adjust.
1197         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1198         comment.
1199         (rs6000_cpu_cpp_builtins): Adjust.
1200         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1201         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1202         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1203         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1204         -maltivec=be support.
1205         (rs6000_split_vec_extract_var): Adjust.
1206         (rs6000_split_v4si_init): Adjust.
1207         (swap_selector_for_mode): Delete.
1208         (altivec_expand_lvx_be, altivec_expand_stvx_be,
1209         altivec_expand_stvex_be): Delete.
1210         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1211         -maltivec=be support.
1212         (rs6000_gimple_fold_builtin): Ditto.
1213         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1214         Adjust.
1215         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1216         (TARGET_DIRECT_MOVE_64BIT): Adjust.
1217         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1218         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1219         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1220         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1221         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1222         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1223         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1224         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1225         anonymous split): Adjust.
1226         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1227         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1228
1229 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
1230
1231         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1232         when --with-gxx-include-dir is also specified.
1233         * configure: Regenerate.
1234
1235 2018-05-09  Jim Wilson  <jimw@sifive.com>
1236
1237         PR target/84797
1238         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1239         * config/riscv/t-withmultilib: New.
1240         * config/riscv/withmultilib.h: New.
1241         * doc/install.texi: Document RISC-V --with-multilib-list support.
1242
1243 2018-05-09  Richard Biener  <rguenther@suse.de>
1244
1245         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1246         vector.
1247         (vect_bb_vectorization_profitable_p): Adjust.  Compute
1248         actual scalar cost using the cost vector and the add_stmt_cost
1249         machinery.
1250
1251 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
1252
1253         PR rtl-optimization/85645
1254         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1255         in the REG_CFA_REGISTER note for LR, don't leave it empty.
1256
1257 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
1258
1259         PR rtl-optimization/85645
1260         * shrink-wrap.c (spread_components): Return a boolean saying if
1261         anything was changed.
1262         (try_shrink_wrapping_separate): Iterate spread_components until
1263         nothing changes anymore.
1264
1265 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
1266
1267         PR rtl-optimization/85645
1268         * regrename.c (build_def_use): Also kill the chains that include the
1269         destination of a REG_CFA_REGISTER note.
1270
1271 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
1272
1273         PR rtl-optimization/85645
1274         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1275         insn that has a REG_CFA_REGISTER note.
1276
1277 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
1278
1279         * cfgexpand.c (expand_clobber): New function.
1280         (expand_gimple_stmt_1): Use it.
1281         * tree-vect-stmts.c (vect_clobber_variable): New function,
1282         split out from...
1283         (vectorizable_simd_clone_call): ...here.
1284         (vectorizable_store): Emit a clobber either side of an
1285         IFN_STORE_LANES sequence.
1286         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1287
1288 2018-05-09  Tom de Vries  <tom@codesourcery.com>
1289
1290         PR target/85626
1291         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1292         (define_insn "trap_if_false"): Add exit after trap.
1293
1294 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
1295
1296         PR rtl-optimization/85638
1297         * bb-reorder.c: Include common/common-target.h.
1298         (create_forwarder_block): New function extracted from...
1299         (fix_up_crossing_landing_pad): ...here.  Rename into...
1300         (dw2_fix_up_crossing_landing_pad): ...this.
1301         (sjlj_fix_up_crossing_landing_pad): New function.
1302         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1303         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1304         from both partitions and exit the loop after one iteration.
1305
1306 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
1307
1308         Revert:
1309         * doc/extend.texi (PowerPC Built-in Functions): Rename this
1310         subsection.
1311         (Basic PowerPC Built-in Functions): The new name of the
1312         subsection previously known as "PowerPC Built-in Functions".
1313         (Basic PowerPC Built-in Functions Available on all Configurations):
1314         New subsubsection.
1315         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1316         subsubsection.
1317         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1318         subsubsection.
1319         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1320         subsubsection.
1321         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1322         subsubsection.
1323
1324 2018-05-08  Jim Wilson  <jimw@sifive.com>
1325
1326         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1327         (LD_EMUL_SUFFIX): New.
1328         (LINK_SPEC): Use it.
1329
1330 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
1331
1332         * doc/extend.texi (PowerPC Built-in Functions): Rename this
1333         subsection.
1334         (Basic PowerPC Built-in Functions): The new name of the
1335         subsection previously known as "PowerPC Built-in Functions".
1336         (Basic PowerPC Built-in Functions Available on all Configurations):
1337         New subsubsection.
1338         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1339         subsubsection.
1340         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1341         subsubsection.
1342         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1343         subsubsection.
1344         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1345         subsubsection.
1346
1347 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
1348
1349         PR target/85683
1350         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1351         after cmpelim optimization.
1352
1353 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
1354
1355         * config.gcc: Support "goldmont".
1356         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1357         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1358         PROCESSOR_GOLDMONT.
1359         * config/i386/i386.c (m_GOLDMONT): Define.
1360         (processor_target_table): Add "goldmont".
1361         (PTA_GOLDMONT): Define.
1362         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1363         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1364         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1365         (fold_builtin_cpu): Add "goldmont".
1366         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1367         (ix86_option_override_internal): Add "goldmont".
1368         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1369         (processor_type): Add PROCESSOR_GOLDMONT.
1370         * config/i386/i386.md: Add CPU "glm".
1371         * config/i386/glm.md: New file.
1372         * config/i386/x86-tune.def: Add m_GOLDMONT.
1373         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1374
1375 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
1376
1377         PR target/85572
1378         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1379         E_V4DImode.
1380         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1381         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
1382         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
1383
1384         PR target/85317
1385         * config/i386/i386.c (ix86_fold_builtin): Handle
1386         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1387
1388         PR target/85480
1389         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1390         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1391
1392 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
1393
1394         PR target/85658
1395         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1396         (check_arch): Likewise.
1397         (check_fpu): Return the result rather than printing it.
1398         (end arch): Fix operator precedence.
1399         (end cpu): Likewise.
1400         (END): Print the result from check_fpu.
1401
1402 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
1403             Alan Hayward  <alan.hayward@arm.com>
1404             David Sherwood  <david.sherwood@arm.com>
1405
1406         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1407         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1408         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1409         (*fcmuo<mode>_and): New patterns.
1410
1411 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
1412
1413         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1414         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1415         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1416         (cmp_op, sve_imm_con): New code attributes.
1417         (SVE_COND_INT_CMP, imm_con): Delete.
1418         (cmp_op): Remove above unspecs from int attribute.
1419         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1420         to...
1421         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
1422         comparison-specific unspecs.
1423         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1424         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1425         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1426         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1427         (*vec_fcm<cmp_op><mode>): Rename to...
1428         (*fcm<cmp_op><mode>): ...this and adjust likewise.
1429         (*vec_fcmuo<mode>): Rename to...
1430         (*fcmuo<mode>): ...this and adjust likewise.
1431         (*pred_fcm<cmp_op><mode>): New pattern.
1432         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1433         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1434         functions.
1435         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1436         and UNORDERED.
1437         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1438         (aarch64_emit_sve_predicated_cond): New function.
1439         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1440         (aarch64_emit_unspec_cond_or): Replace with...
1441         (aarch64_emit_sve_or_conds): ...this new function.  Use
1442         aarch64_emit_sve_ptrue_op for the individual comparisons and
1443         aarch64_emit_binop to OR them together.
1444         (aarch64_emit_inverted_unspec_cond): Replace with...
1445         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
1446         aarch64_emit_sve_ptrue_op for the comparison and
1447         aarch64_emit_unop to invert the result.
1448         (aarch64_expand_sve_vec_cmp_float): Update after the above
1449         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
1450
1451 2018-05-07  Nathan Sidwell  <nathan@acm.org>
1452
1453         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1454         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1455         (Backwards Compatibility): Likewise.
1456
1457 2018-05-07  Luis Machado  <luis.machado@linaro.org>
1458
1459         PR bootstrap/85681
1460         Revert:
1461         2018-05-07  Luis Machado  <luis.machado@linaro.org>
1462
1463         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1464         <prefetch_dynamic_strides>: New const bool field.
1465         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1466         prefetch_dynamic_strides.
1467         (exynosm1_prefetch_tune): Likewise.
1468         (thunderxt88_prefetch_tune): Likewise.
1469         (thunderx_prefetch_tune): Likewise.
1470         (thunderx2t99_prefetch_tune): Likewise.
1471         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1472         (aarch64_override_options_internal): Update to set
1473         PARAM_PREFETCH_DYNAMIC_STRIDES.
1474         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1475         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1476         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1477         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1478         prefetch-dynamic-strides setting.
1479
1480         2018-05-07  Luis Machado  <luis.machado@linaro.org>
1481
1482         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1483         <minimum_stride>: New const int field.
1484         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1485         minimum_stride field.
1486         (exynosm1_prefetch_tune): Likewise.
1487         (thunderxt88_prefetch_tune): Likewise.
1488         (thunderx_prefetch_tune): Likewise.
1489         (thunderx2t99_prefetch_tune): Likewise.
1490         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1491         (aarch64_override_options_internal): Update to set
1492         PARAM_PREFETCH_MINIMUM_STRIDE.
1493         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1494         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1495         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1496         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1497         stride is constant and is below the minimum stride threshold.
1498
1499 2018-05-07  Luis Machado  <luis.machado@linaro.org>
1500
1501         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1502         to 512.
1503
1504 2018-05-07  Luis Machado  <luis.machado@linaro.org>
1505
1506         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1507         <prefetch_dynamic_strides>: New const bool field.
1508         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1509         prefetch_dynamic_strides.
1510         (exynosm1_prefetch_tune): Likewise.
1511         (thunderxt88_prefetch_tune): Likewise.
1512         (thunderx_prefetch_tune): Likewise.
1513         (thunderx2t99_prefetch_tune): Likewise.
1514         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1515         (aarch64_override_options_internal): Update to set
1516         PARAM_PREFETCH_DYNAMIC_STRIDES.
1517         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1518         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1519         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1520         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1521         prefetch-dynamic-strides setting.
1522
1523 2018-05-07  Luis Machado  <luis.machado@linaro.org>
1524
1525         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1526         <minimum_stride>: New const int field.
1527         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1528         minimum_stride field.
1529         (exynosm1_prefetch_tune): Likewise.
1530         (thunderxt88_prefetch_tune): Likewise.
1531         (thunderx_prefetch_tune): Likewise.
1532         (thunderx2t99_prefetch_tune): Likewise.
1533         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1534         (aarch64_override_options_internal): Update to set
1535         PARAM_PREFETCH_MINIMUM_STRIDE.
1536         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1537         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1538         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1539         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1540         stride is constant and is below the minimum stride threshold.
1541
1542 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
1543
1544         PR c++/85659
1545         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1546         the type is addressable.  Don't force op into register if it has
1547         BLKmode.
1548
1549 2018-05-05  Roland McGrath  <mcgrathr@google.com>
1550
1551         PR other/77609
1552         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1553         any section for which we don't know a specific type it should have,
1554         regardless of name.  Previously this was done only for the exact
1555         names ".init_array", ".fini_array", and ".preinit_array".
1556         (default_elf_asm_named_section): Add comment about
1557         relationship with default_section_type_flags and SECTION_NOTYPE.
1558         (get_section): Don't consider it a type conflict if one side has
1559         SECTION_NOTYPE and the other doesn't, as long as neither has the
1560         SECTION_BSS et al used in the default_section_type_flags logic.
1561
1562 2018-05-05  Tom de Vries  <tom@codesourcery.com>
1563
1564         PR target/85653
1565         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1566         (workaround_barsyncs): New function.
1567         (nvptx_reorg): Use workaround_barsyncs.
1568         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1569         (define_expand "nvptx_membar_cta"): New define_expand.
1570         (define_insn "*nvptx_membar_cta"): New insn.
1571
1572 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
1573
1574         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1575         To improve optimization opportunities.
1576         * builtin-types.def: The new needed builtin types for the above.
1577
1578 2018-05-04  Richard Biener  <rguenther@suse.de>
1579
1580         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1581         * gimple-ssa-store-merging.c
1582         (imm_store_chain_info::output_merged_store): Remove redundant create,
1583         release split_store vector contents on failure.
1584         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1585         scalar stmt vector on cache hit.
1586
1587 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
1588
1589         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1590         Xilinx FP support.
1591         * config.gcc (powerpc-xilinx-eabi*): Remove.
1592         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1593         support.
1594         (fusion_addis_mem_combo_load): Ditto.
1595         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1596         FP support.
1597         (rs6000_cpu_cpp_builtins): Ditto.
1598         * config/rs6000/rs6000-linux.c
1599         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1600         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1601         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1602         support.
1603         (rs6000_setup_reg_addr_masks): Ditto.
1604         (rs6000_init_hard_regno_mode_ok): Ditto.
1605         (rs6000_option_override_internal): Ditto.
1606         (legitimate_lo_sum_address_p): Ditto.
1607         (rs6000_legitimize_address): Ditto.
1608         (rs6000_legitimize_reload_address): Ditto.
1609         (rs6000_legitimate_address_p): Ditto.
1610         (abi_v4_pass_in_fpr): Ditto.
1611         (setup_incoming_varargs): Ditto.
1612         (rs6000_gimplify_va_arg): Ditto.
1613         (rs6000_split_multireg_move): Ditto.
1614         (rs6000_savres_strategy): Ditto.
1615         (rs6000_emit_prologue_components): Ditto.
1616         (rs6000_emit_epilogue_components): Ditto.
1617         (rs6000_emit_prologue): Ditto.
1618         (rs6000_emit_epilogue): Ditto.
1619         (rs6000_elf_file_end): Ditto.
1620         (rs6000_function_value): Ditto.
1621         (rs6000_libcall_value): Ditto.
1622         * config/rs6000/rs6000.h: Ditto.
1623         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1624         (TARGET_MINMAX): ... this.  New.
1625         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1626         * config/rs6000/rs6000.md: Remove Xilinx FP support.
1627         (*movsi_internal1_single): Delete.
1628         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1629         mfpu=, mxilinx-fpu): Delete.
1630         * config/rs6000/singlefp.h: Delete.
1631         * config/rs6000/sysv4.h: Remove Xilinx FP support.
1632         * config/rs6000/t-rs6000: Ditto.
1633         * config/rs6000/t-xilinx: Delete.
1634         * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1635         * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1636         (VStype_simple): Delete.
1637         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1638         * config/rs6000/xfpu.h: Delete.
1639         * config/rs6000/xfpu.md: Delete.
1640         * config/rs6000/xilinx.h: Delete.
1641         * config/rs6000/xilinx.opt: Delete.
1642         * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1643         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1644
1645 2018-05-04  Tom de Vries  <tom@codesourcery.com>
1646
1647         PR libgomp/85639
1648         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1649         if ignore == 0.
1650
1651 2018-05-04  Richard Biener  <rguenther@suse.de>
1652
1653         PR middle-end/85627
1654         * tree-complex.c (update_complex_assignment): We are always in SSA form.
1655         (expand_complex_div_wide): Likewise.
1656         (expand_complex_operations_1): Likewise.
1657         (expand_complex_libcall): Preserve EH info of the original stmt.
1658         (tree_lower_complex): Handle removed blocks.
1659         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1660         on complex multiplication and division libcall builtins.
1661
1662 2018-05-04  Richard Biener  <rguenther@suse.de>
1663
1664         PR middle-end/85574
1665         * fold-const.c (negate_expr_p): Restrict negation of operand
1666         zero of a division to when we know that can happen without
1667         overflow.
1668         (fold_negate_expr_1): Likewise.
1669
1670 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
1671
1672         PR libstdc++/85466
1673         * real.h (real_nextafter): Declare.
1674         * real.c (real_nextafter): New function.
1675         * fold-const-call.c (fold_const_nextafter): New function.
1676         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1677         CASE_CFN_NEXTTOWARD.
1678         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1679         even when arg1_mode is different from arg0_mode.
1680
1681 2018-05-03  Nathan Sidwell  <nathan@acm.org>
1682
1683         * doc/extend.texi (Deprecated Features): Remove
1684         -ffriend-injection.
1685         (Backwards Compatibility): Likewise.
1686         * doc/invoke.texi (C++ Language Options): Likewise.
1687         (C++ Dialect Options): Likewise.
1688
1689 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
1690
1691         PR target/85530
1692         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1693         _mm512_mask_mullox_epi64): New intrinsics.
1694
1695 2018-05-03  Tom de Vries  <tom@codesourcery.com>
1696
1697         PR testsuite/85106
1698         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1699         dump files): Add offload-tree.
1700
1701 2018-05-03  Richard Biener  <rguenther@suse.de>
1702
1703         PR tree-optimization/85615
1704         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1705         to loops not nested in BBs loop father to avoid creating multi-entry
1706         loops.
1707
1708 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1709
1710         PR tree-optimization/70291
1711         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1712         arguments.  Change return type to tree.  Emit libcall as a new
1713         statement rather than replacing existing one when inplace_p is true.
1714         (expand_complex_multiplication_components): New function.
1715         (expand_complex_multiplication): Expand floating-point complex
1716         multiplication using the above.
1717         (expand_complex_division): Rename inner_type parameter to type.
1718         Update expand_complex_libcall call-site.
1719         (expand_complex_operations_1): Update expand_complex_multiplication
1720         and expand_complex_division call-sites.
1721
1722 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
1723
1724         PR target/85582
1725         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1726         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1727         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1728         the highest significant bit of the shift count mask is clear.  In
1729         check whether and[sq]i3 is needed verify that all significant bits
1730         of the shift count other than the highest are set.
1731
1732 2018-05-02  Tom de Vries  <tom@codesourcery.com>
1733
1734         PR libgomp/82428
1735         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1736         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1737         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1738         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1739         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1740         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1741         __builtin_goacc_parlevel_size.
1742
1743 2018-05-02  Richard Biener  <rguenther@suse.de>
1744
1745         PR tree-optimization/85597
1746         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1747         do not use split vect_get_vec_defs call but call vect_get_slp_defs
1748         directly.
1749
1750 2018-05-02  Tom de Vries  <tom@codesourcery.com>
1751
1752         PR testsuite/85106
1753         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1754         dump files): Add ltrans-tree.
1755
1756 2018-05-02  Tom de Vries  <tom@codesourcery.com>
1757
1758         PR testsuite/85106
1759         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1760         dump files): Add wpa-ipa.
1761
1762 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
1763
1764         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
1765         powerpc*-*-linux*paired* target.
1766         * config/rs6000/750cl.h: Delete.
1767         * config/rs6000/paired.h: Delete.
1768         * config/rs6000/paired.md: Delete.
1769         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1770         float support.
1771         * config/rs6000/rs6000-builtin.def: Remove paired float support.
1772         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1773         comment.  Remove paired float support.
1774         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1775         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1776         VECTOR_PAIRED.
1777         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1778         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1779         declarations.
1780         * config/rs6000/rs6000.c: Remove paired float support.
1781         (paired_expand_vector_init, paired_expand_vector_move,
1782         paired_emit_vector_compare, paired_emit_vector_cond_expr,
1783         (paired_expand_lv_builtin, paired_expand_stv_builtin,
1784         paired_expand_builtin, paired_expand_predicate_builtin,
1785         paired_init_builtins): Delete.
1786         * config/rs6000/rs6000.h: Remove paired float support.
1787         * config/rs6000/rs6000.md: Remove paired float support.
1788         (move_from_CR_ov_bit): Delete.
1789         * config/rs6000/rs6000.opt (mpaired): Delete.
1790         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1791         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1792
1793 2018-05-02  Richard Biener  <rguenther@suse.de>
1794
1795         PR middle-end/85567
1796         * gimplify.c (gimplify_save_expr): When in SSA form allow
1797         SAVE_EXPRs to compute to SSA vars.
1798
1799 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
1800
1801         PR target/85582
1802         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1803         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1804         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1805         clobber operands[2], instead use a new pseudo.  Formatting fixes.
1806
1807 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
1808
1809         PR tree-optimization/85586
1810         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1811         exit early for statements in the same group if the accesses are
1812         not strided.
1813
1814 2018-05-02  Tom de Vries  <tom@codesourcery.com>
1815
1816         PR lto/85451
1817         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1818         error message.
1819
1820 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
1821
1822         PR tree-optimization/85143
1823         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1824
1825 2018-05-01  Tom de Vries  <tom@codesourcery.com>
1826
1827         PR lto/85451
1828         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1829         not found" error message.
1830
1831 2018-05-01  Tom de Vries  <tom@codesourcery.com>
1832
1833         PR other/83786
1834         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1835         * vec.c (test_ordered_remove_if): New function.
1836         (vec_c_tests): Call test_ordered_remove_if.
1837         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1838         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1839         * tree-vect-patterns.c (vect_pattern_recog_1): Use
1840         VEC_ORDERED_REMOVE_IF.
1841
1842 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
1843
1844         PR tree-optimization/82665
1845         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1846         pointer subtraction where arguments come from a memchr call.
1847
1848 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
1849
1850         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1851         --push-state --as-needed and --pop-state instead of --as-needed and
1852         --no-as-needed if ld supports it.
1853         * configure: Regenerated.
1854
1855         PR web/85578
1856         * doc/install.texi2html: Replace _002d with - and _002a with * in
1857         generated html files using sed.
1858
1859 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
1860
1861         PR c++/85523
1862         * gcc-rich-location.c (blank_line_before_p): New function.
1863         (use_new_line): New function.
1864         (gcc_rich_location::add_fixit_insert_formatted): New function.
1865         * gcc-rich-location.h
1866         (gcc_rich_location::add_fixit_insert_formatted): New function.
1867
1868 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
1869
1870         * selftest.c (assert_streq): Rename "expected" and "actual" to
1871         "val1" and "val2".  Extend NULL-handling to cover both inputs
1872         symmetrically, while still requiring both to be non-NULL for a pass.
1873         * selftest.h (assert_streq): Rename "expected" and "actual" to
1874         "val1" and "val2".
1875         (ASSERT_EQ): Likewise.
1876         (ASSERT_EQ_AT): Likewise.
1877         (ASSERT_KNOWN_EQ): Likewise.
1878         (ASSERT_KNOWN_EQ_AT): Likewise.
1879         (ASSERT_NE): Likewise.
1880         (ASSERT_MAYBE_NE): Likewise.
1881         (ASSERT_MAYBE_NE_AT): Likewise.
1882         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
1883         the assertion to pass.
1884         (ASSERT_STREQ_AT): Likewise.
1885
1886 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
1887
1888         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1889         interaction with -pie.
1890
1891 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
1892
1893         * selftest.h: Fix alphabetization of per-source-file selftest
1894         declarations.
1895
1896 2018-04-30  Jason Merrill  <jason@redhat.com>
1897
1898         PR c++/61982 - dead stores to destroyed objects.
1899         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1900         of clobber.
1901
1902 2018-04-30  Jason Merrill  <jason@redhat.com>
1903
1904         * tree.c (build_clobber): New.
1905         * tree.h: Declare it.
1906         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1907
1908 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
1909
1910         * diagnostic-show-locus.c (layout::layout): Update for
1911         location_get_source_line returning a char_span.
1912         (struct char_span): Move to input.h.
1913         (struct correction): Update for fields in char_span becoming
1914         private.
1915         (struct source_line): Update for location_get_source_line
1916         returning a char_span.
1917         (layout::print_line): Likewise.
1918         * edit-context.c (edited_file::print_content): Likewise.
1919         (edited_file::print_diff_hunk): Likewise.
1920         (edited_file::print_run_of_changed_lines): Likewise.
1921         (edited_file::get_num_lines): Likewise.
1922         (edited_line::edited_line): Likewise.
1923         * final.c (asm_show_source): Likewise.
1924         * input.c (location_get_source_line): Convert return type
1925         from const char * to char_span, losing the final "line_len"
1926         param.
1927         (dump_location_info): Update for the above.
1928         (get_substring_ranges_for_loc): Likewise.  Use a char_span
1929         when handling the literal within the line.
1930         (test_reading_source_line): Update for location_get_source_line
1931         returning a char_span.
1932         * input.h (class char_span): Move here from
1933         diagnostic-show-locus.c, converting from a struct to a class.
1934         Make data members private.
1935         (char_span::operator bool): New.
1936         (char_span::length): New.
1937         (char_span::get_buffer): New.
1938         (char_span::operator[]): New.
1939         (char_span::subspan): Make const.
1940         (char_span::xstrdup): New.
1941         (location_get_source_line): Convert return type from const char *
1942         to char_span, losing the final "line_size" param.
1943
1944 2018-04-30  Jan Hubicka  <jh@suse.cz>
1945
1946         * lto-wrapper.c (ltrans_priorities): New static var.
1947         (cmp_priority): New.
1948         (run_gcc): Read priorities and if doing parallel build order
1949         the Makefile by them.
1950
1951 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
1952
1953         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1954
1955 2018-04-30  Richard Biener  <rguenther@suse.de>
1956
1957         * tree-cfg.c (verify_address): Remove base argument, add
1958         flag whether to check TREE_ADDRESSABLE and do that.
1959         (verify_expr): Remove.
1960         (verify_types_in_gimple_reference): Add pieces from verify_expr.
1961         (verify_gimple_assign_single): Likewise.
1962         (verify_gimple_switch): Likewise.
1963         (verify_expr_location_1): Dereference tp once.  Add (disabled)
1964         piece from verify_expr.
1965         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1966
1967 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
1968
1969         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1970
1971 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
1972
1973         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1974         (small_data_pattern): Likewise.
1975         (arc_rewrite_small_data): Likewise.
1976         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1977         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1978         (get_symbol_alignment): New function.
1979         (legitimate_small_data_address_p): Likewise.
1980         (legitimate_scaled_address): Update, call
1981         legitimate_small_data_address_p.
1982         (output_sdata): New static variable.
1983         (arc_print_operand): Update how we handle small data operands.
1984         (arc_print_operand_address): Likewise.
1985         (arc_legitimate_address_p): Update, use
1986         legitimate_small_data_address_p.
1987         (arc_rewrite_small_data_p): Remove.
1988         (arc_rewrite_small_data_1): Likewise.
1989         (arc_rewrite_small_data): Likewise.
1990         (small_data_pattern): Likewise.
1991         (compact_sda_memory_operand): Update to use
1992         legitimate_small_data_address_p and get_symbol_alignment.
1993         (prepare_move_operands): Don't rewite sdata pattern.
1994         (prepare_extend_operands): Remove.
1995         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1996         pattern.
1997         (zero_extendqisi2): Likewise.
1998         (zero_extendhisi2): Likewise.
1999         (extendqihi2): Likewise.
2000         (extendqisi2): Likewise.
2001         (extendhisi2): Likewise.
2002         (addsi3): Likewise.
2003         (subsi3): Likewise.
2004         (andsi3): Likewise.
2005         * config/arc/constraints.md (Usd): Change it to memory constraint.
2006
2007 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
2008
2009         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2010         as source of std instructions.
2011         * config/arc/arc.md (movsi_insn): Update pattern predicate to
2012         allow 6-bit constants as source for store instructions.
2013         (movdi_insn): Update instruction pattern to allow 6-bit constants
2014         as source for store instructions.
2015
2016 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
2017
2018         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2019
2020 2018-04-30  Nathan Sidwell  <nathan@acm.org>
2021             Sandra Loosemore <sandra@codesourcery.com>
2022
2023         * dumpfile.c (dump_open): Allow '-' for stdout.
2024         * doc/invoke.texi (Developer Options): Document dump filename
2025         determination early.  Document stdin/stdout selection.
2026
2027 2018-04-30 Andrew Sadek  <andrew.sadek.se@gmail.com>
2028
2029         Microblaze Target: PIC data text relative
2030
2031         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2032         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2033         Add declaration.
2034         * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2035         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2036         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2037         New addressing mode for data-text relative position indepenedent code.
2038         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2039         'ADDRESS_SYMBOLIC_TXT_REL'.
2040         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2041         (microblaze_legitimate_pic_operand): Exclude function calls from
2042         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2043         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2044         addresses cases.
2045         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2046         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2047         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2048         for 'address + offset'.
2049         (microblaze_expand_prologue): Add new function prologue call for
2050         'r20' assignation.
2051         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2052         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2053         table in case of TARGET_PIC_DATA_TEXT_REL.
2054         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2055         * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2056         Add new macros 'UNSPEC_TEXT',
2057         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2058         + exclude function calls from 'UNSPEC_PLT' in case of data text
2059         relative mode.
2060         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2061         new target hook for generating address diff vector tables in case of
2062         flag_pic.
2063         * doc/tm.texi : Regenerate.
2064         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2065         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2066         of addr diff vector generation.
2067         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2068         target hook definition.
2069         * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2070         Add default function for generate_pic_addr_diff_vec -> flag_pic.
2071         * doc/invoke.texi (Add new pic option): Add new microblaze pic
2072         option for data text relative.
2073
2074 2018-04-30  Richard Biener  <rguenther@suse.de>
2075
2076         * tree-chrec.h (evolution_function_is_constant_p): Remove
2077         redundant check.
2078         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2079
2080 2018-04-30  Richard Biener  <rguenther@suse.de>
2081
2082         PR bootstrap/85571
2083         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2084
2085 2018-04-30  Richard Biener  <rguenther@suse.de>
2086
2087         PR tree-optimization/28364
2088         PR tree-optimization/85275
2089         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2090         copying first exit test.
2091
2092 2018-04-28  Mark Wielaard  <mark@klomp.org>
2093
2094         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2095         dwarf_version >= 5.
2096         (dwarf_AT): Handle DW_AT_addr_base.
2097         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2098
2099 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
2100
2101         PR target/84431
2102         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2103         (*ashl<dwi>3_doubleword_mask_1): Ditto.
2104         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2105         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2106
2107 2018-04-28  Richard Biener  <rguenther@suse.de>
2108
2109         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2110         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2111         to reflect use.  Only add interesting stmts.
2112
2113 2018-04-27  Martin Jambor  <mjambor@suse.cz>
2114
2115         PR ipa/85549
2116         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2117         the jump function allows for passing through aggregate values.
2118
2119 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
2120
2121         * input.h (in_system_header_at): Convert from macro to inline
2122         function.
2123         (from_macro_expansion_at): Likewise.
2124         (from_macro_definition_at): Likewise.
2125
2126 2018-04-27  Jeff Law  <law@redhat.com>
2127
2128         * config.gcc: Mark tile* targets as deprecated/obsolete.
2129
2130 2018-04-27  Richard Biener  <rguenther@suse.de>
2131
2132         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2133         fix for ILP32.
2134
2135 2018-04-27  Richard Biener  <rguenther@suse.de>
2136
2137         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2138
2139 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
2140
2141         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2142         with Yd constraint. Set "preferred_for_speed" attribute from
2143         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2144         with Yd constraint.
2145         (*movdi_internal): Ditto.
2146         (movti_interunit splitters): Remove
2147         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2148         (movdi_interunit splitters): Ditto.
2149         * config/i386/constraints.md (Ye): Remove.
2150         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2151
2152 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2153
2154         PR target/85512
2155         * config/aarch64/constraints.md (Usg): Limit to 31.
2156         (Usj): Limit to 63.
2157
2158 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
2159
2160         PR tree-optimization/85529
2161         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2162         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2163         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2164         zero extension or masking of the MSB bit.
2165         (optimize_range_tests): Add FIRST_BB argument, pass it through
2166         to optimize_range_tests_var_bound.
2167         (maybe_optimize_range_tests, reassociate_bb): Adjust
2168         optimize_range_tests callers.
2169
2170 2018-04-26  Richard Biener  <rguenther@suse.de>
2171             Jakub Jelinek  <jakub@redhat.com>
2172
2173         * cgraph.h (symbol_table): Just declare debug method here.
2174         * symtab.c (symbol_table::debug): Define.
2175
2176 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
2177
2178         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2179
2180 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
2181
2182         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2183         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2184         (*movdi_internal): Substitute Yi and Yj constraint with x
2185         and Ym and Yn constraint with y constraint.  Update "isa"
2186         attribute and set "preferred_for_speed" attribute from
2187         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2188         (*movsi_internal): Ditto.
2189         (*movdf_internal): Ditto.
2190         (*movsf_internal): Ditto.
2191         (*zero_extendsidi2): Ditto.
2192         * config/i386/sse.md (vec_set<mode>_0): Ditto.
2193         (sse2_loadld): Ditto.
2194         (*vec_extract<ssevecmodelower>_0): Ditto.
2195         (*vec_extractv4si_0_zext_sse4): Ditto.
2196         (vec_concatv2di): Ditto.
2197         (*vec_dup<mode>): Ditto.
2198         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2199         * config/i386/constraints.md (Yi): Remove.
2200         (Yj): Remove.
2201         (Ym): Remove.
2202         (Yn): Remove.
2203
2204 2018-04-26  Nathan Sidwell  <nathan@acm.org>
2205
2206         * dumpfile.c (dump_open): New.
2207         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2208         (dump_finish): Detect stdio/stderr by value not name.
2209
2210 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
2211
2212         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2213
2214 2018-04-26  Tom de Vries  <tom@codesourcery.com>
2215
2216         PR target/84952
2217         * config/nvptx/nvptx.c (verify_neutering_jumps)
2218         (verify_neutering_labels): New function
2219         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2220
2221 2018-04-26  Tom de Vries  <tom@codesourcery.com>
2222
2223         PR target/84025
2224         * config/nvptx/nvptx.c (needs_neutering_p): New function.
2225         (nvptx_single): Use needs_neutering_p to skip over insns that do not
2226         need neutering.
2227
2228 2018-04-26  Richard Biener <rguenther@suse.de>
2229             Tom de Vries  <tom@codesourcery.com>
2230
2231         PR lto/85422
2232         * lto-streamer-out.c (output_function): Fixup loops if required to match
2233         discovery done in the reader.
2234
2235 2018-04-26  Richard Biener  <rguenther@suse.de>
2236
2237         PR tree-optimization/85116
2238         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2239         have a loop exit from the single latch predecessor.  Remove
2240         case of header with just condition.
2241         (ch_base::copy_headers): Exclude infinite loops from any
2242         processing.
2243         (pass_ch::execute): Record exits.
2244
2245 2018-04-26  Richard Biener  <rguenther@suse.de>
2246
2247         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2248         prologue cost vector and pass it to vect_get_load_cost.
2249         (vect_get_peeling_costs_all_drs): Likewise.
2250         (vect_peeling_hash_get_lowest_cost): Likewise.
2251         (vect_enhance_data_refs_alignment): Likewise.
2252
2253 2018-04-26  Richard Biener  <rguenther@suse.de>
2254
2255         PR middle-end/85450
2256         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2257         checking of integer<->pointer conversions.
2258         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2259         sign-/zero-extending pointer types.
2260         (expand_omp_for_static_chunk): Likewise.
2261
2262 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
2263             Jean Lee  <xiaoyur347@gmail.com>
2264
2265         * config/mips/mips.c (mips_asan_shadow_offset): New function.
2266         (TARGET_ASAN_SHADOW_OFFSET): Define.
2267         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2268         true for -fsanitize=address.
2269
2270 2018-04-25  Mark Wielaard  <mark@klomp.org>
2271
2272         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2273         shorter ones.
2274
2275 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
2276
2277         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2278         than "alu", remove explicit "memory" and "imm_disp" attributes.
2279         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2280
2281         PR middle-end/85414
2282         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2283         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2284         gen_lowpart_no_emit.
2285
2286 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
2287
2288         PR target/85473
2289         * config/i386/i386.c (ix86_expand_builtin): Change memory
2290         operand to XI, extend p0 to Pmode.
2291         * config/i386/i386.md: Change unspec volatile and operand
2292         1 mode to XI, change operand 0 mode to P.
2293
2294 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
2295
2296         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2297         GET_MODE_MASK before any checking.
2298         (nds32_can_use_bset_p): Likewise.
2299         (nds32_can_use_btgl_p): Likewise.
2300
2301 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
2302
2303         * config/nds32/nds32-doubleword.md: New define_split pattern for
2304         illegal register number.
2305
2306 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
2307
2308         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2309
2310 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
2311
2312         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2313
2314 2018-04-25  Richard Biener  <rguenther@suse.de>
2315
2316         * lto-streamer.h (LTO_major_version): Bump to 8.
2317
2318 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
2319
2320         * BASE-VER: Set to 9.0.0.
2321
2322 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
2323
2324         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2325         in __abskf2 and __powikf2.
2326
2327 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2328
2329         PR target/85512
2330         * config/aarch64/constraints.md (Usg, Usj): New constraints.
2331         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2332         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2333         Use the above on operand 2.  Reindent.
2334         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2335
2336 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
2337
2338         PR target/85485
2339         * common/config/i386/i386-common.c (ix86_handle_option): Don't
2340         handle OPT_mcet.
2341         * config/i386/i386.opt (mcet): Removed.
2342         * doc/install.texi: Remove -mcet documentation.
2343         * doc/invoke.texi: Likewise.
2344
2345 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
2346
2347         PR target/85485
2348         * doc/install.texi: Remove -mcet from bootstrap-cet.
2349
2350 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
2351
2352         PR target/85511
2353         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2354         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2355         if TARGET_64BIT.
2356
2357         PR target/85503
2358         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2359         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2360         containing a CONST_VECTOR.
2361
2362 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
2363
2364         * doc/install.texi: Update newlib dependency for nvptx.
2365
2366 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
2367
2368         PR target/85508
2369         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2370         instead of INTVAL when shifting x left.
2371
2372 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
2373
2374         PR tree-optimization/85478
2375         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2376         vect_grouped_store_supported for single element vectors.
2377
2378 2018-04-24  Richard Biener  <rguenther@suse.de>
2379
2380         PR target/85491
2381         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2382         load cost increase to the case of non-constant step.
2383
2384 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
2385
2386         PR target/84828
2387         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2388         destination if any_malformed_asm.
2389
2390 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
2391
2392         PR middle-end/85496
2393         * expr.c (store_field): In the bitfield case, if the value comes from
2394         a function call and is returned in registers by means of a PARALLEL,
2395         do not change the mode of the temporary unless BLKmode and VOIDmode.
2396
2397 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
2398
2399         PR rtl-optimization/85423
2400         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2401         dependencies to debug insns when the previous insn is non-debug.
2402
2403 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
2404
2405         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2406         enums into a single definition.
2407         (fls): Fix predicates and printing.
2408         (seti): Likewise.
2409
2410 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
2411
2412         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2413         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2414         and short u6 immediate.
2415         (check_if_valid_sleep_operand): Remove.
2416         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2417
2418 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
2419
2420         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2421         flag_always_save_lp condition.
2422         * config/nds32/nds32.opt (malways-save-lp): New option.
2423
2424 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
2425
2426         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2427         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2428         * config/nds32/nds32.h
2429         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2430         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2431
2432 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
2433
2434         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2435         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2436
2437 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
2438             Chung-Ju Wu  <jasonwucj@gmail.com>
2439
2440         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2441         Declare.
2442         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2443         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2444
2445 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
2446
2447         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2448
2449 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
2450
2451         * config/nds32/nds32-protos.h (nds32_data_alignment,
2452         nds32_local_alignment): Declare.
2453         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2454         nds32_local_alignment): New functions.
2455         (TARGET_CONSTANT_ALIGNMENT): Define.
2456         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2457
2458 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
2459
2460         * config/nds32/nds32.c
2461         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2462         (TARGET_MODES_TIEABLE_P): Likewise.
2463
2464 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
2465
2466         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2467         level Ofast and Og.
2468
2469 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
2470             Chung-Ju Wu  <jasonwucj@gmail.com>
2471
2472         * config/nds32/constants.md (unspec_volatile_element): Add enum values
2473         for unaligned access.
2474         * config/nds32/nds32-intrinsic.c: Implementation of expanding
2475         unaligned access.
2476         * config/nds32/nds32-intrinsic.md: Likewise.
2477         * config/nds32/nds32_intrinsic.h: Likewise.
2478         * config/nds32/nds32.h (nds32_builtins): Likewise.
2479         * config/nds32/nds32.opt (munaligned-access): New option.
2480         * config/nds32/nds32.c (nds32_asm_file_start): Display
2481         flag_unaligned_access status.
2482
2483 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
2484
2485         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2486         -mno-relax is present.
2487         * config/riscv/linux.h (LINK_SPEC): Ditto.
2488
2489 2018-04-20  Martin Sebor  <msebor@redhat.com>
2490
2491         PR c/85365
2492         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2493         for null pointers.
2494         (gimple_fold_builtin_stxcpy_chk): Same.
2495         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2496
2497 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
2498
2499         PR target/85456
2500         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2501         __powikf2 when long double is IEEE 128-bit.
2502
2503 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
2504
2505         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2506         step to make sure stack always aligned.
2507
2508 2018-04-20  Carl Love  <cel@us.ibm.com>
2509
2510         PR target/83402
2511         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2512         size check for arg0.
2513
2514 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
2515             Tom de Vries  <tom@codesourcery.com>
2516
2517         PR target/85445
2518         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2519         Emit insns for calls too.
2520         (nvptx_find_par): Always look for worker-level predecessor insn.
2521         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
2522         calls.
2523         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2524         (nvptx_process_pars): Propagate frames for calls.
2525
2526 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
2527
2528         PR target/85469
2529         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2530         Removed.
2531         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2532         (ix86_handle_option): Don't handle OPT_mibt.
2533         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2534         __SHSTK__.
2535         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2536         has_ibt and ibt.
2537         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2538         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2539         (ix86_target_macros): Define __CET__ with flag_cf_protection
2540         for -fcf-protection.
2541         * config/i386/i386.c (isa2_opts): Remove -mibt.
2542         * config/i386/i386.h (TARGET_IBT): Removed.
2543         (TARGET_IBT_P): Likewise.
2544         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2545         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2546         * config/i386/i386.opt (mcet): Update help message.
2547         (mshstk): Likewise.
2548         (mibt): Removed.
2549         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
2550         -mcet as an alias for -mshstk.
2551
2552 2018-04-20  Richard Biener <rguenther@suse.de>
2553
2554         PR middle-end/85475
2555         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2556         complexity by forcing a single use of the multiply operand.
2557
2558 2018-04-20  Martin Jambor  <mjambor@suse.cz>
2559
2560         ipa/85449
2561         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2562         recursion dependency to only apply to non-clones.
2563
2564 2018-04-20  Martin Jambor  <mjambor@suse.cz>
2565
2566         ipa/85447
2567         * ipa-cp.c (create_specialized_node): Check that clones of
2568         self-recursive edges exist during IPA-CP.
2569
2570 2018-04-19  Toon Moene  <toon@moene.org>
2571
2572         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2573         by -O3.
2574
2575 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
2576
2577         PR tree-optimization/85467
2578         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2579         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
2580         VECTOR_CST element to type.
2581
2582 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
2583
2584         PR target/85397
2585         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2586         * config/i386/i386.md (builtin_setjmp_setup): Removed.
2587         (builtin_longjmp): Likewise.
2588         (save_stack_nonlocal): New pattern.
2589         (restore_stack_nonlocal): Likewise.
2590
2591 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
2592
2593         PR target/85404
2594         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2595         Replace ASM_OUTPUT_LABEL with fprintf.
2596
2597 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
2598
2599         PR target/85417
2600         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2601         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2602         * config/i386/i386-c.c (ix86_target_macros_internal): Also
2603         define __IBT__ and __SHSTK__ for -fcf-protection.
2604         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2605         TARGET_IBT.
2606         (ix86_trampoline_init): Likewise.
2607         (x86_output_mi_thunk): Likewise.
2608         (ix86_notrack_prefixed_insn_p): Likewise.
2609         (ix86_option_override_internal): Don't disallow -fcf-protection.
2610         * config/i386/i386.md (rdssp<mode>): Also enable for
2611         -fcf-protection.
2612         (incssp<mode>): Likewise.
2613         (nop_endbr): Likewise.
2614         * config/i386/i386.opt (mcet): Change help message to built-in
2615         functions only.
2616         (mibt): Likewise.
2617         (mshstk): Likewise.
2618         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2619         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
2620         enable CET built-in functions.
2621
2622 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
2623
2624         * common/config/i386/i386-common.c
2625         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2626         OPTION_MASK_ISA_MOVDIRI_UNSET,
2627         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2628         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2629         * config.gcc (movdirintrin.h): New header.
2630         * config/i386/cpuid.h (bit_MOVDIRI,
2631         bit_MOVDIR64B): New bits.
2632         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2633         and -mmvodir64b.
2634         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2635         (VOID, PVOID, PCVOID)): New function types.
2636         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2637         __builtin_ia32_directstoreu_u64,
2638          __builtin_ia32_movdir64b): New builtins.
2639         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2640         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2641         and -mmovdiri.
2642         (ix86_valid_target_attribute_inner_p): Ditto.
2643         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2644         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2645         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2646         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2647         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2648         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2649         (movdiri<mode>, movdir64b_<mode>): New.
2650         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2651         * config/i386/immintrin.h: Include movdirintrin.h.
2652         * config/i386/movdirintrin.h: New file.
2653         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2654
2655 2018-04-19  Richard Biener  <rguenther@suse.de>
2656
2657         PR middle-end/85455
2658         * cfg.c (clear_bb_flags): When loop state says we have
2659         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2660
2661 2018-04-19  Richard Biener  <rguenther@suse.de>
2662
2663         PR tree-optimization/84737
2664         * tree-vect-data-refs.c (vect_copy_ref_info): New function
2665         copying restrict info.
2666         (vect_setup_realignment): Use it.
2667         * tree-vectorizer.h (vect_copy_ref_info): Declare.
2668         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2669         the first DR to all generated stores.
2670         (vectorizable_load): Likewise for loads.
2671
2672 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
2673
2674         PR tree-optimization/85446
2675         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2676         the integral and pointer types to have the same precision.
2677
2678         * doc/install.texi: Document --disable-cet being the default and
2679         --enable-cet=auto.
2680
2681 2018-04-18  Martin Liska  <mliska@suse.cz>
2682
2683         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2684         style.
2685
2686 2018-04-18  Martin Liska  <mliska@suse.cz>
2687
2688         Revert
2689         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
2690
2691         PR ipa/83983
2692         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2693         arguments if they are comparable.
2694
2695 2018-04-18  Martin Liska  <mliska@suse.cz>
2696
2697         Revert
2698         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
2699
2700         PR lto/84805
2701         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2702         incomplete types.
2703
2704 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
2705
2706         PR target/85388
2707         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2708         ENDBR after calling __morestack.
2709
2710 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
2711
2712         PR jit/85384
2713         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2714         by using gcc_base_ver to generate a gcc_driver_version, and use
2715         it when generating GCC_DRIVER_NAME.
2716         * configure: Regenerate.
2717
2718 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
2719
2720         PR target/81084
2721         * config.gcc: Obsolete powerpc*-*-*spe*.
2722
2723 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
2724
2725         PR debug/84637
2726         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2727         (stabstr_D): Change type of unum from unsigned int to
2728         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
2729         type.
2730
2731 2018-04-17  Jim Wilson  <jimw@sifive.com>
2732
2733         PR 84856
2734         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2735         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2736         Set arg_pointer_offset after using pretend_args_size.
2737
2738 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
2739
2740         PR rtl-optimization/85431
2741         * dse.c (record_store): Ignore zero width stores.
2742
2743         PR sanitizer/85230
2744         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
2745         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2746         __builtin_stack_restore rather than after it.
2747         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2748         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2749         argument instead of virtual_dynamic_stack_rtx.
2750
2751 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
2752
2753         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2754         New prototype.
2755         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2756         Add note to error message to explain internal mapping of overloaded
2757         built-in function name to non-overloaded built-in function name.
2758         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2759         function.
2760
2761 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
2762
2763         PR target/85424
2764         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2765         where the inputs overlap with the output.
2766
2767 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
2768
2769         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2770         (=v, v) alternative and explicit "memory" attribute.
2771         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
2772         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2773         attributes.
2774         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2775         "sselog1" type instead of "sselog".
2776         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2777         "sselog".  Remove explicit "memory" attribute.
2778         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2779         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2780         attributes.
2781         (vec_extract_hi_v32hi): Merge all alternatives into one, use
2782         "sselog1" type instead of "sselog".  Remove explicit "memory"
2783         attribute.
2784         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2785         use "sselog1" type instead of "sselog".  Remove explicit "memory"
2786         attribute.
2787         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2788         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2789         attributes.
2790         (vec_extract_hi_v64qi): Merge all alternatives into one, use
2791         "sselog1" type instead of "sselog".  Remove explicit "memory"
2792         attribute.
2793         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2794         use "sselog1" type instead of "sselog".  Remove explicit "memory"
2795         attribute.
2796
2797         PR target/85430
2798         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2799
2800         PR middle-end/85414
2801         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2802         on a SUBREG.
2803
2804 2018-04-17  Martin Jambor  <mjambor@suse.cz>
2805
2806         PR ipa/85421
2807         * ipa-cp.c (create_specialized_node): Call
2808         expand_all_artificial_thunks if necessary.
2809
2810 2018-04-17  Martin Liska  <mliska@suse.cz>
2811
2812         PR lto/85405
2813         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2814         in message, remote space in between '_G' and '('.
2815
2816 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
2817
2818         PR target/85281
2819         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2820         avx512f_vmcmp<mode>3<round_saeonly_name>,
2821         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2822         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2823         avx512f_rndscale<mode><round_saeonly_name>,
2824         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2825         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2826         Use %<iptr>2 instead of %2 for -masm=intel.
2827         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2828         avx512f_vcvttss2usi<round_saeonly_name>,
2829         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2830         -masm=intel.
2831         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2832         avx512f_vcvttsd2usi<round_saeonly_name>,
2833         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2834         Use %q1 instead of %1 for -masm=intel.
2835         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2836         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2837         of %3 for -masm=intel.
2838         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2839         -masm=intel.
2840         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2841         -masm=intel.
2842         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2843         -masm=intel.
2844         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2845         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2846         %g1.
2847         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2848         -masm=intel.
2849         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2850         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2851         %g1 and one with %0 and %1.
2852         (avx512er_vmrcp28<mode><round_saeonly_name>,
2853         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2854         %1 for -masm=intel.
2855         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2856         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2857         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2858         of %0 and %{%4%} for -masm=intel.
2859         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2860         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2861         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2862         order of %0 and %{%5%}%{z%} for -masm=intel.
2863
2864 2018-04-17  Jan Hubicka  <jh@suse.cz>
2865
2866         PR lto/85405
2867         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2868
2869 2018-04-17  Martin Liska  <mliska@suse.cz>
2870
2871         PR ipa/85329
2872         * multiple_target.c (create_dispatcher_calls): Set apostrophes
2873         for target_clone error message.  Make default implementation
2874         clone to be a local declaration.
2875         (separate_attrs): Add new argument and check for an empty
2876         string.
2877         (expand_target_clones): Handle it.
2878         (ipa_target_clone): Make redirection just for target_clones
2879         functions.
2880
2881 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
2882             Tom de Vries  <tom@codesourcery.com>
2883
2884         PR middle-end/84955
2885         * omp-expand.c (expand_oacc_for): Add dummy false branch for
2886         tiled basic blocks without omp continue statements.
2887
2888 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
2889
2890         PR target/83660
2891         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2892         vec_extract expression as having side effects to make sure it gets
2893         a cleanup point.
2894
2895 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
2896
2897         PR target/85403
2898         * config/i386/i386.c (get_builtin_code_for_version): Check
2899         error_mark_node.
2900
2901 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
2902
2903         PR target/84331
2904         * gcc/config.gcc: Support "skylake".
2905         * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2906         PROCESSOR_SKYLAKE.
2907         * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2908         (processor_target_table): Add "skylake".
2909         (ix86_option_override_internal): Add "skylake".
2910         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2911         PROCESSOR_CANNONLAKE.
2912         (get_builtin_code_for_version): Fix priority for
2913         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2914         PROCESSOR_SKYLAKE-AVX512.
2915         * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2916         (processor_type): Add PROCESSOR_SKYLAKE.
2917
2918 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
2919             Jason Merrill  <jason@redhat.com>
2920
2921         PR c++/85112
2922         * convert.c (convert_to_integer_1): Use direct recursion for
2923         enumeral types and types with a precision less than the number
2924         of bits in their mode.
2925
2926 2018-04-16  Julia Koval  <julia.koval@intel.com>
2927
2928         PR target/84413
2929         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2930         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2931
2932 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
2933
2934         PR target/85293
2935         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2936         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2937         and -mno-direct-move.
2938
2939 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
2940
2941         PR target/83402
2942         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2943         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2944         Ensure negative shifts result in {0}.
2945
2946 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
2947
2948         PR rtl-optimization/79916
2949         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2950         regs (if any) to define how to gnerate SD moves when LRA is in
2951         progress.
2952
2953 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
2954
2955         PR rtl-optimization/85393
2956         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2957         * except.c (expand_dw2_landing_pad_for_region): Make static.
2958         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2959         a label and unconditional jump to old_bb, rather than
2960         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2961         basic block.
2962
2963         PR rtl-optimization/85376
2964         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2965         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2966         instead of a specific value.
2967
2968 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
2969             Bin Cheng  <bin.cheng@arm.com>
2970
2971         PR tree-optimization/82965
2972         PR tree-optimization/83991
2973         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2974         by_profile_only parameter.
2975         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2976         information if the loop was predicted to iterate too many times.
2977         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2978
2979 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
2980
2981         PR lto/71991
2982         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2983         always inline.
2984
2985 2018-04-13  Martin Liska  <mliska@suse.cz>
2986             Jakub Jelinek  <jakub@redhat.com>
2987
2988         PR middle-end/81657
2989         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2990         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2991         * builtins.c (expand_builtin_memory_copy_args): Use
2992         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2993         handle dest_addr == pc_rtx.
2994
2995 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
2996
2997         PR target/85291
2998         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2999         asked to not generate direct moves.
3000         (fix_trunc<mode>si2_stfiwx): Similar.
3001         (fix_trunc<mode>si2_internal): Similar.
3002
3003 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
3004
3005         PR debug/83157
3006         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3007         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3008         lookup if dest in some wider mode is known to be const0_rtx and
3009         if so, record permanent equivalence for it to be ZERO_EXTEND of
3010         the narrower mode destination.
3011
3012 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
3013
3014         * lto-streamer-out.c (output_function): Revert 259346.
3015         * omp-expand.c (expand_oacc_for): Likewise.
3016
3017 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
3018
3019         PR rtl-optimization/85354
3020         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3021         * sel-sched.c (sel_global_init): ... here.
3022
3023 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
3024
3025         PR target/85238
3026         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3027         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3028         mode for PE-COFF targets.
3029         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3030         (i386_pe_asm_lto_end): Likewise.
3031         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3032         (TARGET_ASM_LTO_END): Likewise.
3033         * config/i386/winnt.c (saved_debug_info_level): New static variable.
3034         (i386_pe_asm_lto_start): New function.
3035         (i386_pe_asm_lto_end): Likewise.
3036
3037 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
3038             Richard Biener  <rguenther@suse.de>
3039
3040         PR middle-end/84955
3041         * lto-streamer-out.c (output_function): Fix CFG loop state before
3042         streaming out.
3043         * omp-expand.c (expand_oacc_for): Handle calls to internal
3044         functions like regular functions.
3045
3046 2018-04-12  Richard Biener  <rguenther@suse.de>
3047
3048         PR lto/85371
3049         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3050         for the early LTO debug to properly generate references to it
3051         during DIE emission.  Do not re-use that for the skeleton for
3052         split-dwarf.
3053         (dwarf2out_early_finish): Likewise.
3054
3055 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
3056
3057         PR target/85328
3058         * config/i386/sse.md
3059         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3060         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3061         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3062         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3063         and output is a reg, avoid creating invalid lowpart subreg, but
3064         instead split into a 512-bit move.  Don't split if not AVX512VL,
3065         input is xmm16+ reg and output is a mem.
3066         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3067         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3068         xmm16+ reg and output is a mem.
3069
3070 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
3071
3072         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3073         also for flag_dwarf2_cfi_asm.
3074
3075 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
3076
3077         PR rtl-optimization/85342
3078         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3079         a bool scalar var inside of the loop instead.  Don't try to update
3080         recog_data.operand after failed apply_change_group.
3081
3082 2018-04-12  Tom de Vries  <tom@codesourcery.com>
3083
3084         PR target/85296
3085         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3086         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
3087         array with flexible array member as array without given dimension.
3088         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3089         argument for undefined param to true.
3090
3091 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
3092
3093         PR target/85321
3094         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3095         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3096         from PowerPC section.
3097         * config/rs6000/sysv4.opt (mcall-): Improve help text.
3098         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3099         help text that is too long.
3100         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3101         help text that is too long.
3102         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3103         help text that is too long.
3104
3105 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
3106
3107         * config/alpha/alpha.md (stack_probe_internal): Rename
3108         from "probe_stack".  Update all callers.
3109
3110 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
3111
3112         PR rtl-optimization/84566
3113         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3114         sched_macro_fuse_insns.
3115
3116 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
3117
3118         PR target/84301
3119         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3120         (compute_block_dependences): ... from here.
3121
3122 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
3123
3124         PR tree-optimization/85331
3125         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3126         from int to HOST_WIDE_INT.
3127
3128 2018-04-11  Martin Jambor  <mjambor@suse.cz>
3129
3130         PR ipa/84149
3131         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3132         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3133         not the same as the source val.
3134         (cgraph_edge_brings_value_p): New parameter.
3135         (gather_edges_for_value): Pass destination value to
3136         cgraph_edge_brings_value_p.
3137         (perhaps_add_new_callers): Likewise.
3138         (get_info_about_necessary_edges): Likewise and exclude values brought
3139         only by self-recursive edges.
3140         (create_specialized_node): Redirect only clones of self-calling edges.
3141         (+self_recursive_pass_through_p): New function.
3142         (find_more_scalar_values_for_callers_subset): Use it.
3143         (find_aggregate_values_for_callers_subset): Likewise.
3144         (known_aggs_to_agg_replacement_list): Removed.
3145         (decide_whether_version_node): Re-calculate known constants for all
3146         remaining context clones.
3147
3148 2018-04-11  Richard Biener  <rguenther@suse.de>
3149
3150         PR lto/85339
3151         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3152         from early DWARF output.
3153         (dwarf2out_early_finish): Output line info unconditionally into
3154         early DWARF and add reference to it.
3155
3156 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
3157
3158         PR target/85281
3159         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3160         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3161         other than V2DFmode using iptr mode attribute.
3162         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3163
3164 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
3165
3166         PR rtl-optimization/84659
3167         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3168
3169 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
3170
3171         PR debug/85302
3172         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3173         SIZEP is NULL.
3174         (output_loc_list): Pass address of a dummy size variable even in the
3175         locview handling loop.
3176         (index_location_lists): Add comment on why skip_loc_list_entry can't
3177         call size_of_locs.
3178
3179 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
3180
3181         PR target/85261
3182         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3183         into register.
3184
3185 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
3186
3187         PR target/85321
3188         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3189         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3190         and -mstring-compare-inline-limit.
3191
3192 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
3193
3194         PR target/85287
3195         * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3196         for stack clash protection in a register whenever we need it to be in
3197         a register.
3198
3199 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
3200
3201         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3202         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3203
3204 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
3205
3206         PR target/85321
3207         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3208         the help text.
3209         (mlong-double-): Ditto.
3210         * config/rs6000/sysv4.opt (msdata=): Ditto.
3211         (mtls-size=): Ditto.
3212
3213 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
3214
3215         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3216         erroneous entries for
3217         "vector int vec_ldl (int, long int *)", and
3218         "vector unsigned int vec_ldl (int, unsigned long int *)".
3219         Add comments and entries for
3220         "vector bool char vec_ldl (int, bool char *)",
3221         "vector bool short vec_ldl (int, bool short *)",
3222         "vector bool int vec_ldl (int, bool int *)",
3223         "vector bool long long vec_ldl (int, bool long long *)",
3224         "vector pixel vec_ldl (int, pixel *)",
3225         "vector long long vec_ldl (int, long long *)",
3226         "vector unsigned long long vec_ldl (int, unsigned long long *)".
3227         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3228         type tree bool_long_long_type_node and correct definition of
3229         bool_V2DI_type_node to make reference to this new type tree.
3230         (rs6000_mangle_type): Replace erroneous reference to
3231         bool_long_type_node with bool_long_long_type_node.
3232         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3233         comments to emphasize sign distinctions for char and int types and
3234         replace RS6000_BTI_bool_long constant with
3235         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
3236         use of RS6000_BTI_pixel.
3237         (bool_long_type_node): Remove this macro definition.
3238         (bool_long_long_type_node): New macro definition
3239
3240 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
3241
3242         PR rtl-optimization/85300
3243         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3244         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3245         simplify_unary_operation fails.
3246
3247 2018-04-10  Martin Liska  <mliska@suse.cz>
3248
3249         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3250         cgraph_edge and ipa_ref.
3251
3252 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
3253
3254         PR target/85177
3255         PR target/85255
3256         * config/i386/sse.md
3257         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3258         computation of the VEC_MERGE selector from mask.
3259         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3260         Fix decoding of the VEC_MERGE selector into mask.
3261
3262 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
3263
3264         PR tree-optimization/85286
3265         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3266
3267 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
3268
3269         * final.c (final_1): Set insn_last_address as well as
3270         insn_current_address.
3271
3272 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3273
3274         PR target/85173
3275         * explow.c (emit_stack_probe): Call validize_mem on memory location
3276         before passing it to gen_probe_stack.  Create address operand and
3277         legitimize it for the probe_stack_address case.
3278
3279 2018-04-09  Jan Hubicka  <jh@suse.cz>
3280
3281         PR lto/85078
3282         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3283         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3284         * tree.c (free_lang_data_in_type): Fix handling of binfos;
3285         walk basetypes.
3286         (free_lang_data): Rebuild type inheritance graph.
3287
3288 2018-04-09  Martin Sebor  <msebor@redhat.com>
3289
3290         * invoke.texi (-finline-small-functions): Mention other optimization
3291         options.
3292         (-findirect-inlining, -fpartial-inlining): Same.
3293         (-finline-functions-called-once): Same.
3294         (-freorder-blocks-and-partition): Same.
3295
3296 2018-04-09  Jan Hubicka  <jh@suse.cz>
3297
3298         PR rtl/84058
3299         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3300         jumps; choose last target that matches the criteria (i.e.
3301         no partition changes for non-crossing jumps).
3302         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3303         support for redirecting crossing jumps to non-crossing.
3304
3305 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
3306
3307         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3308         also for naked functions.
3309
3310 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
3311
3312         * config/arc/arc.md (add_shift): New pattern.
3313         (add_shift2): Likewise.
3314         (sub_shift): Likewise.
3315         (sub_shift_cmp0_noout): Likewise.
3316         (compare_si_ashiftsi): Likewise.
3317         (xbfu_cmp0_noout): New combine pattern.
3318         (xbfu_cmp0"): Likewise.
3319         (movsi_set_cc_insn): Place the predicable variant first.
3320         (commutative_binary_cmp0_noout): Remove clobber.
3321         (commutative_binary_cmp0): New pattern.
3322         (noncommutative_binary_cmp0): Likewise.
3323         (noncommutative_binary_cmp0_noout): Likewise.
3324         (noncommutative_binary_comparison_result_used): Removed.
3325         (rsub_cmp0): New pattern.
3326         (rsub_cmp0_noout): Likewise.
3327         (extzvsi): Changed, keep only meaningful variants.
3328         (SQH, SEZ): New iterators.
3329         (SQH_postfix): New mode attribute.
3330         (SEZ_prefix): New code attribute.
3331         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3332         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3333         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3334         of numerical value.
3335         (noncommutative_operator): Check the availability of barrel
3336         shifter option.
3337
3338 2018-04-09  Richard Biener  <rguenther@suse.de>
3339
3340         PR tree-optimization/85284
3341         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3342         Only use the niter constraining form of simple_iv when the exit
3343         is always executed.
3344
3345 2018-04-09  Tom de Vries  <tom@codesourcery.com>
3346
3347         PR target/84041
3348         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3349         (define_expand "*memory_barrier"): New define_expand.
3350         (define_insn "memory_barrier"): New insn.
3351
3352 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
3353
3354         PR rtl-optimization/80463
3355         PR rtl-optimization/83972
3356         PR rtl-optimization/83480
3357
3358         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3359         correct producer for the insn.
3360         (tidy_control_flow): Fixup seqnos in case of debug insns.
3361
3362 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
3363
3364         PR rtl-optimization/83913
3365
3366         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3367         different sched-times when merging exprs.
3368
3369 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
3370
3371         PR rtl-optimization/83962
3372
3373         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3374         tidy_fallthru_edge and tidy_control_flow.
3375
3376 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
3377
3378         PR rtl-optimization/83530
3379
3380         * sel-sched.c (force_next_insn): New global variable.
3381         (remove_insn_for_debug): When force_next_insn is true, also leave only
3382         next insn in the ready list.
3383         (sel_sched_region): When the region wasn't scheduled, make another pass
3384         over it with force_next_insn set to 1.
3385
3386 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
3387
3388         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3389         into tm_file.
3390         * config/nds32/constants.md (unspec_volatile_element): Add enum values
3391         for interrupt control.
3392         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3393         functions for interrupt control.
3394         * config/nds32/nds32-intrinsic.md: Likewise.
3395         * config/nds32/nds32_intrinsic.h: Likewise.
3396         * config/nds32/nds32.h (nds32_builtins): Likewise.
3397
3398 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
3399
3400         * config/nds32/nds32.c (nds32_init_machine_status,
3401         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3402         strict_aligned_p field.
3403         (nds32_expand_to_rtl_hook): New function.
3404         (TARGET_EXPAND_TO_RTL_HOOK): Define.
3405         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3406
3407 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
3408             Chung-Ju Wu  <jasonwucj@gmail.com>
3409
3410         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3411         * config/nds32/nds32-n7.md: New file.
3412         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3413         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3414         pipeline.
3415         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3416         * config/nds32/nds32.md (pipeline_model): Add n7.
3417         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3418         * config/nds32/pipelines.md: Include n7 settings.
3419
3420 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
3421             Chung-Ju Wu  <jasonwucj@gmail.com>
3422
3423         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3424         * config/nds32/nds32-e8.md: New file.
3425         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3426         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3427         pipeline.
3428         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3429         * config/nds32/nds32.md (pipeline_model): Add e8.
3430         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3431         * config/nds32/pipelines.md: Include e8 settings.
3432
3433 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
3434             Chung-Ju Wu  <jasonwucj@gmail.com>
3435
3436         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3437         * config/nds32/nds32-n8.md: New file.
3438         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3439         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3440         pipeline.
3441         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3442         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3443         * config/nds32/nds32.md (pipeline_model): Add n8.
3444         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3445         * config/nds32/pipelines.md: Include n8 settings.
3446
3447 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
3448             Chung-Ju Wu  <jasonwucj@gmail.com>
3449
3450         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3451         * config/nds32/nds32-n9-2r1w.md: New file.
3452         * config/nds32/nds32-n9-3r2w.md: New file.
3453         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3454         nds32_register_ports): New or modify for cpu n9.
3455         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3456         pipeline.
3457         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3458         * config/nds32/nds32-utils.c: New file.
3459         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3460         TARGET_MUL_SLOW): Define.
3461         * config/nds32/nds32.md (pipeline_model): New attribute.
3462         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3463         New options that support cpu n9.
3464         * config/nds32/pipelines.md: Include n9 settings.
3465         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3466
3467 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
3468
3469         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3470         information if necessary.
3471         (output_cond_branch_compare_zero): Likewise.
3472         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3473         (nds32_target_alignment): Refine for alignment.
3474         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3475         (FUNCTION_BOUNDARY): Modify.
3476         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3477         align case.
3478         * config/nds32/nds32.opt (malways-align, malign-functions): New.
3479
3480 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
3481
3482         * config/nds32/constants.md (unspec_volatile_element): Add values for
3483         TLB operation and data prefetch.
3484         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3485         functions for TLB operation and data prefetch.
3486         * config/nds32/nds32-intrinsic.md: Likewise.
3487         * config/nds32/nds32_intrinsic.h: Likewise.
3488         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3489         (nds32_print_operand): Likewise.
3490         * config/nds32/nds32.h (nds32_builtins): Likewise.
3491
3492 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
3493         Andrew Pinski <pinsika@gcc.gnu.org>
3494
3495         PR middle-end/82976
3496         * match.pd: Use constant_boolean_node of correct type instead of
3497         boolean_true_node or boolean_false_node for simplifying
3498         pointer comparisons to zero.
3499
3500 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
3501
3502         PR tree-optimization/80021
3503         * tree.c (verify_type_variant): Make error call in verify_variant_match
3504         translatable and remove final full stop.
3505
3506 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
3507
3508         * config/nds32/constants.md (unspec_volatile_element): Add
3509         UNSPEC_VOLATILE_EH_RETURN.
3510         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3511         nds32_output_stack_pop): Support dwarf exception handling process.
3512         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3513         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3514         exception handling process.
3515         (nds32_compute_stack_frame): Likewise.
3516         (nds32_return_addr_rtx): Likewise.
3517         (nds32_initial_elimination_offset): Likewise.
3518         (nds32_expand_prologue): Likewise.
3519         (nds32_expand_epilogue): Likewise.
3520         (nds32_dynamic_chain_address): New function.
3521         * config/nds32/nds32.h (machine_function): Add fields for dwarf
3522         exception handling.
3523         (DYNAMIC_CHAIN_ADDRESS): Define.
3524         (EH_RETURN_DATA_REGNO): Define.
3525         (EH_RETURN_STACKADJ_RTX): Define.
3526         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3527         patterns for dwarf exception handling.
3528
3529 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
3530
3531         * config/nds32/nds32.h: Clean up obsolete macros.
3532
3533 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
3534
3535         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3536         Add enum values for particular instructions.
3537         * config/nds32/nds32-intrinsic.c: Implementation of expanding
3538         particular intrinsic functions.
3539         * config/nds32/nds32-intrinsic.md: Likewise.
3540         * config/nds32/nds32_intrinsic.h: Likewise.
3541         * config/nds32/nds32.h (nds32_builtins): Likewise.
3542         * config/nds32/nds32.md (type): Add pbsad and pbsada.
3543         (btst, ave): New patterns for particular instructions.
3544
3545 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
3546
3547         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3548         Add enum values for atomic load/store and memory sync.
3549         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3550         and memory sync.
3551         * config/nds32/nds32-intrinsic.md: Likewise.
3552         * config/nds32/nds32_intrinsic.h: Likewise.
3553         * config/nds32/nds32.h (nds32_builtins): Likewise.
3554
3555 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
3556
3557         PR tree-optimization/85257
3558         * fold-const.c (native_encode_vector): If not all elts could fit
3559         and off is -1, return 0 rather than offset.
3560         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3561         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3562         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
3563         adjust buffer in native_interpret_expr call.
3564
3565 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
3566
3567         * config/nds32/constants.md (unspec_volatile_element): Add cache
3568         control enum values.
3569         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3570         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3571         * config/nds32/nds32.c (nds32_cctl_names): New.
3572         (nds32_print_operand): Handle cache control register names.
3573         * config/nds32/nds32.h (nds32_builtins): New enum values.
3574         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3575         macros.
3576         * config/nds32/nds32.md (type): Add mmu.
3577         * config/nds32/pipelines.md (simple_insn): Add mmu.
3578
3579 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
3580
3581         * config/nds32/nds32.md (type): Remove call.
3582         * config/nds32/pipelines.md (simple_insn): Likewise.
3583
3584 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
3585
3586         * config/nds32/constants.md (unspec_volatile_element): Add
3587         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3588         UNSPEC_VOLATILE_FMFCFG.
3589         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3590         description for fmfcfg and fmfcsr.
3591         (bdesc_1arg): Add fmtcsr.
3592         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3593         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3594         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3595         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3596         unspec_fmfcfg): New patterns.
3597         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3598         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3599         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3600         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3601         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3602         __nds32__fmfcfg): Define.
3603
3604 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
3605
3606         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3607         intrinsic register names.
3608         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3609         intrinsic register enum values and macros.
3610
3611 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
3612
3613         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3614         for load/store addressing form.
3615         (nds32_print_operand_address): Likewise.
3616
3617 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
3618
3619         PR target/85196
3620         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3621         based on LABEL_REF.  Remove useless assertion.
3622         (pic_address_needs_scratch): Fix formatting.
3623         (sparc_legitimize_pic_address): Minor tweaks.
3624         (sparc_delegitimize_address): Adjust assertion accordingly.
3625         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3626         into symbolic_operand.
3627         (movsi_high_pic_label_ref): Likewise.
3628         (movsi_lo_sum_pic_label_ref): Likewise.
3629         (movdi_pic_label_ref): Likewise.
3630         (movdi_high_pic_label_ref): Likewise.
3631         (movdi_lo_sum_pic_label_ref): Likewise.
3632
3633 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
3634
3635         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3636         custom LIB_SPEC setup.
3637
3638 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
3639             Kito Cheng  <kito.cheng@gmail.com>
3640
3641         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3642         * config/riscv/freebsd.h: New.
3643
3644 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
3645
3646         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3647         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3648         file.
3649
3650 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
3651             Kito Cheng  <kito.cheng@gmail.com>
3652
3653         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3654         nds32_output_call, nds32_symbol_binds_local_p): New functions.
3655         * config/nds32/nds32-protos.h (nds32_output_call,
3656         nds32_output_return): Declare.
3657         * config/nds32/nds32.md: Refine all the call and return patterns.
3658
3659 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
3660
3661         PR debug/85252
3662         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3663         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3664
3665         PR rtl-optimization/84872
3666         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3667         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3668         EDGE_CROSSING edge.
3669
3670 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
3671
3672         * expr.c (copy_blkmode_to_reg): Revert 254862.
3673         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3674
3675 2018-04-06  Richard Biener  <rguenther@suse.de>
3676
3677         PR middle-end/85244
3678         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3679         after seeing a component reference with an adjacent field.  Treat
3680         refs to arrays at struct end of external decls similar to
3681         refs to unconstrained commons.
3682
3683 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
3684
3685         PR sanitizer/85213
3686         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3687         look through SAVE_EXPRs with non-side-effects argument.  Adjust
3688         recursive calls.
3689         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3690         save_p here.
3691
3692 2018-04-06  Richard Biener  <rguenther@suse.de>
3693
3694         PR middle-end/85180
3695         * alias.c (find_base_term): New wrapper around find_base_term
3696         unwinding CSELIB_VAL_PTR changes.
3697         (find_base_term): Do not restore CSELIB_VAL_PTR during the
3698         recursion.
3699
3700 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
3701
3702         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3703         instructions.
3704         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3705         constant definitions.
3706         ("nop"): lr 0,0 -> nopr r0
3707         ("nop_lr0", "nop_lr1"): New insn definitions.
3708
3709 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
3710
3711         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3712         NDS32_V3PUSH_AVAILABLE_P macro.
3713
3714 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
3715             Chung-Ju Wu  <jasonwucj@gmail.com>
3716
3717         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3718         (nds32*-*-*): Add float and fpu_config into supported_defaults.
3719         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3720         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3721         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3722         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3723         * config/nds32/constraints.md: New constraints and checking for hard
3724         float configuration.
3725         * config/nds32/iterators.md: New mode iterator and attribute for hard
3726         float configuration.
3727         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3728         patterns.
3729         * config/nds32/nds32-fpu.md: New file.
3730         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3731         deal with hard float code generation.
3732         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3733         ARCH_V3S.
3734         (abi_type, float_reg_number): New enum type.
3735         * config/nds32/nds32-predicates.c: New predicates for hard float.
3736         * config/nds32/nds32-protos.h: Declare functions for hard float.
3737         * config/nds32/nds32.c: Implementation for hard float configuration.
3738         * config/nds32/nds32.h: Definitions for hard float configuration.
3739         * config/nds32/nds32.md: Include hard float machine description and
3740         modify patterns for hard float configuration.
3741         * config/nds32/nds32.opt: New options for hard float configuration.
3742         * config/nds32/predicates.md: New predicates for hard float
3743         configuration.
3744
3745 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
3746
3747         * common/config/nds32/nds32-common.c
3748         (nds32_option_optimization_table): Enable -mreleax-hint by default.
3749
3750 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
3751
3752         PR middle-end/85195
3753         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3754         CONSTRUCTOR_ELT (ctor, ...)->value.
3755
3756 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
3757
3758         PR target/85193
3759         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3760
3761 2018-04-05  Tom de Vries  <tom@codesourcery.com>
3762
3763         PR target/85204
3764         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3765         cond jump.
3766
3767 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
3768             Kito Cheng  <kito.cheng@gmail.com>
3769
3770         * config/nds32/constraints.md (U33): Fine-tune checking condition.
3771         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3772         * config/nds32/nds32.h (nds32_16bit_address_type): Add
3773         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3774
3775 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
3776             Kito Cheng  <kito.cheng@gmail.com>
3777
3778         * config/nds32/constraints.md (Ufe): New memory constraint.
3779         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3780         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3781         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3782         operands.
3783         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3784         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3785
3786 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
3787
3788         * config/nds32/nds32.md: Use optimize_size in the condition for
3789         alu-shift instructions.
3790
3791 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
3792
3793         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3794
3795 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
3796
3797         * config/nds32/nds32.md (negsi2): Refine pattern.
3798
3799 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
3800             Chung-Ju Wu  <jasonwucj@gmail.com>
3801
3802         * config/nds32/iterators.md (shift_rotate): New code iterator.
3803         (shift): New code attribute.
3804         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3805         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3806         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3807         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3808         bit-wise operations.
3809         (andsi3, *andsi3): Ditto.
3810         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3811         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3812         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3813         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3814         nds32_ior_operand, nds32_xor_operand): New predicates.
3815
3816 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
3817
3818         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3819         (addsi3, subsi3): ... this.
3820
3821 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
3822
3823         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3824
3825 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
3826
3827         * config/nds32/nds32.md: Adjust indention.
3828
3829 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
3830
3831         * config/nds32/nds32.md (feature): New attribute.
3832
3833 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
3834
3835         * config/nds32/nds32.md (subtype): New attribute.
3836
3837 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
3838
3839         PR target/85203
3840         * config/arm/arm-builtins.c (arm_expand_builtin): Change
3841         expansion to perform a bitwise AND of the argument followed by a
3842         boolean negation of the result.
3843
3844 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
3845
3846         PR rtl-optimization/84878
3847         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3848         the basic block.  Assert the use reference is not artificial and that
3849         it has an associated insn.
3850
3851 2018-04-04  Michael Matz  <matz@suse.de>
3852
3853         * builtins.c (compute_objsize): Pass correct operand
3854         to array_at_struct_end_p.
3855
3856 2018-04-04  Richard Biener  <rguenther@suse.de>
3857
3858         PR lto/85176
3859         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3860         from contexts for DINFO_LEVEL_TERSE and below.
3861
3862 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
3863
3864         * config/nds32/nds32-doubleword.md (move_<mode>): Require
3865         resiter_operand condition.
3866         * config/nds32/nds32.md (*move<mode>): Ditto.
3867
3868 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
3869             Monk Chiang  <sh.chiang04@gmail.com>
3870
3871         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3872
3873 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
3874
3875         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3876
3877 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
3878             Kito Cheng  <kito.cheng@gmail.com>
3879
3880         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3881         nds32_cond_code_str, output_cond_branch,
3882         output_cond_branch_compare_zero, nds32_expand_cbranch,
3883         nds32_expand_cstore, nds32_expand_movcc,
3884         nds32_output_cbranchsi4_equality_zero,
3885         nds32_output_cbranchsi4_equality_reg,
3886         nds32_output_cbranchsi4_equality_reg_or_const_int,
3887         nds32_output_cbranchsi4_greater_less_zero: New functions.
3888         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3889         nds32_expand_cstore, nds32_expand_movcc,
3890         nds32_output_cbranchsi4_equality_zero,
3891         nds32_output_cbranchsi4_equality_reg,
3892         nds32_output_cbranchsi4_equality_reg_or_const_int,
3893         nds32_output_cbranchsi4_greater_less_zero): Declare.
3894         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3895         nds32_rimm11s_operand): New predicates.
3896         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3897         * config/nds32/nds32.md: Rewrite all the branch and conditional move
3898         patterns.
3899
3900 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
3901
3902         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3903         * config/nds32/nds32.md: Ditto.
3904         * config/nds32/pipelines.md: Ditto.
3905
3906 2018-04-04  Richard Biener  <rguenther@suse.de>
3907
3908         PR tree-optimization/85168
3909         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3910         propagating abnormals.
3911
3912 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
3913
3914         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3915
3916 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
3917             Kito Cheng  <kito.cheng@gmail.com>
3918
3919         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3920         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3921         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3922         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3923         * config/nds32/nds32.md (sibcall_internal): New.
3924         (sibcall_register): Remove.
3925         (sibcall_immediate): Remove.
3926         (sibcall_value_internal): New.
3927         (sibcall_value_register): Remove.
3928         (sibcall_value_immediate): Remove.
3929         * config/nds32/predicates.md (nds32_general_register_operand): New.
3930         (nds32_call_address_operand): New.
3931
3932 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
3933
3934         PR rtl-optimization/85167
3935         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3936         bb_defs if *split_p, instead preinitialize it to NULL.
3937
3938         PR tree-optimization/85156
3939         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3940         evaluating the argument multiple times.
3941
3942 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
3943
3944         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3945         than vector.
3946         (_mm_cvtpd_ps): Likewise.
3947         (_mm_cvttpd_epi32): Likewise.
3948         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3949         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3950         vector, pixel, and bool following altivec.h include.
3951
3952 2018-04-03  Martin Sebor  <msebor@redhat.com>
3953
3954         * doc/extend.texi (Common Function Attributes): Clarify.
3955         (const attribute): Likewise.
3956         (pure attribute): Likewise.
3957
3958 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
3959
3960         PR target/85169
3961         * config/i386/i386.c (ix86_expand_vector_set): Use
3962         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
3963
3964 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
3965
3966         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3967         instructions when changing rounding bits to preserve precision bits
3968         in the x87 control word.
3969
3970 2018-04-03  Martin Liska  <mliska@suse.cz>
3971
3972         PR tree-optimization/82491
3973         * rtl.h (strip_offset_and_add): Replace += suboffset with
3974         poly_uint64 () + suboffset.
3975
3976 2018-03-29  Martin Liska  <mliska@suse.cz>
3977             Martin Jambor  <mjambor@suse.cz>
3978
3979         PR ipa/84947
3980         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3981         param_type is not an integral or pointer type.
3982
3983 2018-04-03  Richard Biener  <rguenther@suse.de>
3984
3985         * sese.h (recompute_all_dominators): Remove.
3986
3987 2018-04-02  Martin Sebor  <msebor@redhat.com>
3988
3989         * doc/invoke.texi (-Wrestrict): Fix typos.
3990
3991 2018-04-02  Jim Wilson  <jimw@sifive.com>
3992
3993         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3994         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3995         (<optab>di3, <optab>si3_extend): Likewise.
3996         (<optab>si3_mask, <optab>si3_mask_1): New.
3997         (<optab>di3_mask, <optab>di3_mask_1): New.
3998         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3999         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4000         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4001
4002 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
4003
4004         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4005         example.
4006
4007 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
4008
4009         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4010         (nds32_canonicalize_comparison): New function.
4011
4012 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
4013             Kito Cheng  <kito.cheng@gmail.com>
4014             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
4015
4016         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4017         * config/nds32/constants.md (unspec_volatile_element): Add
4018         UNSPEC_VOLATILE_RELAX_GROUP.
4019         * config/nds32/nds32-relax-opt.c: New file.
4020         * config/nds32/nds32-predicates.c
4021         (nds32_symbol_load_store_p): New function.
4022         * config/nds32/nds32-protos.h
4023         (nds32_symbol_load_store_p): Declare function.
4024         (make_pass_nds32_relax_opt): Declare new rtl pass function.
4025         * config/nds32/nds32.c
4026         (nds32_register_pass): New function to register pass.
4027         (nds32_register_passes): New function to register passes.
4028         * config/nds32/nds32.md (relax_group): New pattern.
4029         * config/nds32/nds32.opt (mrelax-hint): New option.
4030         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4031
4032 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
4033
4034         * config/nds32/t-nds32: Modify files dependency.
4035
4036 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
4037
4038         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4039         (PROFILE_HOOK): Define its implementation.
4040
4041 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
4042
4043         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4044         type and 32-bit size.
4045
4046 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
4047
4048         PR middle-end/85090
4049         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4050         (V_128_256): New mode iterator.
4051         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4052         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4053         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4054         of V.
4055         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4056         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4057
4058 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
4059
4060         PR target/83315
4061         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4062         NaN inputs correctly.
4063
4064 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
4065
4066         PR target/80546
4067         * config/rs6000/vsx.md (??r): New mode attribute.
4068         (*vsx_mov<mode>_64bit): Use it.
4069         (*vsx_mov<mode>_32bit): Likewise.
4070
4071 2018-03-30  Martin Sebor  <msebor@redhat.com>
4072
4073         PR tree-optimization/84818
4074         * builtins.c (check_access): Use warning_n.
4075
4076 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
4077
4078         PR target/83822
4079         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4080         condition.
4081         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4082         condition.
4083
4084 2018-03-30  Julia Koval  <julia.koval@intel.com>
4085
4086         PR target/84413
4087         * x86-tune.def (movx, partial_reg_dependency): Enable for
4088         m_SKYLAKE_AVX512.
4089
4090 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
4091
4092         PR inline-asm/84985
4093         * lra-constraints.c (process_alt_operands): Move setting
4094         this_alternative_matches below.
4095
4096 2018-03-29  Martin Liska  <mliska@suse.cz>
4097
4098         PR lto/84995.
4099         * doc/invoke.texi: Document how LTO works with debug info.
4100         Describe auto-load support of binutils.  Mention 'x86-64'
4101         as valid option value of -march option.
4102
4103 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
4104
4105         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4106
4107         PR c/85094
4108         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4109         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4110         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4111         checking.
4112
4113 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
4114
4115         PR target/84912
4116         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4117         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4118         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4119         for RS6000_BTM_POWERPC64.
4120         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4121         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4122         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4123         definition.
4124         (DIVDE): Use it.
4125         (DIVDEU): Likewise.
4126
4127 2018-03-28 Carl Love  <cel@us.ibm.com>
4128
4129         Revert
4130         2017-09-27  Carl Love  <cel@us.ibm.com>
4131
4132         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4133         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4134         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4135         fctiw instruction.
4136
4137 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4138
4139         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4140         instead of __vector bool.
4141         (_mm_max_pu8): Likewise.
4142         (_mm_min_pi16): Likewise.
4143
4144 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
4145
4146         PR target/84912
4147         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4148         (DIVWEUO): Likewise.
4149         (DIVDEO): Likewise.
4150         (DIVDEUO): Likewise.
4151         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4152         DIVWEUO and DIVDEUO.
4153         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4154         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4155         (div_extend): Likewise.
4156         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4157         builtin function.
4158         (__builtin_divweuo): Likewise.
4159         (__builtin_divdeo): Likewise.
4160         (__builtin_divdeuo): Likewise.
4161
4162 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
4163
4164         PR target/85095
4165         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4166         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4167
4168         PR tree-optimization/82004
4169         * gimple-match-head.c (optimize_pow_to_exp): New function.
4170         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4171         Don't fold to exp if optimize_pow_to_exp is false.
4172
4173 2018-03-28  Martin Liska  <mliska@suse.cz>
4174
4175         PR other/84819
4176         * calls.c (initialize_argument_information): Fix trailing space.
4177         * common.opt: Fix typo and provide better explanation for
4178         -fsanitize-coverage option.
4179         * config/i386/i386.opt: Fix typo.
4180
4181 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
4182             Martin Liska  <mliska@suse.cz>
4183
4184         PR sanitizer/85081
4185         * gimplify.c (asan_poison_variable): Don't do the check for
4186         gimplify_omp_ctxp here.
4187         (gimplify_decl_expr): Do it here.
4188         (gimplify_target_expr): Likewise.
4189
4190 2018-03-28  Martin Liska  <mliska@suse.cz>
4191
4192         PR target/84988
4193         * config/i386/i386.c (ix86_function_arg_advance): Do not call
4194         chkp_type_bounds_count if MPX is not enabled.
4195
4196 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
4197
4198         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4199
4200 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
4201
4202         PR target/84914
4203         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4204         function to create the function decl for complex long double
4205         multiply and divide for -mabi=ieeelongdouble.
4206         (init_float128_ieee): Call it.
4207
4208 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
4209
4210         PR target/85044
4211         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4212         -fcf-protection=branch -mibt.
4213         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4214
4215 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
4216
4217         PR target/81863
4218         * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4219
4220 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
4221
4222         PR target/85056
4223         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4224         extern array declarations.
4225
4226 2018-03-27  Richard Biener  <rguenther@suse.de>
4227
4228         PR middle-end/84067
4229         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4230         explicit single_use checks.
4231
4232 2018-03-27  Richard Biener  <rguenther@suse.de>
4233
4234         PR tree-optimization/85082
4235         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4236         Valueize the VUSE.
4237
4238 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
4239
4240         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4241         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4242         Turn on fasynchronous-unwind-tables and funwind-tables.
4243
4244 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
4245
4246         PR target/85073
4247         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4248         (*bmi_blsr_<mode>_ccz): Ditto.
4249
4250 2018-03-26  Tom de Vries  <tom@codesourcery.com>
4251
4252         PR tree-optimization/85063
4253         * omp-general.c (offloading_function_p): New function.  Factor out
4254         of ...
4255         * omp-offload.c (pass_omp_target_link::gate): ... here.
4256         * omp-general.h (offloading_function_p): Declare.
4257         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4258         with attribute omp declare target for offloading functions.
4259
4260 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
4261
4262         PR tree-optimization/84005
4263         * tree-data-ref.h (get_base_for_alignment): Declare.
4264         * tree-data-ref.c (get_base_for_alignment_1): New function.
4265         (get_base_for_alignment): Likewise.
4266         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4267         get_base_for_alignment to find a suitable base object, instead
4268         of always using drb->base_address.
4269
4270 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
4271
4272         PR inline-asm/85022
4273         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4274         known size by default.
4275
4276 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
4277
4278         PR inline-asm/85030
4279         * lra-constraints.c (process_alt_operands): Don't match BLKmode
4280         and non BLKmode operands.
4281
4282 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4283
4284         PR target/85026
4285         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4286         Clean up attributes.
4287
4288 2018-03-23  Richard Biener  <rguenther@suse.de>
4289
4290         PR debug/85020
4291         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4292         we are going to emit early debug for LTO.
4293
4294 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
4295
4296         PR inline-asm/85034
4297         * function.c (match_asm_constraints_1): Don't optimize if input
4298         doesn't satisfy general_operand predicate for output's mode.
4299
4300         PR inline-asm/85022
4301         * alias.c (write_dependence_p): Don't require for x_canonicalized
4302         non-VOIDmode if x has VOIDmode.
4303
4304         PR sanitizer/85029
4305         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4306         just don't try to optimize it rather than assert it never happens.
4307
4308 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
4309
4310         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4311         macro expansions for definition of ST_INTERNAL_<mode> and
4312         LD_INTERNAL_<mode> builtins.
4313         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4314         Remove prototype.
4315         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4316         function.
4317         (altivec_expand_st_builtin): Likewise.
4318         (altivec_expand_builtin): Remove calls to deleted functions.
4319         (rs6000_address_for_altivec): Delete this function.
4320         * config/rs6000/vector.md: Remove expands for
4321         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4322
4323 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
4324
4325         PR target/84826
4326         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4327         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4328         re-computing once computed.
4329         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4330         (arm_init_machine_status): Initialize
4331         machine->static_chain_stack_bytes.
4332
4333 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
4334
4335         PR target/84760
4336         * doc/extend.texi: Add four new prototypes for vec_ld.
4337         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4338         definitions for more logical presentation.
4339         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4340         entries for V1TI variants of __builtin_altivec_ld builtin.
4341         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4342         handling of V1TI variant of LVX icode pattern.
4343         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4344         (rs6000_gimple_fold_builtin): Likewise.
4345         (altivec_init_builtins): Add code to define
4346         __builtin_altivec_lvx_v1ti function.
4347
4348 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
4349
4350         PR inline-asm/84941
4351         * function.c (match_asm_constraints_1): Don't do the optimization
4352         if input isn't a REG, SUBREG, MEM or constant.
4353
4354 2018-03-22  Tom de Vries  <tom@codesourcery.com>
4355
4356         PR tree-optimization/84956
4357         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4358         bb_has_abnormal_pred.
4359
4360 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
4361
4362         PR sanitizer/85018
4363         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4364         DECL_INITIAL (decl) to decl at the end.
4365         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4366         adjust the comment.
4367
4368 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
4369
4370         * doc/extend.texi (__builtin_tgmath): Document when complex
4371         integer types are treated as _Complex _Float64.
4372
4373 2018-03-21  Tom de Vries  <tom@codesourcery.com>
4374
4375         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4376
4377 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
4378
4379         PR tree-optimization/84960
4380         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4381         if it is ENTRY block, move them into single succ of ENTRY in that case.
4382
4383 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
4384
4385         PR tree-optimization/84811
4386         * poly-int.h (poly_span_traits): Remove the T3 parameter and
4387         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4388         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4389         (known_subrange_p): Update accordingly.  Cast each value involved
4390         in the size comparison, rather than casting the result of the
4391         subtraction.
4392
4393 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
4394
4395         PR tree-optimization/84982
4396         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4397         by flipping the least significant bit rather than all bits from
4398         bitpos to bitpos + bitsize - 1.
4399
4400 2018-03-21  Nathan Sidwell  <nathan@acm.org>
4401
4402         * doc/extend.texi (Deprecated Features): Remove mention of
4403         long-deleted deprecations.
4404
4405 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4406
4407         PR jit/84288
4408         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4409         * configure: Regenerate.
4410
4411 2018-03-21  Tom de Vries  <tom@codesourcery.com>
4412
4413         PR tree-optimization/83126
4414         * tree-parloops.c (num_phis): New function.
4415         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4416
4417 2018-03-21  Nathan Sidwell  <nathan@acm.org>
4418
4419         * doc/extend.texi (Deprecated Features): Update deprecated flags,
4420         mention anon-struct/union members and trailing attributes.
4421
4422 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
4423
4424         PR tree-optimization/84969
4425         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4426         builtin memset partitions if they set different rhs values.
4427
4428 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
4429
4430         PR rtl-optimization/84989
4431         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4432         VEC_DUPLICATE with scalar result mode.
4433
4434 2018-03-21  Martin Liska  <mliska@suse.cz>
4435
4436         PR ipa/84963
4437         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4438         not intended return statement.
4439
4440 2018-03-21  Martin Liska  <mliska@suse.cz>
4441
4442         PR target/84988
4443         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4444         (chkp_find_bound_slots_1): Limit number of iterations.
4445
4446 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
4447
4448         PR target/84838
4449         * Minor grammar fixes for x86 options.
4450
4451 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
4452
4453         PR debug/84875
4454         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4455         holding REG_CFA_RESTORE notes, instead turn them into a USE.
4456
4457 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
4458
4459         PR target/83789
4460         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4461         (altivec_lvx_<mode>_1op): Likewise.
4462         (altivec_stvx_<mode>_2op): Likewise.
4463         (altivec_stvx_<mode>_1op): Likewise.
4464         (altivec_lvx_<VM2:mode>): New define_expand.
4465         (altivec_stvx_<VM2:mode>): Likewise.
4466         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4467         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4468         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4469         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4470         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4471         (rs6000_gen_lvx): Likewise.
4472         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4473         (altivec_expand_stv_builtin): Likewise.
4474         (altivec_expand_builtin): Likewise.
4475         * config/rs6000/vector.md: Likewise.
4476
4477 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4478
4479         PR target/82518
4480         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4481         BYTES_BIG_ENDIAN.
4482
4483 2018-03-20  Richard Biener  <rguenther@suse.de>
4484
4485         PR target/84986
4486         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4487         sign-conversions as zero, fall back to standard scalar_stmt
4488         cost for the rest.
4489
4490 2018-03-20  Martin Liska  <mliska@suse.cz>
4491
4492         PR ipa/84825
4493         * predict.c (rebuild_frequencies): Handle case when we have
4494         PROFILE_ABSENT, but flag_guess_branch_prob is false.
4495
4496 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
4497
4498         PR target/84990
4499         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4500         flag_section_anchors.
4501         * varasm.c (use_blocks_for_decl_p): Remove hack for
4502         dw2_force_const_mem.
4503
4504         PR target/84845
4505         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4506         to ...
4507         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
4508         be created, use lowpart_subreg of operands[0] rather than operands[0]
4509         itself.
4510         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4511         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4512         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4513         and n constraint instead of aarch64_shift_imm_di and Usd.
4514         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4515         (*aarch64_<optab>_reg_minus<mode>3): ... this.
4516
4517 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
4518
4519         PR target/82989
4520         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4521         to favor GPR over NEON registers.
4522         (<shift>di3_neon): Likewise.
4523
4524 2018-03-20  Tom de Vries  <tom@codesourcery.com>
4525
4526         PR target/84952
4527         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4528         (nvptx_process_pars): Emit bar.sync asap and alap.
4529
4530 2018-03-20  Tom de Vries  <tom@codesourcery.com>
4531
4532         PR target/84954
4533         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4534         seen_label if seen_label is already set.
4535
4536 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
4537
4538         PR target/84945
4539         * config/i386/i386.c (fold_builtin_cpu): For features above 31
4540         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4541         Use 1U instead of 1.  Formatting fixes.
4542
4543         PR c/84953
4544         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4545         instead of TREE_TYPE (s1) for the return value.
4546
4547 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
4548
4549         PR tree-optimization/84946
4550         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4551         bitsize + bitsize in poly_uint64 rather than poly_int64.
4552
4553         PR sanitizer/78651
4554         * dwarf2asm.c: Include fold-const.c.
4555         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4556         of decl rather than decl itself.
4557
4558         PR rtl-optimization/84643
4559         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4560
4561 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
4562
4563         PR sanitizer/78651
4564         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4565         calling assemble_variable.
4566
4567 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
4568
4569         PR target/81647
4570         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4571         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4572
4573 2018-03-19  Jim Wilson  <jimw@sifive.com>
4574
4575         PR bootstrap/84856
4576         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4577         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4578         (riscv_first_stack_step): Likewise.
4579         (riscv_option_override): Use STACK_BOUNDARY instead of
4580         MIN_STACK_BOUNDARY.
4581         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4582         MIN_STACK_BOUNDARY.
4583         (BIGGEST_ALIGNMENT): Set to 128.
4584         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4585         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4586         STACK_BOUNDARY.
4587
4588 2018-03-19  Richard Biener  <rguenther@suse.de>
4589
4590         PR tree-optimization/84933
4591         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4592         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4593
4594 2018-03-19  Richard Biener  <rguenther@suse.de>
4595
4596         PR tree-optimization/84859
4597         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4598         (cond_if_else_store_replacement): Perform sinking operation on
4599         single-store BBs regardless of MAX_STORES_TO_SINK setting.
4600         Generalize what a BB with a single eligible store is.
4601
4602 2018-03-19  Richard Biener  <rguenther@suse.de>
4603
4604         PR tree-optimization/84929
4605         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4606         chrec_is_positive against non-chrec arg.
4607
4608 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
4609
4610         PR target/84711
4611         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4612
4613 2018-03-18  Martin Liska  <mliska@suse.cz>
4614
4615         PR rtl-optimization/84635
4616         * regrename.c (build_def_use): Use matches_mode only when
4617         matches >= 0.
4618
4619 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
4620
4621         PR tree-optimization/84913
4622         * tree-vect-loop.c (vectorizable_reduction): Don't try to
4623         vectorize chains of COND_EXPRs.
4624
4625 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
4626
4627         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4628
4629 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
4630
4631         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4632
4633 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
4634
4635         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4636
4637 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
4638             Kito Cheng  <kito.cheng@gmail.com>
4639
4640         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4641         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4642         (nds32_adjust_reg_alloc_order): New function.
4643         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4644
4645 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
4646
4647         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4648         nds32_print_operand, nds32_print_operand_address): Use
4649         HOST_WIDE_INT_PRINT_DEC instead.
4650
4651 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
4652
4653         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4654
4655 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
4656
4657         PR target/84902
4658         * config/i386/i386.c (initial_ix86_tune_features,
4659         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4660         unsigned long long.
4661         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4662         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4663         rather than 1u << ix86_tune.  Formatting fix.
4664         (ix86_option_override_internal): Change ix86_arch_mask from
4665         unsigned int to unsigned HOST_WIDE_INT, initialize to
4666         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4667         (ix86_function_specific_restore): Likewise.
4668
4669 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
4670
4671         PR target/84899
4672         * postreload.c (reload_combine_recognize_pattern): Perform
4673         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4674         truncate_int_for_mode the result for the destination's mode.
4675
4676         PR c/84909
4677         * hsa-gen.c (mem_type_for_type): Fix comment typo.
4678         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4679         Likewise.
4680         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4681         Likewise.
4682
4683 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
4684
4685         PR target/84876
4686         * lra-assigns.c (lra_split_hard_reg_for): Don't use
4687         regno_allocno_class_array and sorted_pseudos.
4688         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4689         insns where regno is used.
4690
4691 2018-03-16  Martin Liska  <mliska@suse.cz>
4692
4693         PR ipa/84833
4694         * multiple_target.c (create_dispatcher_calls): Redirect
4695         reference in the symbol table.
4696
4697 2018-03-16  Martin Liska  <mliska@suse.cz>
4698
4699         PR ipa/84722
4700         * multiple_target.c (create_dispatcher_calls): Redirect also
4701         an alias.
4702
4703 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
4704
4705         PR c++/79937
4706         PR c++/82410
4707         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4708         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4709         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4710
4711 2018-03-16  Julia Koval  <julia.koval@intel.com>
4712
4713         * doc/invoke.texi (Skylake Server): Add CLWB.
4714         Cannonlake): Remove CLWB.
4715
4716 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
4717
4718         PR tree-optimization/84841
4719         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4720         1 << 3.
4721         (FLOAT_ONE_CONST_TYPE): Define.
4722         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4723         (sort_by_operand_rank): Put entries with higher constant_type last
4724         rather than first to match comments.
4725
4726 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
4727
4728         * config/nios2/nios2.md (movsi_internal): Fix thinko in
4729         split predicate.
4730
4731 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
4732
4733         PR c++/79085
4734         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4735         check and use address of target always.
4736
4737 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
4738
4739         PR target/84574
4740         * config/i386/i386.c (indirect_thunk_needed): Update comments.
4741         (indirect_thunk_bnd_needed): Likewise.
4742         (indirect_thunks_used): Likewise.
4743         (indirect_thunks_bnd_used): Likewise.
4744         (indirect_return_needed): New.
4745         (indirect_return_bnd_needed): Likewise.
4746         (output_indirect_thunk_function): Add a bool argument for
4747         function return.
4748         (output_indirect_thunk_function): Don't generate alias for
4749         function return thunk.
4750         (ix86_code_end): Call output_indirect_thunk_function to generate
4751         function return thunks.
4752         (ix86_output_function_return): Set indirect_return_bnd_needed
4753         and indirect_return_needed instead of indirect_thunk_bnd_needed
4754         and indirect_thunk_needed.
4755
4756 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
4757
4758         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4759         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4760         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4761
4762 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
4763             Paul Hua <paul.hua.gm@gmail.com>
4764
4765         PR c/84852
4766         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4767
4768 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
4769
4770         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4771         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4772         resp. SFmode cases.
4773
4774 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
4775
4776         PR target/84711
4777         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4778         instead of GET_MODE_SIZE when comparing Units.
4779
4780 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
4781
4782         PR target/68256
4783         * varasm.c (hash_section): Return an unchangeble hash value
4784         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4785         Return !aarch64_can_use_per_function_literal_pools_p ().
4786
4787 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
4788
4789         PR target/84860
4790         * optabs.c (emit_conditional_move): Pass address of cmode's copy
4791         rather than address of cmode as last argument to prepare_cmp_insn.
4792
4793 2018-03-15  Julia Koval  <julia.koval@intel.com>
4794
4795         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4796         F_AVX512VNNI, F_AVX512BITALG): New.
4797
4798 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
4799
4800         PR target/83451
4801         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
4802         insn for floating-point loads and stores.
4803
4804 2018-03-14  Carl Love  <cel@us.ibm.com>
4805
4806         * config/rs6000/rs6000-c.c: Add macro definitions for
4807         ALTIVEC_BUILTIN_VEC_PERMXOR.
4808         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4809         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4810         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4811         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4812         UNSPEC_VPERMXOR.
4813         * config/doc/extend.texi: Add prototypes for vec_permxor.
4814
4815 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
4816
4817         PR c/84852
4818         * diagnostic-show-locus.c (class layout_point): Convert m_line
4819         from int to linenum_type.
4820         (line_span::comparator): Use linenum "compare" function when
4821         comparing line numbers.
4822         (test_line_span): New function.
4823         (layout_range::contains_point): Convert param "row" from int to
4824         linenum_type.
4825         (layout_range::intersects_line_p): Likewise.
4826         (layout::will_show_line_p): Likewise.
4827         (layout::print_source_line): Likewise.
4828         (layout::should_print_annotation_line_p): Likewise.
4829         (layout::print_annotation_line): Likewise.
4830         (layout::print_leading_fixits): Likewise.
4831         (layout::annotation_line_showed_range_p): Likewise.
4832         (struct line_corrections): Likewise for field m_row.
4833         (line_corrections::line_corrections): Likewise for param "row".
4834         (layout::print_trailing_fixits): Likewise.
4835         (layout::get_state_at_point): Likewise.
4836         (layout::get_x_bound_for_row): Likewise.
4837         (layout::print_line): Likewise.
4838         (diagnostic_show_locus): Likewise for locals "last_line" and
4839         "row".
4840         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4841         * input.c (selftest::test_linenum_comparisons): New function.
4842         (selftest::input_c_tests): Call it.
4843         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4844         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4845         * selftest.h (ASSERT_GT): New macro.
4846         (ASSERT_GT_AT): New macro.
4847         (ASSERT_LT): New macro.
4848         (ASSERT_LT_AT): New macro.
4849
4850 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
4851
4852         PR rtl-optimization/84780
4853         * combine.c (distribute_links): Don't make a link based on pc_rtx.
4854
4855 2018-03-14  Martin Liska  <mliska@suse.cz>
4856
4857         * tree.c (record_node_allocation_statistics): Use
4858         get_stats_node_kind.
4859         (get_stats_node_kind): New function extracted from
4860         record_node_allocation_statistics.
4861         (free_node): Use get_stats_node_kind.
4862
4863 2018-03-14  Richard Biener  <rguenther@suse.de>
4864
4865         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4866         that the value-set of ANTIC_IN doesn't grow.
4867
4868         Revert
4869         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4870         member.
4871         (BB_VISITED_WITH_VISITED_SUCCS): New define.
4872         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4873
4874 2018-03-14  Julia Koval  <julia.koval@intel.com>
4875
4876         * config.gcc (icelake-client, icelake-server): New.
4877         (icelake): Remove.
4878         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4879         (initial_ix86_arch_features): Ditto.
4880         (PTA_SKYLAKE): Add SGX.
4881         (PTA_ICELAKE): Remove.
4882         (PTA_ICELAKE_CLIENT): New.
4883         (PTA_ICELAKE_SERVER): New.
4884         (ix86_option_override_internal): Split up icelake on icelake client and
4885         icelake server.
4886         (get_builtin_code_for_version): Ditto.
4887         (fold_builtin_cpu): Ditto.
4888         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4889         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4890         * config/i386/i386.h (processor_type): Ditto.
4891         * doc/invoke.texi: Ditto.
4892
4893 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
4894
4895         PR sanitizer/83392
4896         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4897         INTEGER_CST offset, add it together with bitpos / 8 and
4898         sign extend based on POINTER_SIZE.
4899
4900         PR target/84844
4901         Revert
4902         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
4903
4904         PR target/78090
4905         * config/i386/constraints.md (Yc): New register constraint.
4906         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4907         Use Yc constraint for alternative 2 of operand 0.  Remove
4908         preferred_for_speed attribute.
4909
4910 2018-03-14  Richard Biener  <rguenther@suse.de>
4911
4912         PR tree-optimization/84830
4913         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4914         with the old one to avoid oscillations.
4915
4916 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
4917
4918         PR target/83712
4919         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4920         pseudos.
4921         (assign_by_spills): Return a flag of reload assignment failure.
4922         Do not process the reload assignment failures.  Do not spill other
4923         reload pseudos if they has the same reg class.  Update n if
4924         necessary.
4925         (lra_assign): Add a return arg.  Set up from the result of
4926         assign_by_spills call.
4927         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4928         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
4929         usage_insns if it is not NULL.
4930         (spill_hard_reg_in_range): New function.
4931         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4932         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4933         function prototypes.
4934         (lra_assign): Change prototype.
4935         * lra.c (lra): Add code to deal with fails by splitting hard reg
4936         live ranges.
4937
4938 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
4939
4940         * config/riscv/riscv.opt (mrelax): New option.
4941         * config/riscv/riscv.c (riscv_file_start): Emit ".option
4942         "norelax" when riscv_mrelax is disabled.
4943         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4944
4945 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
4946
4947         PR target/84743
4948         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4949         reassociation for int modes.
4950
4951 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
4952
4953         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4954         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4955         for big-endian.
4956         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4957         * config/aarch64/aarch64-sve.md
4958         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4959         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4960         (*extend<mode><Vwide>2): Rename to...
4961         (aarch64_sve_extend<mode><Vwide>2): ...this.
4962         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4963         renaming the old pattern to...
4964         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
4965         unsigned packs.
4966         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4967         define_expand, renaming the old pattern to...
4968         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4969         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4970         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4971         account when deciding which SVE instruction the optab should use.
4972         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4973
4974 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
4975
4976         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4977         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4978         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4979         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4980         (tlsdesc_small_<mode>): Turn a define_expand and use
4981         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
4982         (tlsdesc_small_advsimd_<mode>): ...this.
4983         (tlsdesc_small_sve_<mode>): New pattern.
4984
4985 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
4986
4987         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4988         (UNSPEC_UMUL_HIGHPART): New constants.
4989         (MUL_HIGHPART): New int iteraor.
4990         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4991         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4992         define_expand.
4993         (*<su>mul<mode>3_highpart): New define_insn.
4994
4995 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
4996
4997         PR lto/84805
4998         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4999         incomplete types.
5000
5001 2018-03-13  Martin Liska  <mliska@suse.cz>
5002
5003         PR ipa/84658.
5004         * (sem_item_optimizer::sem_item_optimizer): Initialize new
5005         vector.
5006         (sem_item_optimizer::~sem_item_optimizer): Release it.
5007         (sem_item_optimizer::merge_classes): Register variable aliases.
5008         (sem_item_optimizer::fixup_pt_set): New function.
5009         (sem_item_optimizer::fixup_points_to_sets): Likewise.
5010         * ipa-icf.h: Declare new variables and functions.
5011
5012 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
5013
5014         PR middle-end/84834
5015         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5016         integer_pow2p@2 and test integer_pow2p in condition.
5017         (A < 0 ? C : 0): Similarly for @1.
5018
5019         PR middle-end/84831
5020         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5021         characters starting at p contain '\0' character, don't look beyond
5022         that.
5023
5024         PR target/84827
5025         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5026         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5027
5028         PR target/84828
5029         * reg-stack.c (change_stack): Change update_end var from int to
5030         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5031         also call set_block_for_insn on the newly added insns and rescan.
5032
5033         PR target/84786
5034         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5035         on the last operand.
5036
5037         PR c++/84704
5038         * tree.c (stabilize_reference_1): Return save_expr (e) for
5039         STATEMENT_LIST even if it doesn't have side-effects.
5040
5041 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
5042
5043         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5044
5045 2018-03-12  Renlin Li  <renlin.li@arm.com>
5046
5047         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5048         aarch64_output_scalar_simd_mov_immediate.
5049
5050 2018-03-12  Martin Sebor  <msebor@redhat.com>
5051
5052         PR tree-optimization/83456
5053         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5054         for perfectly overlapping calls to memcpy.
5055         (gimple_fold_builtin_memory_chk): Same.
5056         (gimple_fold_builtin_strcpy): Handle no-warning.
5057         (gimple_fold_builtin_stxcpy_chk): Same.
5058         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5059
5060 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
5061
5062         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5063         parameter.  Use it for SFmode.
5064         (rs6000_function_arg_advance_1): Adjust.
5065         (rs6000_function_arg): Adjust.
5066         (rs6000_gimplify_va_arg): Pass false for that new parameter.
5067
5068 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
5069
5070         PR rtl-optimization/84169
5071         PR rtl-optimization/84780
5072         * combine.c (can_combine_p): Check for a 2-insn combination whether
5073         the destination register is used between the two insns, too.
5074
5075 2018-03-12  Richard Biener  <rguenther@suse.de>
5076
5077         PR tree-optimization/84803
5078         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5079         for refs DR analysis didn't process.
5080
5081 2018-03-12  Richard Biener  <rguenther@suse.de>
5082
5083         PR tree-optimization/84777
5084         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5085         force-vectorize loops ignore whether we are optimizing for size.
5086
5087 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
5088
5089         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5090         (TARGET_MD_ASM_ADJUST): Define.
5091
5092 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
5093             Kito Cheng  <kito.cheng@gmail.com>
5094             Chung-Ju Wu  <jasonwucj@gmail.com>
5095
5096         * config/nds32/nds32.c (nds32_compute_stack_frame,
5097         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5098         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5099         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5100         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5101         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5102         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5103         * config/nds32/nds32.md (prologue, epilogue): Use macro
5104         NDS32_V3PUSH_AVAILABLE_P to do checking.
5105
5106 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
5107
5108         PR debug/58150
5109         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5110         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5111         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
5112         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
5113         addition of most attributes on !orig_type_die or the attribute not
5114         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5115
5116 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
5117             Chung-Ju Wu  <jasonwucj@gmail.com>
5118
5119         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5120         __NDS32_VH__ macro.
5121         * config/nds32/nds32.opt (mvh): New option.
5122
5123 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
5124             Chung-Ju Wu  <jasonwucj@gmail.com>
5125
5126         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5127         function.
5128         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5129         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5130         definition.
5131
5132 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
5133             Chung-Ju Wu  <jasonwucj@gmail.com>
5134
5135         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5136         function.
5137         * config/nds32/nds32-multiple.md (strlensi): New pattern.
5138         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5139
5140 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
5141             Kito Cheng  <kito.cheng@gmail.com>
5142             Chung-Ju Wu  <jasonwucj@gmail.com>
5143
5144         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5145         UNSPEC_FFMISM and UNSPEC_FLMISM.
5146         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5147         for ffb, ffmism and flmism.
5148         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5149         (unspec_ffmism): Ditto.
5150         (unspec_flmism): Ditto.
5151         (nds32_expand_builtin_impl): Check if string extension is available.
5152         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5153         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5154
5155 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
5156
5157         Reverting patch:
5158         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
5159
5160         PR target/83712
5161         * lra-assigns.c (assign_by_spills): Return a flag of reload
5162         assignment failure.  Do not process the reload assignment
5163         failures.  Do not spill other reload pseudos if they has the same
5164         reg class.
5165         (lra_assign): Add a return arg.  Set up from the result of
5166         assign_by_spills call.
5167         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5168         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
5169         usage_insns if it is not NULL.
5170         (spill_hard_reg_in_range): New function.
5171         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5172         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5173         function prototypes.
5174         (lra_assign): Change prototype.
5175         * lra.c (lra): Add code to deal with fails by splitting hard reg
5176         live ranges.
5177
5178 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
5179
5180         PR target/84807
5181         * config/i386/i386.opt: Replace Enforcment with Enforcement.
5182
5183 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
5184
5185         PR debug/84620
5186         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5187         (dw_val_node): Add val_symbolic_view.
5188         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5189         (symview_upper_bound): New.
5190         (new_line_info_table): Initialize symviews_since_reset.
5191         (dwarf2out_source_line): Count symviews_since_reset and set
5192         symview_upper_bound.
5193         (dw_val_equal_p): Handle symview.
5194         (add_AT_symview): New.
5195         (print_dw_val): Handle symview.
5196         (attr_checksum, attr_checksum_ordered): Likewise.
5197         (same_dw_val_p, size_of_die): Likewise.
5198         (value_format, output_die): Likewise.
5199         (add_high_low_attributes): Use add_AT_symview for entry_view.
5200         (dwarf2out_finish): Reset symview_upper_bound, clear
5201         zero_view_p.
5202
5203 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
5204
5205         PR target/83969
5206         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5207         Add strict argument and use it.
5208         (rs6000_split_multireg_move): Update for new strict argument.
5209         (mem_operand_gpr): Disallow all non-offsettable addresses.
5210         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5211
5212 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
5213
5214         PR target/84772
5215         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5216         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5217         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5218
5219         PR c++/84767
5220         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5221         decl, use remap_type if we want to use the type.
5222
5223 2018-03-09  Martin Sebor  <msebor@redhat.com>
5224
5225         PR tree-optimization/84526
5226         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5227         Remove dead code.
5228         (builtin_access::generic_overlap): Be prepared to handle non-array
5229         base objects.
5230
5231 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
5232
5233         PR rtl-optimization/84682
5234         * lra-constraints.c (process_address_1): Check is_address flag
5235         for address constraints.
5236         (process_alt_operands): Likewise.
5237         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5238         preprocess_constraints.
5239         * recog.h (preprocess_constraints): Add oploc parameter.
5240         Adjust callers.
5241         * recog.c (preprocess_constraints): Test address_operand for
5242         CT_ADDRESS constraints.
5243
5244 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
5245
5246         PR target/83712
5247         * lra-assigns.c (assign_by_spills): Return a flag of reload
5248         assignment failure.  Do not process the reload assignment
5249         failures.  Do not spill other reload pseudos if they has the same
5250         reg class.
5251         (lra_assign): Add a return arg.  Set up from the result of
5252         assign_by_spills call.
5253         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5254         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
5255         usage_insns if it is not NULL.
5256         (spill_hard_reg_in_range): New function.
5257         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5258         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5259         function prototypes.
5260         (lra_assign): Change prototype.
5261         * lra.c (lra): Add code to deal with fails by splitting hard reg
5262         live ranges.
5263
5264 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5265
5266         PR target/83193
5267         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5268         Accept complain bool parameter.  Only emit errors if it is true.
5269         (arm_parse_cpu_option_name): Likewise.
5270         (arm_target_thumb_only): Adjust callers of the above.
5271         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5272         prototype to take a default true bool parameter.
5273         (arm_parse_arch_option_name): Likewise.
5274
5275 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
5276             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
5277
5278         PR jit/64089
5279         PR jit/84288
5280         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5281         * configure: Regenerate.
5282         * configure.ac ("linker --version-script option"): New.
5283         ("linker soname option"): New.
5284
5285 2018-03-09  Richard Biener  <rguenther@suse.de>
5286
5287         PR tree-optimization/84775
5288         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5289         immediate uses of predicate stmts and mark them modified.
5290
5291         Revert
5292         PR tree-optimization/84178
5293         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5294         to caller.
5295         (version_loop_for_if_conversion): Delay update_ssa call.
5296         (tree_if_conversion): Delay update_ssa until after predicate
5297         insertion.
5298
5299 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
5300
5301         PR target/84763
5302         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5303         when the function accesses prior frames.
5304
5305 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
5306
5307         PR debug/84456
5308         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5309         gen_llsym, otherwise call maybe_gen_llsym.
5310
5311         PR inline-asm/84742
5312         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5313         has ',' character inside of it.
5314
5315 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5316
5317         PR target/84748
5318         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5319         as clobbering CC_REGNUM.
5320
5321 2018-03-08  Richard Biener  <rguenther@suse.de>
5322
5323         PR middle-end/84552
5324         * tree-scalar-evolution.c: Include tree-into-ssa.h.
5325         (follow_copies_to_constant): Do not follow SSA names registered
5326         for update.
5327
5328 2018-03-08  Richard Biener  <rguenther@suse.de>
5329
5330         PR tree-optimization/84178
5331         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5332         to caller.
5333         (version_loop_for_if_conversion): Delay update_ssa call.
5334         (tree_if_conversion): Delay update_ssa until after predicate
5335         insertion.
5336
5337 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
5338
5339         PR tree-optimization/84178
5340         * tree-if-conv.c (release_bb_predicate): Remove the
5341         the assertion that the stmts have NULL use_ops.
5342         Discard the statements, asserting that they haven't
5343         yet been added to a BB.
5344
5345 2018-03-08  Richard Biener  <rguenther@suse.de>
5346
5347         PR tree-optimization/84746
5348         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5349         (phi_translate): Pass in destination ANTIC_OUT set.
5350         (phi_translate_1): Likewise.  For a simplified result lookup
5351         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5352         (phi_translate_set): Adjust.
5353         (do_pre_regular_insertion): Likewise.
5354         (do_pre_partial_partial_insertion): Likewise.
5355
5356 2018-03-08  Martin Liska  <mliska@suse.cz>
5357
5358         PR gcov-profile/84735
5359         * doc/gcov.texi: Document usage of profile files.
5360         * gcov-io.h: Document changes in the format.
5361
5362 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
5363
5364         PR debug/84404
5365         PR debug/84408
5366         * dwarf2out.c (struct dw_line_info_table): Update comments for
5367         view == -1.
5368         (FORCE_RESET_NEXT_VIEW): New.
5369         (FORCE_RESETTING_VIEW_P): New.
5370         (RESETTING_VIEW_P): Check for -1 too.
5371         (ZERO_VIEW_P): Likewise.
5372         (new_line_info_table): Force-reset next view.
5373         (dwarf2out_begin_function): Likewise.
5374         (dwarf2out_source_line): Simplify zero_view_p initialization.
5375         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5376         view directly.  Omit view when omitting .loc at line 0.
5377
5378 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
5379
5380         PR tree-optimization/84740
5381         * tree-switch-conversion.c (process_switch): Call build_constructors
5382         only if info.phi_count is non-zero.
5383
5384         PR tree-optimization/84739
5385         * tree-tailcall.c (find_tail_calls): Check call arguments against
5386         DECL_ARGUMENTS (current_function_decl) rather than
5387         DECL_ARGUMENTS (func) when checking for tail recursion.
5388
5389 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
5390
5391         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5392         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5393         Volker Reichelt's entry and add entries for people that perform
5394         GCC fuzzy testing and report numerous bugs.
5395
5396 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
5397
5398         PR target/82411
5399         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5400         readonly data in sdata, if that is disabled.
5401         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5402         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5403         -mreadonly-in-sdata option.
5404
5405 2018-03-07  Martin Sebor  <msebor@redhat.com>
5406
5407         PR tree-optimization/84468
5408         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5409         basic block when looking for nul assignment.
5410
5411 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
5412
5413         PR target/84277
5414         * except.h (output_function_exception_table): Adjust prototype.
5415         * except.c (output_function_exception_table): Remove FNNAME parameter
5416         and add SECTION parameter.  Ouput one part of the table at a time.
5417         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5418         the first part of the exception table and emit unwind directives.
5419         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5420         (i386_pe_seh_cold_init): Likewise.
5421         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5422         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5423         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5424         (ix86_output_call_insn): Emit a nop in one more case for SEH.
5425         * config/i386/winnt.c: Include except.h.
5426         (struct seh_frame_state): Add reg_offset, after_prologue and
5427         in_cold_section fields.
5428         (i386_pe_seh_end_prologue): Set seh->after_prologue.
5429         (i386_pe_seh_cold_init): New function.
5430         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5431         to seh->in_cold_section.
5432         (seh_emit_push): Record the offset of the push.
5433         (seh_emit_save): Record the offet of the save.
5434         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5435         Test seh->after_prologue to disregard the epilogue.
5436         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5437         (i386_pe_end_cold_function): New function.
5438
5439 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
5440
5441         PR fortran/84565
5442         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5443         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5444
5445         PR c++/84704
5446         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5447         on tmp_var.
5448         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5449         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5450
5451         PR middle-end/84723
5452         * multiple_target.c: Include tree-inline.h and intl.h.
5453         (expand_target_clones): Diagnose and fail if node->definition and
5454         !tree_versionable_function_p (node->decl).
5455
5456 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
5457
5458         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5459         sprint_ul.
5460         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5461         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5462         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5463
5464 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
5465
5466         PR target/84710
5467         * combine.c (try_combine): Use reg_or_subregno instead of handling
5468         just paradoxical SUBREGs and REGs.
5469
5470 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
5471
5472          * config/arc/arc.c (arc_finalize_pic): Remove function.
5473          (arc_must_save_register): We use single base PIC register, remove
5474          checks to save/restore the PIC register.
5475          (arc_expand_prologue): Likewise.
5476          * config/arc/arc-protos.h (arc_set_default_type_attributes):
5477          Remove.
5478          (arc_verify_short): Likewise.
5479          (arc_attr_type): Likewise.
5480          * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5481          (walk_stores): Likewise.
5482          (arc_address_cost): Make it static.
5483          (arc_verify_short): Likewise.
5484          (branch_dest): Likewise.
5485          (arc_attr_type): Likewise.
5486          * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5487          (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5488          (arc_final_prescan_insn): Remove inserting the nops due to
5489          hardware hazards.  It is done in reorg step.
5490          (insn_length_variant_t): Remove.
5491          (insn_length_parameters_t): Likewise.
5492          (arc_insn_length_parameters): Likewise.
5493          (arc_get_insn_variants): Likewise.
5494          * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5495
5496 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
5497
5498         PR inline-asm/84683
5499         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5500         assertion failure.
5501
5502         PR tree-optimization/84687
5503         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5504         on new_node->decl.
5505         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5506
5507 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5508
5509         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5510         Rename to ppc_speculation_barrier.
5511         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5512         __builtin_ppc_speculation_barrier.
5513
5514 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
5515
5516         PR target/84700
5517         * combine.c (combine_simplify_rtx): Don't try to simplify if
5518         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5519         are equal to x.
5520
5521 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
5522
5523         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5524         to 32 bytes when compiling for POWER9.
5525
5526 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
5527
5528         PR target/84564
5529         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5530         regparm >= 3 with no arg reg available also for calls with
5531         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
5532
5533         PR target/84524
5534         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5535         orig,vex.
5536         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
5537
5538 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
5539
5540         PR target/84264
5541         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5542
5543 2018-03-05  Richard Biener  <rguenther@suse.de>
5544
5545         PR tree-optimization/84486
5546         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5547         When inserting a __builtin_assume_aligned call set the LHS
5548         SSA name alignment info accordingly.
5549
5550 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
5551
5552         PR tree-optimization/84114
5553         * config/aarch64/aarch64.c (aarch64_reassociation_width)
5554         Avoid reassociation of FLOAT_MODE addition.
5555
5556 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
5557
5558         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5559         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5560         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5561         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5562         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5563         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5564         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5565         and -mwbnoinvd.
5566         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5567         __builtin_ia32_wbinvd): New builtins.
5568         (SPECIAL_ARGS2): New.
5569         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5570         (SPECIAL_ARGS2): New.
5571         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5572         (ix86_valid_target_attribute_inner_p): Ditto.
5573         (ix86_init_mmx_sse_builtins): Add special_args2.
5574         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5575         TARGET_WBNOINVD_P): New.
5576         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5577         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5578         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5579         * config/i386/immintrin.h (_wbinvd): New intrinsic.
5580         * config/i386/pconfigintrin.h: New file.
5581         * config/i386/wbnoinvdintrin.h: Ditto.
5582         * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5583         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5584
5585 2018-03-05  Richard Biener  <rguenther@suse.de>
5586
5587         PR tree-optimization/84670
5588         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5589         member.
5590         (BB_VISITED_WITH_VISITED_SUCCS): New define.
5591         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5592         (compute_antic_aux): Only assert the number of values in ANTIC_IN
5593         doesn't grow if all successors (recursively) were visited at least
5594         once.
5595
5596 2018-03-05  Richard Biener  <rguenther@suse.de>
5597
5598         PR tree-optimization/84650
5599         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5600         if executed in the loop pipeline.
5601
5602 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
5603
5604         * doc/configfiles.texi (Configuration Files): Move info about
5605         conditionalizing $target-protos.h to...
5606         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
5607         differs from $target-protos.h.
5608
5609 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
5610             Chung-Ju Wu  <jasonwucj@gmail.com>
5611
5612         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5613         * config/nds32/nds32-multiple.md (setmemsi): Define.
5614         * config/nds32/nds32-memory-manipulation.c
5615         (nds32_gen_dup_4_byte_to_word_value): New.
5616         (emit_setmem_word_loop): New.
5617         (emit_setmem_byte_loop): New.
5618         (nds32_expand_setmem_loop): New.
5619         (nds32_expand_setmem_loop_v3m): New.
5620         (nds32_expand_setmem_unroll): New.
5621         (nds32_expand_setmem): New.
5622
5623 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
5624             Chung-Ju Wu  <jasonwucj@gmail.com>
5625
5626         * config/nds32/nds32-memory-manipulation.c
5627         (nds32_emit_load_store): New.
5628         (nds32_emit_post_inc_load_store): New.
5629         (nds32_emit_mem_move): New.
5630         (nds32_emit_mem_move_block): New.
5631         (nds32_expand_movmemsi_loop_unknown_size): New.
5632         (nds32_expand_movmemsi_loop_known_size): New.
5633         (nds32_expand_movmemsi_loop): New.
5634         (nds32_expand_movmemsi_unroll): New.
5635         (nds32_expand_movmemqi): Rename ...
5636         (nds32_expand_movmemsi): ... to this.
5637         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5638         (movmemsi): ... to this.
5639         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5640         (nds32_expand_movmemsi): ... to this.
5641
5642 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
5643             Monk Chiang  <sh.chiang04@gmail.com>
5644             Chung-Ju Wu  <jasonwucj@gmail.com>
5645
5646         * config/nds32/nds32-protos.h
5647         (nds32_expand_load_multiple): New arguments.
5648         (nds32_expand_store_multiple): Ditto.
5649         (nds32_valid_multiple_load_store): Rename ...
5650         (nds32_valid_multiple_load_store_p): ... to this.
5651         * config/nds32/nds32-memory-manipulation.c
5652         (nds32_expand_load_multiple): Refine implementation.
5653         (nds32_expand_store_multiple): Ditto.
5654         * config/nds32/nds32-multiple.md
5655         (load_multiple): Update nds32_expand_load_multiple interface.
5656         (store_multiple): Update nds32_expand_store_multiple interface.
5657         * config/nds32/nds32-predicates.c
5658         (nds32_valid_multiple_load_store): Rename ...
5659         (nds32_valid_multiple_load_store_p): ... to this and refine
5660         implementation.
5661         * config/nds32/predicates.md
5662         (nds32_load_multiple_and_update_address_operation): New predicate.
5663         (nds32_store_multiple_and_update_address_operation): New predicate.
5664
5665 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
5666             Chung-Ju Wu  <jasonwucj@gmail.com>
5667
5668         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5669         (combo): New attribute.
5670         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5671
5672 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
5673
5674         * config/nds32/nds32.opt: Change -mcmodel= default value.
5675
5676 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
5677             Monk Chiang  <sh.chiang04@gmail.com>
5678             Chung-Ju Wu  <jasonwucj@gmail.com>
5679
5680         * config/nds32/constants.md (unspec_element): New enum.
5681         * config/nds32/constraints.md (Umw): New constraint.
5682         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5683         * config/nds32/nds32-intrinsic.md: Likewise.
5684         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5685         (nds32_valid_smw_lwm_base_p): New.
5686         (nds32_output_smw_single_word): New.
5687         (nds32_output_lmw_single_word): New.
5688         (nds32_expand_unaligned_load): New.
5689         (nds32_expand_unaligned_store): New.
5690         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5691         (nds32_output_smw_single_word): Declare.
5692         (nds32_output_lmw_single_word): Declare.
5693         (nds32_expand_unaligned_load): Declare.
5694         (nds32_expand_unaligned_store): Declare.
5695         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5696         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5697         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5698         NDS32_BUILTIN_UASTORE_DW.
5699         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5700         predicate.
5701
5702 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
5703             Kito Cheng  <kito.cheng@gmail.com>
5704             Chung-Ju Wu  <jasonwucj@gmail.com>
5705
5706         * config/nds32/nds32-intrinsic.c
5707         (nds32_expand_builtin_null_ftype_reg): Delete.
5708         (nds32_expand_builtin_reg_ftype_imm): Ditto.
5709         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5710         (nds32_read_argument): New.
5711         (nds32_legitimize_target): Ditto.
5712         (nds32_legitimize_argument): Ditto.
5713         (nds32_check_constant_argument): Ditto.
5714         (nds32_expand_unop_builtin): Ditto.
5715         (nds32_expand_unopimm_builtin): Ditto.
5716         (nds32_expand_binop_builtin): Ditto.
5717         (nds32_builtin_decl_impl): Ditto.
5718         (builtin_description): Ditto.
5719         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5720         (nds32_init_builtins_impl): Ditto.
5721         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5722         (nds32_builtin_decl): New.
5723         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5724         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5725
5726 2018-03-02  Jeff Law  <law@redhat.com>
5727
5728         * reorg.c (stop_search_p): Handle DEBUG_INSN.
5729         (redundant_insn, fill_simple_delay_slots): Likewise.
5730         (fill_slots_from_thread): Likewise.
5731         * resource.c (mark_referenced_resources): Likewise.
5732         (mark_set_resources, find_dead_or_set_registers): Likewise.
5733
5734 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
5735
5736         * substring-locations.h (format_warning_va): Formatting fix for
5737         ATTRIBUTE_GCC_DIAG.
5738         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5739         argument.
5740         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5741         * substring-locations.c: Include intl.h.
5742         (format_warning_va): Turned into small wrapper around
5743         format_warning_n_va, renamed to ...
5744         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5745         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5746         use ngettext.
5747         (format_warning_at_substring_n): New function.
5748         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5749         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
5750         format_warning_at_substring with just a shorter name instead of
5751         const function pointer.
5752         (fmtwarn_n): New function.
5753         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5754         appropriate, get rid of all the fmtstr temporaries, move conditionals
5755         with G_() wrapped string literals directly into fmtwarn arguments,
5756         cast dir.len to (int), formatting fixes.
5757
5758 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
5759
5760         * doc/invoke.texi: Remove "Cilk Plus" references.
5761
5762 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
5763             Richard Biener  <rguenther@suse.de>
5764
5765         PR ipa/84628
5766         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5767         for error or warning attributes if CALL_FROM_THUNK_P is set.
5768         Formatting fixes.
5769
5770 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
5771
5772         PR target/56540
5773         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5774         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5775
5776         PR target/56540
5777         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5778         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5779
5780         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5781         instead of -1U in last predictors element's probability member.
5782
5783 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
5784
5785         PR ipa/83983
5786         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5787         arguments if they are comparable.
5788
5789 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
5790
5791         PR tree-optimization/84634
5792         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5793         masks and masked_loop_p with a single loop_masks, making sure it's
5794         null for bb vectorization.
5795
5796 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
5797
5798         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5799         (vect_analyze_data_ref_access): Use loop->safe_len rather than
5800         loop->force_vectorize to check whether there is no alias.
5801
5802 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
5803
5804         PR target/84614
5805         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5806         prototypes.
5807         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5808         comments.
5809         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5810         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5811         instead of a loop around prev_real_insn.
5812         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5813         prev_real_insn.
5814
5815         PR inline-asm/84625
5816         * config/i386/i386.c (ix86_print_operand): Use conditional
5817         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5818         zero vector.
5819
5820 2018-03-02  Richard Biener  <rguenther@suse.de>
5821
5822         PR tree-optimization/84427
5823         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5824         (bitmap_set_subtract_values): Rewrite to handle multiple
5825         exprs per value.
5826         (clean): Likewise.
5827         (prune_clobbered_mems): Likewise.
5828         (phi_translate): Take edge instead of pred/phiblock.
5829         (phi_translate_1): Likewise.
5830         (phi_translate_set): Likewise.  Insert all translated
5831         exprs for a value into the set, keeping possibly multiple
5832         expressions per value.
5833         (compute_antic_aux): Adjust for phi_translate changes.
5834         When intersecting union the expressions and prune those
5835         not in the final value set, keeping possibly multiple
5836         expressions per value.  Do not use value-insertion
5837         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5838         all expressions.  Add verification that the value-sets
5839         only shrink during iteration.
5840         (compute_partial_antic_aux): Adjust for the phi_translate changes.
5841         (do_pre_regular_insertion): Likewise.
5842         (do_pre_partial_partial_insertion): Likewise.
5843
5844 2018-03-02  Richard Biener  <rguenther@suse.de>
5845
5846         PR target/82005
5847         * config/darwin.c (saved_debug_info_level): New static global.
5848         (darwin_asm_lto_start): Disable debug info generation for LTO out.
5849         (darwin_asm_lto_end): Restore debug info generation settings.
5850
5851 2018-03-01  Martin Liska  <mliska@suse.cz>
5852
5853         PR sanitizer/82484
5854         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5855         volatile arguments.
5856
5857 2018-03-01  Richard Biener  <rguenther@suse.de>
5858
5859         PR debug/84645
5860         * dwarf2out.c (gen_variable_die): Properly handle late VLA
5861         type annotation with LTO when debug was disabled at compile-time.
5862
5863 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
5864
5865         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5866         XINT with INTVAL.
5867         (mips_final_postscan_insn): Likewise.
5868
5869 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
5870
5871         PR rtl-optimization/84528
5872         * alias.c (init_alias_target): Add commentary.
5873         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5874         a unique base value if the frame pointer is not eliminated
5875         to the stack pointer.
5876
5877 2018-03-01  Tom de Vries  <tom@codesourcery.com>
5878
5879         PR rtl-optimization/83327
5880         * lra-int.h (hard_regs_spilled_into): Declare.
5881         * lra.c (hard_regs_spilled_into): Define.
5882         (init_reg_info): Init hard_regs_spilled_into.
5883         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5884         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5885         (process_bb_lives): Handle hard_regs_spilled_into.
5886         (lra_create_live_ranges_1): Before doing liveness propagation, clear
5887         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5888
5889 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
5890
5891         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5892         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5893         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5894         * config/rs6000/aix72.h: New file.
5895
5896 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
5897
5898         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5899         instead of warning_at with conditional singular and plural messages
5900         where possible.
5901
5902         PR target/52991
5903         * stor-layout.c (update_alignment_for_field): For
5904         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5905         && !DECL_PACKED (field), do the alignment update, just use
5906         only desired_align instead of MAX (type_align, desired_align)
5907         as the alignment.
5908         (place_field): Don't do known_align < desired_align handling
5909         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5910         is non-NULL, instead do it after rli->prev_field handling and
5911         only if not within a bitfield word.  For DECL_PACKED (field)
5912         use type_align of BITS_PER_UNIT.
5913
5914 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
5915
5916         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5917         superfluous parentheses and trailing spaces.
5918
5919 2018-02-28  Richard Biener  <rguenther@suse.de>
5920
5921         PR tree-optimization/84584
5922         * graphite-scop-detection.c (scop_detection::add_scop): Discard
5923         SCoPs with fake exit edge.
5924
5925 2018-02-28  Martin Liska  <mliska@suse.cz>
5926
5927         PR testsuite/84597
5928         * timevar.c (timer::print): Fix format to properly print 100%
5929         values.
5930
5931 2018-02-28  Richard Biener  <rguenther@suse.de>
5932
5933         PR middle-end/84607
5934         * genmatch.c (capture_info::walk_match): Do not mark
5935         captured expressions without operands as expr_p given
5936         they act more like predicates and should be subject to
5937         "lost tail" side-effect preserving.
5938
5939 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
5940
5941         PR rtl-optimization/81611
5942         * auto-inc-dec.c (attempt_change): Move dead note from
5943         mem_insn if it's the next use of regno
5944         (find_address): Take address use of reg holding
5945         non-incremented value.  Add parm to limit search to the named
5946         reg only.
5947         (merge_in_block): Attempt to use a mem insn that is the next
5948         use of the original regno.
5949
5950 2018-02-27  Martin Sebor  <msebor@redhat.com>
5951
5952         PR c++/83871
5953         * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5954         * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5955
5956 2018-02-27  Martin Sebor  <msebor@redhat.com>
5957
5958         PR translation/84207
5959         * diagnostic-core.h (warning_n, error_n, inform_n): Change
5960         n argument to unsigned HOST_WIDE_INT.
5961         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5962         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
5963         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5964         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5965
5966 2018-02-27  Richard Biener  <rguenther@suse.de>
5967
5968         PR tree-optimization/84512
5969         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5970         Do not use the estimate returned from record_stmt_cost for
5971         the scalar iteration cost but sum properly using add_stmt_cost.
5972
5973 2018-02-27  Richard Biener  <rguenther@suse.de>
5974
5975         PR tree-optimization/84466
5976         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5977         Adjust last change to less strictly validate use operands.
5978
5979 2018-02-27  Martin Liska  <mliska@suse.cz>
5980
5981         PR gcov-profile/84548
5982         * gcov.c (process_file): Allow partial overlap and consider it
5983         also as group functions.
5984         (output_lines): Properly calculate range of lines for a group.
5985
5986 2018-02-27  Martin Liska  <mliska@suse.cz>
5987
5988         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5989         'ggc' suffixes.  Change first column width.
5990         (timer::print): Fix formatting of the column.
5991
5992 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
5993
5994         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5995         preserve inline entry blocks for the sake of debug inline
5996         entry point markers alone.
5997         (remove_unused_locals): Suggest in comments a better place to
5998         force the preservation of inline entry blocks that are
5999         otherwise unused, but do not preserve them.
6000
6001 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
6002
6003         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6004
6005 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
6006
6007         PR target/84039
6008         * config/i386/constraints.md (Bs): Replace
6009         ix86_indirect_branch_register with
6010         TARGET_INDIRECT_BRANCH_REGISTER.
6011         (Bw): Likewise.
6012         * config/i386/i386.md (indirect_jump): Likewise.
6013         (tablejump): Likewise.
6014         (*sibcall_memory): Likewise.
6015         (*sibcall_value_memory): Likewise.
6016         Peepholes of indirect call and jump via memory: Likewise.
6017         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6018         (*sibcall_value_GOT_32): Likewise.
6019         * config/i386/predicates.md (indirect_branch_operand): Likewise.
6020         (GOT_memory_operand): Likewise.
6021         (call_insn_operand): Likewise.
6022         (sibcall_insn_operand): Likewise.
6023         (GOT32_symbol_operand): Likewise.
6024         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6025
6026 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
6027
6028         PR rtl-optimization/83496
6029         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6030         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
6031         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6032         redundant insn, if any.
6033         (relax_delay_slots): Likewise.
6034         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6035
6036 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
6037
6038         PR tree-optimization/83965
6039         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6040         that grouped statements are part of a reduction chain.  Return
6041         true if the statement is not marked as a reduction itself but
6042         is part of a group.
6043         (vect_recog_dot_prod_pattern): Don't check whether the statement
6044         is part of a group here.
6045         (vect_recog_sad_pattern): Likewise.
6046         (vect_recog_widen_sum_pattern): Likewise.
6047
6048 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
6049
6050         PR debug/84545
6051         * final.c (rest_of_clean_state): Also look for calls inside sequences.
6052
6053 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
6054
6055         PR target/84530
6056         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6057         the bool argument.
6058         (ix86_output_indirect_function_return): New prototype.
6059         (ix86_split_simple_return_pop_internal): Likewise.
6060         * config/i386/i386.c (indirect_return_via_cx): New.
6061         (indirect_return_via_cx_bnd): Likewise.
6062         (indirect_thunk_name): Handle return va CX_REG.
6063         (output_indirect_thunk_function): Create alias for
6064         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6065         (ix86_output_indirect_jmp): Remove the bool argument.
6066         (ix86_output_indirect_function_return): New function.
6067         (ix86_split_simple_return_pop_internal): Likewise.
6068         * config/i386/i386.md (*indirect_jump): Don't pass false
6069         to ix86_output_indirect_jmp.
6070         (*tablejump_1): Likewise.
6071         (simple_return_pop_internal): Change it to define_insn_and_split.
6072         Call ix86_split_simple_return_pop_internal to split it for
6073         -mfunction-return=.
6074         (simple_return_indirect_internal): Call
6075         ix86_output_indirect_function_return instead of
6076         ix86_output_indirect_jmp.
6077
6078 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
6079
6080         PR bootstrap/84405
6081         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6082         memset and value initialization afterwards.
6083
6084 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
6085
6086         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6087
6088 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
6089
6090         PR target/84521
6091         * common/config/aarch64/aarch64-common.c
6092         (aarch_option_optimization_table[]): Switch
6093         off fomit-frame-pointer
6094
6095 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
6096             Chung-Ju Wu  <jasonwucj@gmail.com>
6097
6098         * config/nds32/nds32-multiple.md (load_multiple): Disallow
6099         volatile memory.
6100         (store_multiple): Ditto.
6101
6102 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
6103
6104         * config.gcc: Add --with-cpu support for nds32 target.
6105         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6106         * config/nds32/nds32.opt: Add -mcpu= option.
6107
6108 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
6109
6110         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6111         isel=yes): Warn for these deprecated options.
6112
6113 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
6114
6115         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6116         ISA_2_5_MASKS_EMBEDDED.
6117
6118 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
6119
6120         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6121         p->max as pointers rather than using iterative_hash_expr.
6122
6123 2018-02-23  Carl Love  <cel@us.ibm.com>
6124
6125         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6126         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6127         BU_P8V_OVERLOAD_2.
6128         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6129         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6130         P8V_BUILTIN_VEC_VUNSIGNED2.
6131
6132 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
6133
6134         PR target/81572
6135         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6136         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6137         LRA_UNKNOWN_ALT.
6138         * lra-constraints.c (curr_insn_transform): Set up
6139         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
6140         LRA_UNKNOWN_ALT.
6141         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6142         * lra-eliminations.c (spill_pseudos): Ditto.
6143         (process_insn_for_elimination): Ditto.
6144         * lra-lives.c (reg_early_clobber_p): Use the new macros.
6145         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6146         LRA_NON_CLOBBERED_ALT.
6147
6148 2018-02-22  Martin Sebor  <msebor@redhat.com>
6149
6150         PR tree-optimization/84480
6151         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6152         to maybe_diag_stxncpy_trunc.  Call it.
6153         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6154         from gimple_fold_builtin_strcpy.  Print inlining stack.
6155         (handle_builtin_stxncpy): Print inlining stack.
6156         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6157
6158 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
6159
6160         PR target/84176
6161         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6162         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6163         and -fcheck-pointer-bounds are used together.
6164         (indirect_thunk_prefix): New enum.
6165         (indirect_thunk_need_prefix): New function.
6166         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
6167         "_nt" instead of "_bnd" for NOTRACK prefix.
6168         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6169         (output_indirect_thunk_function): Likewise.
6170         (): Likewise.
6171         (ix86_code_end): Update output_indirect_thunk_function calls.
6172         (ix86_output_indirect_branch_via_reg): Replace
6173         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6174         (ix86_output_indirect_branch_via_push): Likewise.
6175         (ix86_output_function_return): Likewise.
6176         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6177         incompatible with -fcf-protection=branch and
6178         -fcheck-pointer-bounds.
6179
6180 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
6181
6182         PR target/83335
6183         * config/aarch64/aarch64.c (aarch64_print_address_internal):
6184         Change gcc_assert call to output_operand_lossage.
6185
6186 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
6187
6188         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6189
6190 2018-02-22  DJ Delorie  <dj@redhat.com>
6191             Sebastian Perta  <sebastian.perta@renesas.com>
6192             Oleg Endo  <olegendo@gcc.gnu.org>
6193
6194         * config/rx/rx.c (rx_rtx_costs): New function.
6195         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6196
6197 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6198
6199         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6200
6201 2018-02-22  Martin Liska  <mliska@suse.cz>
6202
6203         PR driver/83193
6204         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6205         Add "native" as a possible value.
6206
6207 2018-02-22  Martin Liska  <mliska@suse.cz>
6208
6209         PR driver/83193
6210         * config/i386/i386.c (ix86_option_override_internal):
6211         Add "native" as a possible value for -march and -mtune.
6212
6213 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
6214
6215         PR target/84502
6216         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6217         to all type variants.
6218
6219         PR tree-optimization/84503
6220         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6221         width as info->bitpos + info->bitsize - start.
6222         (merged_store_group::merge_overlapping): Simplify width computation.
6223         (check_no_overlap): New function.
6224         (imm_store_chain_info::try_coalesce_bswap): Compute expected
6225         start + width and last_order of the group, fail if check_no_overlap
6226         fails.
6227         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6228         to group if check_no_overlap fails.
6229
6230 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
6231
6232         * config/rs6000/altivec.md: Delete contraint arguments to
6233         define_expand, define_split, and define_peephole2, and in
6234         define_insn_and_split if always unused.
6235         * config/rs6000/darwin.md: Ditto.
6236         * config/rs6000/dfp.md: Ditto.
6237         * config/rs6000/rs6000.md: Ditto.
6238         * config/rs6000/sync.md: Ditto.
6239         * config/rs6000/vector.md: Ditto.
6240         * config/rs6000/vsx.md: Ditto.
6241
6242 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
6243
6244         * config/rs6000/altivec.md: Write output control strings as braced
6245         blocks instead of double-quoted strings.
6246         * config/rs6000/darwin.md: Ditto.
6247         * config/rs6000/rs6000.md: Ditto.
6248         * config/rs6000/vector.md: Ditto.
6249         * config/rs6000/vsx.md: Ditto.
6250
6251 2018-02-21  Jason Merrill  <jason@redhat.com>
6252
6253         PR c++/84314 - ICE with templates and fastcall attribute.
6254         * attribs.c (build_type_attribute_qual_variant): Remove assert.
6255
6256 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
6257
6258         * ipa-cp.c (determine_versionability): Fix comment typos.
6259
6260 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
6261
6262         PR c/84229
6263         * ipa-cp.c (determine_versionability): Do not version functions caling
6264         va_arg_pack.
6265
6266 2018-02-21  Martin Liska  <mliska@suse.cz>
6267
6268         PR driver/83193
6269         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6270         Add "native" as a possible value.
6271         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
6272         the macro when native cpu detection is available.
6273
6274 2018-02-21  Martin Liska  <mliska@suse.cz>
6275
6276         PR driver/83193
6277         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6278         Add "native" as a possible value.
6279         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6280         when native cpu detection is available.
6281
6282 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
6283             Martin Sebor  <msebor@redhat.com>
6284
6285         PR tree-optimization/84478
6286         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6287         false.
6288         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6289         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6290         support which is conservatively correct, for 2 only stay conservative
6291         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
6292         argument to the 2 argument get_range_strlen, adjust 6 arg
6293         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6294         false.
6295         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6296         (gimple_fold_builtin_strlen): Pass true as last argument to
6297         get_range_strlen.
6298
6299 2018-02-20  Martin Sebor  <msebor@redhat.com>
6300
6301         PR middle-end/84095
6302         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6303         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
6304         (builtin_memref::builtin_memref): Factor out parts into
6305         set_base_and_offset and call it.
6306
6307 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
6308
6309         PR middle-end/84406
6310         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6311         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6312         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
6313         search at the associated MODE_INT.
6314
6315 2018-02-20  Jeff Law  <law@redhat.com>
6316
6317         PR middle-end/82123
6318         PR tree-optimization/81592
6319         PR middle-end/79257
6320         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6321         for range data rather than using global data.
6322         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6323         range data rather than using global data.
6324         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6325         pass it to children as needed.
6326         (struct directive::fmtresult): Similarly.
6327         (struct directive::set_width): Similarly.
6328         (struct directive::set_precision): Similarly.
6329         (format_integer, format_directive, parse_directive): Similarly.
6330         (format_none): Accept unnamed vr_values parameter.
6331         (format_percent, format_floating, format_character): Similarly.
6332         (format_string, format_plain): Similarly.
6333         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6334         the EVRP range analyzer for range data rather than using global data.
6335         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6336         gimple-ssa-evrp-analyze.h
6337         (class sprintf_dom_walker): Add after_dom_children member function.
6338         Add evrp_range_analyzer member.
6339         (sprintf_dom_walker::before_dom_children): Call into the EVRP
6340         range analyzer as needed.
6341         (sprintf_dom_walker::after_dom_children): New member function.
6342         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6343         if not optimizing.
6344         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6345         (evrp_range_analyzer::pop_to_marker): Likewise.
6346
6347 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
6348
6349         PR tree-optimization/84419
6350         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6351         with the required type if its current type is compatible but
6352         different.
6353
6354 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
6355
6356         PR middle-end/82004
6357         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6358         after vectorization.
6359
6360 2018-02-20  Martin Liska  <mliska@suse.cz>
6361
6362         PR driver/83193
6363         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6364         possible values if we don't have a hint.
6365
6366 2018-02-20  Martin Liska  <mliska@suse.cz>
6367
6368         PR c/84310
6369         PR target/79747
6370         * final.c (shorten_branches): Build align_tab array with one
6371         more element.
6372         * opts.c (finish_options): Add alignment option limit check.
6373         (MAX_CODE_ALIGN): Likewise.
6374         (MAX_CODE_ALIGN_VALUE): Likewise.
6375         * doc/invoke.texi: Document maximum allowed option value for
6376         all -falign-* options.
6377
6378 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
6379
6380         PR target/84146
6381         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6382         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6383         * var-tracking.c (emit_note_insn_var_location): Remove all references
6384         to NOTE_INSN_CALL_ARG_LOCATION.
6385         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6386         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6387         Use copy_rtx_if_shared.
6388         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6389         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6390         (dwarf2out_var_location): Remove handling of
6391         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6392         on call_insn.
6393         * final.c (final_scan_insn): Remove all references to
6394         NOTE_INSN_CALL_ARG_LOCATION.
6395         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
6396         before dumping final insns.
6397         * except.c (emit_note_eh_region_end): Remove all references to
6398         NOTE_INSN_CALL_ARG_LOCATION.
6399         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6400         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6401         * config/arc/arc.c (hwloop_optimize): Likewise.
6402         * config/arm/arm.c (create_fix_barrier): Likewise.
6403         * config/s390/s390.c (s390_chunkify_start): Likewise.
6404         * config/sh/sh.c (find_barrier): Likewise.
6405         * config/i386/i386.c (rest_of_insert_endbranch,
6406         ix86_seh_fixup_eh_fallthru): Likewise.
6407         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6408         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6409         * config/frv/frv.c (frv_function_prologue): Likewise.
6410         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
6411         reg note.
6412         (note_outside_basic_block_p): Remove all references to
6413         NOTE_INSN_CALL_ARG_LOCATION.
6414         * gengtype.c (adjust_field_rtx_def): Likewise.
6415         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6416         Likewise.
6417         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6418         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6419
6420         PR c++/84444
6421         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6422         is ADDR_EXPR.
6423
6424         PR tree-optimization/84452
6425         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6426         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6427         is NULL.
6428
6429 2018-02-19  Martin Liska  <mliska@suse.cz>
6430
6431         PR sanitizer/82183
6432         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6433
6434 2018-02-19  Martin Liska  <mliska@suse.cz>
6435             Richard Sandiford  <richard.sandiford@linaro.org>
6436
6437         PR tree-optimization/82491
6438         * gimple-fold.c (get_base_constructor): Make earlier bail out
6439         to prevent ubsan.
6440
6441 2018-02-19  Carl Love  <cel@us.ibm.com>
6442
6443         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6444         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6445         BU_P8V_OVERLOAD_1.
6446         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6447         P8V_BUILTIN_VEC_NEG.
6448
6449 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
6450
6451         * config/rl78/rl78.md (movdf): New define expand.
6452
6453 2018-02-19  Martin Liska  <mliska@suse.cz>
6454
6455         PR other/80589
6456         * doc/invoke.texi: Fix typo.
6457         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6458
6459 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
6460
6461         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6462         handle rs6000_single_float and rs6000_double_float specially for
6463         e500 family CPUs.
6464
6465 2018-02-16  Jeff Law  <law@redhat.com>
6466
6467         * config/rx/rx.c (add_pop_cfi_notes): New function.;
6468         (pop_regs): Use it.
6469
6470 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
6471
6472         PR ipa/84425
6473         * ipa-inline.c (inline_small_functions): Fix a typo.
6474
6475 2018-02-16  Nathan Sidwell  <nathan@acm.org>
6476
6477         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6478
6479 2018-02-16  Carl Love  <cel@us.ibm.com>
6480
6481         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6482         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6483         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6484         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6485         expansion to P8V_BUILTIN_VEC_FLOAT2.
6486
6487 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
6488
6489         PR rtl-optimization/70023
6490         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6491         src_regno into account.
6492
6493 2018-02-16  Carl Love  <cel@us.ibm.com>
6494
6495         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6496         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6497         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6498         * config/rs6000/rs6000.c: Remove case statements for
6499         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6500         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6501         and P9V_BUILTIN_VEC_VINSERT4B.
6502         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6503         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6504         * config/rs6000/vsx.md:
6505         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6506         vec_insert4b.
6507
6508 2018-02-16  Carl Love  <cel@us.ibm.com>
6509
6510         * config/rs6000/altivec.h: Add builtin names vec_extract4b
6511         vec_insert4b.
6512         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6513         definitions.
6514         * config/rs6000/rs6000-c.c: Add the definitions for
6515         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6516         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6517         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6518         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
6519         definition for insert4b and define insn *insert3b_internal.
6520         * doc/extend.texi: Add documentation for vec_extract4b.
6521
6522 2018-02-16  Nathan Sidwell  <nathan@acm.org>
6523
6524         * doc/extend.texi (Backwards Compatibility): Mention friend
6525         injection.  Note for-scope is deprecated.
6526         * doc/invoke.texi (-ffriend-injection): Deprecate.
6527
6528 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
6529
6530         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6531         that moved to I2, also allow destinations that are a paradoxical
6532         subreg (instead of a normal reg).
6533
6534 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
6535
6536         PR target/83831
6537         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6538         to QImode.
6539
6540 2018-02-16  Richard Biener  <rguenther@suse.de>
6541
6542         PR tree-optimization/84037
6543         PR tree-optimization/84016
6544         PR target/82862
6545         * config/i386/i386.c (ix86_builtin_vectorization_cost):
6546         Adjust vec_construct for the fact we need additional higher latency
6547         128bit inserts for AVX256 and AVX512 vector builds.
6548         (ix86_add_stmt_cost): Scale vector construction cost for
6549         elementwise loads.
6550
6551 2018-02-16  Richard Biener  <rguenther@suse.de>
6552
6553         PR tree-optimization/84417
6554         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6555         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6556         (non_rewritable_lvalue_p): Likewise, use poly-ints.
6557
6558 2018-02-16  Martin Liska  <mliska@suse.cz>
6559
6560         PR sanitizer/84307
6561         * internal-fn.def (ASAN_CHECK): Set proper flags.
6562         (ASAN_MARK): Likewise.
6563
6564 2018-02-16  Julia Koval  <julia.koval@intel.com>
6565
6566         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6567         from PTA_CANNONLAKE.
6568
6569 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
6570
6571         PR target/84272
6572         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6573         Use ++iter rather than iter++ for std::list iterators.
6574         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
6575         defer deleting them until all nodes in the forest are processed.  Do
6576         free even leaf nodes.  Change to_process into auto_vec.
6577
6578         PR bootstrap/84405
6579         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6580         * vec.h (vec_default_construct): Use memset instead of placement new
6581         if BROKEN_VALUE_INITIALIZATION is defined.
6582         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6583         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6584         is defined.
6585
6586         PR rtl-optimization/83723
6587         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6588         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
6589         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
6590         recursive calls.
6591         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6592         callers.
6593         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6594
6595 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
6596
6597         PR rtl-optimization/81443
6598         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6599         from inner REGs to paradoxical SUBREGs.
6600
6601 2018-02-16  Richard Biener  <rguenther@suse.de>
6602
6603         PR tree-optimization/84399
6604         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6605         For operands we can analyze at their definition make sure we can
6606         analyze them at each use as well.
6607
6608 2018-02-16  Richard Biener  <rguenther@suse.de>
6609
6610         PR tree-optimization/84190
6611         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6612         volatile accesses if the decl isn't volatile.
6613
6614 2018-02-15  Jason Merrill  <jason@redhat.com>
6615
6616         PR c++/84314 - ICE with templates and fastcall attribute.
6617         * attribs.c (build_type_attribute_qual_variant): Don't clobber
6618         TYPE_CANONICAL on an existing type.
6619
6620 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
6621
6622         PR tree-optimization/84383
6623         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6624         dstoff nor call operand_equal_p if dstbase is NULL.
6625
6626         PR tree-optimization/84334
6627         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6628         also a CONSTANT_CLASS_P, punt.
6629
6630 2018-02-14  Jim Wilson  <jimw@sifive.com>
6631
6632         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6633         first SMALL_OPERAND check.  New local min_second_step.  Move assert
6634         to where locals are set.  Add TARGET_RVC support.
6635         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6636
6637 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
6638
6639         * doc/invoke.texi: Correct -Wformat-overflow code sample.
6640
6641 2018-02-14  Martin Sebor  <msebor@redhat.com>
6642
6643         PR tree-optimization/83698
6644         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6645         arrays constrain the offset range to their bounds.
6646         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6647         (builtin_access::overlap): Avoid setting the size of overlap if it's
6648         already been set.
6649         (maybe_diag_overlap): Also consider arrays when deciding what values
6650         of offsets to include in diagnostics.
6651
6652 2018-02-14  Martin Sebor  <msebor@redhat.com>
6653
6654         PR c/84108
6655         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6656         that correspond to the kind of a declaration.
6657
6658 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
6659
6660         PR target/83984
6661         * config/pa/pa.md: Load address of PIC label using the linkage table
6662         if the label is nonlocal.
6663
6664 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
6665
6666         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6667         warning message if user requests -maltivec=be.
6668         * doc/invoke.texi: Document deprecation of -maltivec=be.
6669
6670 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
6671
6672         PR target/84220
6673         * config/rs6000/rs6000-c.c: Update definitions for
6674         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6675         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6676
6677 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
6678
6679         PR target/84239
6680         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6681         add _get_ssp intrinsics. Remove argument from
6682         __builtin_ia32_rdssp[d|q].
6683         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6684         * config/i386/i386-builtin.def: Remove argument from
6685         __builtin_ia32_rdssp[d|q].
6686         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6687         ix86_expand_special_args_builtin for _rdssp[d|q].
6688         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6689         Clear register before usage.
6690         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6691         Add documentation for new _get_ssp and _inc_ssp intrinsics.
6692
6693 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
6694
6695         PR tree-optimization/84357
6696         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6697         operand 1 of an ARRAY_REF too.
6698
6699 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
6700
6701         PR target/83831
6702         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6703         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6704         declarations.
6705         (set_of_reg): New struct.
6706         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6707         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6708         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6709         functions.
6710         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6711         Split into bitclr, bitset, bitinvert patterns if appropriate.
6712         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6713         use rx_fuse_in_memory_bitop.
6714         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6715         to named insn, correct maximum insn length.
6716
6717 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
6718
6719         PR target/79242
6720         * machmode.def: Define a complex mode for PARTIAL_INT.
6721         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6722         MODE_PARTIAL_INT.
6723         * doc/rtl.texi: Document CSPImode.
6724         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6725         handling.
6726         (msp430_hard_regno_nregs_with_padding): Likewise.
6727
6728 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
6729
6730         PR target/84279
6731         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6732
6733 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
6734
6735         PR rtl-optimization/84169
6736         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
6737         we generated a parallel as new i3 and we split that to new i2 and i3
6738         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
6739         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
6740         those to i2, not i1.  Partially rewrite this scan code.
6741
6742 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
6743
6744         PR c/82210
6745         * stor-layout.c (place_field): For variable length fields, adjust
6746         offset_align afterwards not just based on the field's alignment,
6747         but also on the size.
6748
6749         PR middle-end/84309
6750         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6751         of exps and logs in the use_exp2 case.
6752
6753 2018-02-13  Jeff Law  <law@redhat.com>
6754
6755         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6756         entry for "vector".
6757
6758         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6759         ARGS as unused.
6760
6761 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
6762
6763         PR debug/84342
6764         PR debug/84319
6765         * common.opt (gas-loc-support, gas-locview-support): New.
6766         (ginline-points, ginternal-reset-location-views): New.
6767         * doc/invoke.texi: Document them.  Use @itemx where intended.
6768         (gvariable-location-views): Adjust.
6769         * target.def (reset_location_view): New.
6770         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6771         (TARGET_RESET_LOCATION_VIEW): New.
6772         * doc/tm.texi: Rebuilt.
6773         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6774         (dwarf2out_default_as_locview_support): New.
6775         (output_asm_line_debug_info): Use option variables.
6776         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6777         (output_loc_list): Likewise.
6778         (add_high_low_attributes): Check option variables.
6779         Don't output entry view attribute in strict mode.
6780         (gen_inlined_subroutine_die): Check option variables.
6781         (dwarf2out_inline_entry): Likewise.
6782         (init_sections_and_labels): Likewise.
6783         (dwarf2out_early_finish): Likewise.
6784         (maybe_reset_location_view): New, from...
6785         (dwarf2out_var_location): ... here.  Call it.
6786         * debug.h (dwarf2out_default_as_loc_support): Declare.
6787         (dwarf2out_default_as_locview_support): Declare.
6788         * hooks.c (hook_int_rtx_insn_0): New.
6789         * hooks.h (hook_int_rtx_insn_0): Declare.
6790         * toplev.c (process_options): Take -gas-loc-support and
6791         -gas-locview-support from dwarf2out.  Enable
6792         -gvariable-location-views by default only with locview
6793         assembler support.  Enable -ginternal-reset-location-views by
6794         default only if the target defines the corresponding hook.
6795         Enable -ginline-points by default if location views are
6796         enabled; force it disabled if statement frontiers are
6797         disabled.
6798         * tree-inline.c (expand_call_inline): Check option variables.
6799         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6800
6801 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
6802
6803         PR tree-optimization/84321
6804         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6805         handling.  Also check whether the anti-range contains any values
6806         that satisfy the mask; switch to a VR_RANGE if not.
6807
6808 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
6809
6810         PR sanitizer/84340
6811         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6812
6813 2018-02-13  Martin Jambor  <mjambor@suse.cz>
6814
6815         PR c++/83990
6816         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6817         of call statements, also set location of a load to a temporary.
6818
6819 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
6820
6821         * config/rl78/rl78.c (add_vector_labels): New function.
6822         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6823         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6824         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6825         which checks that no arguments are passed.
6826         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6827         * doc/extend.texi: Documentation for the new attribute.
6828
6829 2018-02-13  Andreas Schwab  <schwab@suse.de>
6830
6831         * config/riscv/linux.h (CPP_SPEC): Define.
6832
6833 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
6834
6835         PR target/84335
6836         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6837         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6838         OPTION_MASK_ISA_AES as first argument to def_builtin_const
6839         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6840         instead of OPTION_MASK_ISA_PCLMUL as first argument to
6841         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6842         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6843         temporarily for AES and PCLMUL builtins.
6844
6845         PR tree-optimization/84339
6846         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6847         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6848         Formatting fixes.
6849
6850         PR middle-end/84309
6851         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6852         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6853         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6854         inline function.
6855         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6856         inline function.
6857         * omp-simd-clone.h: New file.
6858         * omp-simd-clone.c: Include omp-simd-clone.h.
6859         (expand_simd_clones): No longer static.
6860         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6861         cgraph.h and omp-simd-clone.h.
6862         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6863         (vect_recog_widen_shift_pattern): Formatting fix.
6864         (vect_pattern_recog_1): Don't check optab for calls.
6865
6866         PR target/84336
6867         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6868         operands[2] into a REG before using gen_lowpart on it.
6869
6870 2018-02-12  Jeff Law  <law@redhat.com>
6871
6872         PR target/83760
6873         * config/sh/sh.c (find_barrier): Consider a sibling call
6874         a barrier as well.
6875
6876         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6877         successfully back substituting a reg.
6878
6879 2018-02-12  Richard Biener  <rguenther@suse.de>
6880
6881         PR tree-optimization/84037
6882         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6883         parameter, move visited init to caller.
6884         (vect_slp_analyze_operations): Separate cost from validity
6885         check, initialize visited once for all instances.
6886         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6887         for all instances.
6888         * tree-vect-stmts.c (vect_model_simple_cost): Make early
6889         out an assert.
6890         (vect_model_promotion_demotion_cost): Likewise.
6891         (vectorizable_bswap): Guard cost modeling with !slp_node
6892         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6893         SLP stmts.
6894         (vectorizable_call): Likewise.
6895         (vectorizable_conversion): Likewise.
6896         (vectorizable_assignment): Likewise.
6897         (vectorizable_shift): Likewise.
6898         (vectorizable_operation): Likewise.
6899         (vectorizable_store): Likewise.
6900         (vectorizable_load): Likewise.
6901         (vectorizable_condition): Likewise.
6902         (vectorizable_comparison): Likewise.
6903
6904 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
6905
6906         PR sanitizer/84307
6907         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6908         (ASAN_MARK): Fix fnspec to account for return value, change pointer
6909         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6910
6911 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
6912
6913         PR middle-end/83665
6914         * params.def (inline-min-speedup): Increase from 8 to 15.
6915         (max-inline-insns-auto): Decrease from 40 to 30.
6916         * ipa-split.c (consider_split): Add some buffer for function to
6917         be considered inlining candidate.
6918         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6919         default values.
6920
6921 2018-02-12  Richard Biener  <rguenther@suse.de>
6922
6923         PR tree-optimization/84037
6924         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6925         matched stmts if we cannot swap the non-matched ones.
6926
6927 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
6928
6929         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6930         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6931         _mm_maskz_scalef_round_ss): New intrinsics.
6932         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6933         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6934         __builtin_ia32_scalefss_round): Remove.
6935         (__builtin_ia32_scalefsd_mask_round,
6936         __builtin_ia32_scalefss_mask_round): New intrinsics.
6937         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6938         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6939         ((match_operand:VF_128 2 "<round_nimm_predicate>"
6940         "<round_constraint>")): Changed to ...
6941         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6942         "<round_scalar_constraint>")): ... this.
6943         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6944         %0, %1, %2<round_op3>}"): Changed to ...
6945         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6946         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6947         %2<round_scalar_mask_op3>}"): ... this.
6948         * config/i386/subst.md (round_scalar_nimm_predicate): New.
6949
6950 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
6951
6952         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6953         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6954         (_mm_maskz_sqrt_round_ss): New intrinsics.
6955         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6956         (__builtin_ia32_sqrtsd_mask_round)
6957         (__builtin_ia32_sqrtss_mask_round): New builtins.
6958         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6959         (__builtin_ia32_sqrtss_round): Remove.
6960         (__builtin_ia32_sqrtsd_mask_round)
6961         (__builtin_ia32_sqrtss_mask_round): New builtins.
6962         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6963         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6964         ((match_operand:VF_128 1 "vector_operand"
6965         "xBm,<round_constraint>")): Changed to ...
6966         ((match_operand:VF_128 1 "vector_operand"
6967         "xBm,<round_scalar_constraint>")): ... this.
6968         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6969         %0, %2, %<iptr>1<round_op3>}): Changed to ...
6970         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6971         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6972         %<iptr>1<round_scalar_mask_op3>}): ... this.
6973         ((set_attr "prefix" "<round_prefix>")): Changed to ...
6974         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6975
6976 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
6977
6978         PR target/84266
6979         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6980         Cast vec_cmpeq result to correct type.
6981         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6982         Cast vec_cmpgt result to correct type.
6983
6984 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
6985
6986         * final.c (final_scan_insn_1): Renamed from...
6987         (final_scan_insn): ... this.  New wrapper, to recover
6988         seen from the outermost call in recursive ones.
6989         * config/sparc/sparc.c (output_return): Drop seen from call.
6990         (output_sibcall): Likewise.
6991         * config/visium/visium.c (output_branch): Likewise.
6992
6993 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
6994
6995         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6996         function label.
6997
6998 2018-02-10  Alan Modra  <amodra@gmail.com>
6999
7000         PR target/84300
7001         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7002         Specify LR as an input.
7003
7004 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
7005
7006         PR sanitizer/83987
7007         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7008         remove_member_access_dummy_vars): New functions.
7009         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7010         lower_omp_1, execute_lower_omp): Use them.
7011
7012         PR rtl-optimization/84308
7013         * shrink-wrap.c (spread_components): Release todo vector.
7014
7015 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
7016
7017         PR rtl-optimization/57193
7018         * ira-color.c (struct allocno_color_data): Add member
7019         conflict_allocno_hard_prefs.
7020         (update_conflict_allocno_hard_prefs): New.
7021         (bucket_allocno_compare_func): Add a preference based on
7022         conflict_allocno_hard_prefs.
7023         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7024         (color_allocnos): Remove a dead code.  Initiate
7025         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
7026
7027 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
7028
7029         PR target/84226
7030         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7031         constraint from =wa to wa.  Avoid a subreg on the output operand,
7032         instead use a pseudo and subreg it in a move.
7033         (p9_xxbrd_<mode>): Changed to ...
7034         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7035         (p9_xxbrd_v2df): New expander.
7036         (p9_xxbrw_<mode>): Changed to ...
7037         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7038         (p9_xxbrw_v4sf): New expander.
7039
7040 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
7041
7042         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7043
7044 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
7045
7046         PR target/83926
7047         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7048         multiply in 32-bit mode.
7049         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7050         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7051         mode.
7052
7053 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
7054
7055         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7056         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7057         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7058         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7059
7060 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
7061
7062         PR lto/84213
7063         * dwarf2out.c (is_trivial_indirect_ref): New function.
7064         (dwarf2out_late_global_decl): Do not generate a location
7065         attribute for variables that have a non-trivial DECL_VALUE_EXPR
7066         and that are not defined in the current unit.
7067
7068 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
7069
7070         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7071         instead of a libcall for UNORDERED.
7072
7073 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
7074
7075         PR target/82641
7076         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7077         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7078
7079 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
7080
7081         PR target/PR84295
7082         * config/s390/s390.c (s390_set_current_function): Invoke
7083         s390_indirect_branch_settings also if fndecl didn't change.
7084
7085 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
7086
7087         * config/rs6000/rs6000.md (blockage): Set length to zero.
7088
7089 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
7090
7091         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7092
7093 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
7094
7095         PR sanitizer/84285
7096         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7097         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7098         -static-lib*san.
7099
7100         PR debug/84252
7101         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7102         PARALLEL incoming that failed vt_get_decl_and_offset check.
7103
7104         PR middle-end/84237
7105         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7106         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7107         TREE_READONLY bit.
7108         (get_variable_section): For decls in named .bss* sections pass true as
7109         second argument to bss_initializer_p.
7110
7111 2018-02-09  Marek Polacek  <polacek@redhat.com>
7112             Jakub Jelinek  <jakub@redhat.com>
7113
7114         PR c++/83659
7115         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7116         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
7117         Sync some changes from cxx_fold_indirect_ref.
7118
7119 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
7120
7121         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7122         markers.
7123         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7124         (BLOCK_INLINE_ENTRY_LABEL): New.
7125         (dwarf2out_var_location): Disregard inline entry markers.
7126         (inline_entry_data): New struct.
7127         (inline_entry_data_hasher): New hashtable type.
7128         (inline_entry_data_hasher::hash): New.
7129         (inline_entry_data_hasher::equal): New.
7130         (inline_entry_data_table): New variable.
7131         (add_high_low_attributes): Add DW_AT_entry_pc and
7132         DW_AT_GNU_entry_view attributes if a pending entry is found
7133         in inline_entry_data_table.  Add old entry_pc attribute only
7134         if debug nonbinding markers are disabled.
7135         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7136         markers are enabled.
7137         (block_within_block_p, dwarf2out_inline_entry): New.
7138         (dwarf2out_finish): Check that no entries remained in
7139         inline_entry_data_table.
7140         * final.c (reemit_insn_block_notes): Handle inline entry notes.
7141         (final_scan_insn, notice_source_line): Likewise.
7142         (rest_of_clean_state): Skip inline entry markers.
7143         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7144         markers.
7145         * gimple.c (gimple_build_debug_inline_entry): New.
7146         * gimple.h (enum gimple_debug_subcode): Add
7147         GIMPLE_DEBUG_INLINE_ENTRY.
7148         (gimple_build_debug_inline_entry): Declare.
7149         (gimple_debug_inline_entry_p): New.
7150         (gimple_debug_nonbind_marker_p): Adjust.
7151         * insn-notes.def (INLINE_ENTRY): New.
7152         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7153         inline entry marker notes.
7154         (print_insn): Likewise.
7155         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7156         (INSN_DEBUG_MARKER_KIND): Likewise.
7157         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7158         * tree-inline.c (expand_call_inline): Build and insert
7159         debug_inline_entry stmt.
7160         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7161         inline entry blocks early, if nonbind markers are enabled.
7162         (dump_scope_block): Dump fragment info.
7163         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7164         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7165         (gimple_build_debug_inline_entry): New.
7166         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7167         Enable/disable inline entry points too.
7168         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7169         (DEBUG_INSN): Describe inline entry markers.
7170
7171         * common.opt (gvariable-location-views): New.
7172         (gvariable-location-views=incompat5): New.
7173         * config.in: Rebuilt.
7174         * configure: Rebuilt.
7175         * configure.ac: Test assembler for view support.
7176         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7177         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7178         * dwarf2out.c (var_loc_view): New typedef.
7179         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7180         (dwarf2out_locviews_in_attribute): New.
7181         (dwarf2out_locviews_in_loclist): New.
7182         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7183         (enum dw_line_info_opcode): Add LI_adv_address.
7184         (struct dw_line_info_table): Add view.
7185         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7186         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7187         (zero_view_p): New variable.
7188         (ZERO_VIEW_P): New macro.
7189         (output_asm_line_debug_info): New.
7190         (struct var_loc_node): Add view.
7191         (add_AT_view_list, AT_loc_list): New.
7192         (add_var_loc_to_decl): Add view param.  Test it against last.
7193         (new_loc_list): Add view params.  Record them.
7194         (AT_loc_list_ptr): Handle loc and view lists.
7195         (view_list_to_loc_list_val_node): New.
7196         (print_dw_val): Handle dw_val_class_view_list.
7197         (size_of_die): Likewise.
7198         (value_format): Likewise.
7199         (loc_list_has_views): New.
7200         (gen_llsym): Set vl_symbol too.
7201         (maybe_gen_llsym, skip_loc_list_entry): New.
7202         (dwarf2out_maybe_output_loclist_view_pair): New.
7203         (output_loc_list): Output view list or entries too.
7204         (output_view_list_offset): New.
7205         (output_die): Handle dw_val_class_view_list.
7206         (output_dwarf_version): New.
7207         (output_compilation_unit_header): Use it.
7208         (output_skeleton_debug_sections): Likewise.
7209         (output_rnglists, output_line_info): Likewise.
7210         (output_pubnames, output_aranges): Update version comments.
7211         (output_one_line_info_table): Output view numbers in asm comments.
7212         (dw_loc_list): Determine current endview, pass it to new_loc_list.
7213         Call maybe_gen_llsym.
7214         (loc_list_from_tree_1): Adjust.
7215         (add_AT_location_description): Create view list attribute if
7216         needed, check it's absent otherwise.
7217         (convert_cfa_to_fb_loc_list): Adjust.
7218         (maybe_emit_file): Call output_asm_line_debug_info for test.
7219         (dwarf2out_var_location): Reset views as needed.  Precompute
7220         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
7221         attribute.  Set view.
7222         (new_line_info_table): Reset next view.
7223         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7224         (dwarf2out_source_line): Likewise.  Output view resets and labels to
7225         the assembler, or select appropriate line info opcodes.
7226         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7227         (optimize_string_length): Catch it.  Adjust.
7228         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
7229         dw_val_class_view_list, and remove it if no longer needed.
7230         (hash_loc_list): Hash view numbers.
7231         (loc_list_hasher::equal): Compare them.
7232         (optimize_location_lists): Check whether a view list symbol is
7233         needed, and whether the locview attribute is present, and
7234         whether they match.  Remove the locview attribute if no longer
7235         needed.
7236         (index_location_lists): Call skip_loc_list_entry for test.
7237         (dwarf2out_finish): Call output_asm_line_debug_info for test.
7238         Use output_dwarf_version.
7239         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7240         (struct dw_val_node): Add val_view_list.
7241         * final.c (SEEN_NEXT_VIEW): New.
7242         (set_next_view_needed): New.
7243         (clear_next_view_needed): New.
7244         (maybe_output_next_view): New.
7245         (final_start_function): Rename to...
7246         (final_start_function_1): ... this.  Take pointer to FIRST,
7247         add SEEN parameter.  Emit param bindings in the initial view.
7248         (final_start_function): Reintroduce SEEN-less interface.
7249         (final): Rename to...
7250         (final_1): ... this.  Take SEEN parameter.  Output final pending
7251         next view at the end.
7252         (final): Reintroduce seen-less interface.
7253         (final_scan_insn): Output pending next view before switching
7254         sections or ending a block.  Mark the next view as needed when
7255         outputting variable locations.  Notify debug backend of section
7256         changes, and of location view changes.
7257         (rest_of_handle_final): Adjust.
7258         * toplev.c (process_options): Autodetect value for debug variable
7259         location views option.  Warn on incompat5 without -gdwarf-5.
7260         * doc/invoke.texi (gvariable-location-views): New.
7261         (gvariable-location-views=incompat5): New.
7262         (gno-variable-location-views): New.
7263
7264 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
7265
7266         PR tree-optimization/84136
7267         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7268         that the result of find_edge is non-NULL.
7269
7270 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
7271
7272         PR target/83008
7273         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7274         storing integer register in SImode.  Fix cost of 256 and 512
7275         byte aligned SSE register store.
7276
7277 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
7278
7279         * config/i386/i386.c (ix86_multiplication_cost): Fix
7280         multiplication cost for TARGET_AVX512DQ.
7281
7282 2018-02-08  Marek Polacek  <polacek@redhat.com>
7283
7284         PR tree-optimization/84238
7285         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7286         get_range_strlen.
7287
7288 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
7289
7290         PR tree-optimization/84265
7291         * tree-vect-stmts.c (vectorizable_store): Don't treat
7292         VMAT_CONTIGUOUS accesses as grouped.
7293         (vectorizable_load): Likewise.
7294
7295 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
7296
7297         PR tree-optimization/81635
7298         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7299         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7300         (test_round_for_mask): New functions.
7301         (wide_int_cc_tests): Call test_round_for_mask.
7302         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7303         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7304         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7305         range returned by get_range_info.
7306
7307 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
7308
7309         PR ipa/81360
7310         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7311         * symtab.c: Include builtins.h
7312         (symtab_node::output_to_lto_symbol_table_p): Move here
7313         from lto-streamer-out.c:output_symbol_p.
7314         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7315         (output_symbol_p): Move all logic to symtab.c
7316         (produce_symtab): Update.
7317
7318 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
7319
7320         * config/s390/s390-opts.h (enum indirect_branch): Define.
7321         * config/s390/s390-protos.h (s390_return_addr_from_memory)
7322         (s390_indirect_branch_via_thunk)
7323         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7324         (enum s390_indirect_branch_type): Define.
7325         * config/s390/s390.c (struct s390_frame_layout, struct
7326         machine_function): Remove.
7327         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7328         (indirect_branch_table_label_no, indirect_branch_table_name):
7329         Define variables.
7330         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7331         (enum s390_indirect_branch_option): Define.
7332         (s390_return_addr_from_memory): New function.
7333         (s390_handle_string_attribute): New function.
7334         (s390_attribute_table): Add new attribute handler.
7335         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7336         (s390_indirect_branch_via_thunk): New function.
7337         (s390_indirect_branch_via_inline_thunk): New function.
7338         (s390_function_ok_for_sibcall): When jumping via thunk disallow
7339         sibling call optimization for non z10 compiles.
7340         (s390_emit_call): Force indirect branch target to be a single
7341         register.  Add r1 clobber for non-z10 compiles.
7342         (s390_emit_epilogue): Emit return jump via return_use expander.
7343         (s390_reorg): Handle JUMP_INSNs as execute targets.
7344         (s390_option_override_internal): Perform validity checks for the
7345         new command line options.
7346         (s390_indirect_branch_attrvalue): New function.
7347         (s390_indirect_branch_settings): New function.
7348         (s390_set_current_function): Invoke s390_indirect_branch_settings.
7349         (s390_output_indirect_thunk_function):  New function.
7350         (s390_code_end): Implement target hook.
7351         (s390_case_values_threshold): Implement target hook.
7352         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7353         macros.
7354         * config/s390/s390.h (struct s390_frame_layout)
7355         (struct machine_function): Move here from s390.c.
7356         (TARGET_INDIRECT_BRANCH_NOBP_RET)
7357         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7358         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7359         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7360         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7361         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7362         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7363         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7364         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7365         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7366         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7367         (mnemonic attribute): Add values which aren't recognized
7368         automatically.
7369         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7370         pattern for branch conversion.  Fix mnemonic attribute.
7371         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7372         indirect branch via thunk if requested.
7373         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7374         ("*indirect_jump"): Disable for branch conversion using out of
7375         line thunks.
7376         ("indirect_jump_via_thunk<mode>_z10")
7377         ("indirect_jump_via_thunk<mode>")
7378         ("indirect_jump_via_inlinethunk<mode>_z10")
7379         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7380         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7381         ("casesi_jump_via_inlinethunk<mode>_z10")
7382         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7383         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7384         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7385         ("*indirect2_jump"): Disable for branch conversion.
7386         ("casesi_jump"): Turn into expander and expand patterns for branch
7387         conversion.
7388         ("return_use"): New expander.
7389         ("*return"): Emit return via thunk and rename it to ...
7390         ("*return<mode>"): ... this one.
7391         * config/s390/s390.opt: Add new options and and enum for the
7392         option values.
7393
7394 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
7395
7396         * lra-constraints.c (match_reload): Unconditionally use
7397         gen_lowpart_SUBREG, rather than selecting between that
7398         and equivalent gen_rtx_SUBREG code.
7399
7400 2018-02-08  Richard Biener  <rguenther@suse.de>
7401
7402         PR tree-optimization/84233
7403         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7404         changed flag instead of boguously re-using phi_inserted.
7405
7406 2018-02-08  Martin Jambor  <mjambor@suse.cz>
7407
7408         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7409         static local variables.
7410
7411 2018-02-08  Richard Biener  <rguenther@suse.de>
7412
7413         PR tree-optimization/84278
7414         * tree-vect-stmts.c (vectorizable_store): When looking for
7415         smaller vector types to perform grouped strided loads/stores
7416         make sure the mode is supported by the target.
7417         (vectorizable_load): Likewise.
7418
7419 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
7420
7421         * config/aarch64/aarch64.c (aarch64_components_for_bb):
7422         Increase LDP/STP opportunities by adding adjacent callee-saves.
7423
7424 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
7425
7426         PR rtl-optimization/84068
7427         PR rtl-optimization/83459
7428         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7429
7430 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
7431
7432         PR tree-optimization/84224
7433         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7434         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7435         non-zero arguments.
7436
7437 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
7438
7439         PR target/84113
7440         * config/rs6000/altivec.md (*restore_world): Remove LR use.
7441         * config/rs6000/predicates.md (restore_world_operation): Adjust op
7442         count, remove one USE.
7443
7444 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
7445
7446         * doc/install.texi (Configuration): Document the
7447         --with-long-double-format={ibm,ieee} PowerPC configuration
7448         options.
7449
7450         PR target/84154
7451         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7452         Convert from define_expand to be define_insn_and_split.  Rework
7453         float/double/_Float128 conversions to QI/HI/SImode to work with
7454         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
7455         conversions to QI/HImode types did a store and then a load to
7456         truncate the value.  For conversions to VSX registers, don't split
7457         the insn, instead emit the code directly.  Use the code iterator
7458         any_fix to combine signed and unsigned conversions.
7459         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7460         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7461         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7462         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7463         (fix_<mode>di2_hw): Likewise.
7464         (fixuns_<mode>di2_hw): Likewise.
7465         (fix_<mode>si2_hw): Likewise.
7466         (fixuns_<mode>si2_hw): Likewise.
7467         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7468         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7469         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7470         fix<uns>_trunc<SFDF:mode>si2_p8.
7471         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7472         used.
7473         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7474         (fix<uns>_<mode>_mem): Likewise.
7475         (fctiw<u>z_<mode>_mem): Likewise.
7476         (fix<uns>_<mode>_mem): Likewise.
7477         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7478         the register allocator from doing a direct move to the GPRs to do
7479         a store, and instead use the ISA 3.0 store byte/half-word from
7480         vector register instruction.  For IEEE 128-bit floating point,
7481         also optimize stores of 32-bit ints.
7482         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7483
7484 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
7485
7486         * genextract.c (push_pathstr_operand): New function to support
7487         [a-zA-Z].
7488         (walk_rtx): Call push_pathstr_operand.
7489         (print_path): Support [a-zA-Z].
7490
7491 2018-02-07  Richard Biener  <rguenther@suse.de>
7492
7493         PR tree-optimization/84037
7494         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7495         (cse_and_gimplify_to_preheader): Declare.
7496         (vect_get_place_in_interleaving_chain): Likewise.
7497         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7498         ivexpr_map.
7499         (_loop_vec_info::~_loop_vec_info): Delete it.
7500         (cse_and_gimplify_to_preheader): New function.
7501         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7502         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7503         (vectorizable_load): Likewise.  For grouped stores always base
7504         the IV on the first element.
7505         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7506         condition before gimplifying.
7507
7508 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
7509
7510         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7511         *DIV_EXPR and *MOD_EXPR.
7512
7513 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
7514
7515         PR target/84248
7516         * config/i386/i386.c (ix86_option_override_internal): Mask out
7517         the CF_SET bit when checking -fcf-protection.
7518
7519 2018-02-07  Tom de Vries  <tom@codesourcery.com>
7520
7521         PR libgomp/84217
7522         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7523         enough.
7524
7525 2018-02-07  Richard Biener  <rguenther@suse.de>
7526
7527         PR tree-optimization/84204
7528         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7529         this place.
7530
7531         PR tree-optimization/84205
7532         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7533         special-case isl_ast_op_zdiv_r.
7534
7535         PR tree-optimization/84223
7536         * graphite-scop-detection.c (gather_bbs::before_dom_children):
7537         Only add conditions from within the region.
7538         (gather_bbs::after_dom_children): Adjust.
7539
7540 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
7541
7542         PR target/84209
7543         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7544         * config/avr/avr.md: Only post-reload split REG-REG moves if
7545         either register is GENERAL_REG_P.
7546
7547 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
7548
7549         PR tree-optimization/84235
7550         * tree-ssa-scopedtables.c
7551         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7552         if the subtraction is performed in floating point type where NaNs are
7553         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7554         build 1.  Formatting fix.
7555
7556 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
7557
7558         PR target/84146
7559         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7560         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7561         and skip it regardless of bb boundaries.  Use CALL_P macro,
7562         don't test INSN_P (insn) together with CALL_P or JUMP_P check
7563         unnecessarily, formatting fix.
7564
7565 2018-02-06  Michael Collison  <michael.collison@arm.com>
7566
7567         * config/arm/thumb2.md:
7568         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7569         (*thumb_mov_notscc): Ditto.
7570
7571 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
7572
7573         PR target/84154
7574         * config/rs6000/rs6000.md (su code attribute): Use "u" for
7575         unsigned_fix, not "s".
7576
7577 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7578
7579         * configure.ac (gcc_fn_eh_frame_ro): New function.
7580         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7581         correct .eh_frame permissions.
7582         * configure: Regenerate.
7583
7584 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
7585
7586         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7587         irrelevant options.
7588
7589 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
7590
7591         * config/rs6000/rs6000.c (rs6000_option_override_internal):
7592         Display warning message for -mno-speculate-indirect-jumps.
7593
7594 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
7595
7596         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7597         Undocumented.
7598         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7599
7600 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
7601
7602         PR tree-optimization/84225
7603         * tree-eh.c (find_trapping_overflow): Only call
7604         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7605
7606 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
7607
7608         PR target/84145
7609         * config/i386/i386.c: Reimplement the check of possible options
7610         -mibt/-mshstk conbination. Change error messages.
7611         * doc/invoke.texi: Fix a typo: remove extra '='.
7612
7613 2018-02-06  Marek Polacek  <polacek@redhat.com>
7614
7615         PR tree-optimization/84228
7616         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7617
7618 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
7619
7620         PR target/82641
7621         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7622         emitted arch directives.
7623         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7624         __ARM_FEATURE_COPROC before changing architectures.
7625
7626 2018-02-06  Richard Biener  <rguenther@suse.de>
7627
7628         * config/i386/i386.c (print_reg): Fix typo.
7629         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7630
7631 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
7632
7633         * configure: Regenerate.
7634
7635 2018-02-05  Martin Sebor  <msebor@redhat.com>
7636
7637         PR tree-optimization/83369
7638         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7639         inlining context.
7640
7641 2018-02-05  Martin Liska  <mliska@suse.cz>
7642
7643         * doc/invoke.texi: Cherry-pick upstream r323995.
7644
7645 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
7646
7647         * ira.c (ira_init_register_move_cost): Adjust comment.
7648
7649 2018-02-05  Martin Liska  <mliska@suse.cz>
7650
7651         PR gcov-profile/84137
7652         * doc/gcov.texi: Fix typo in documentation.
7653
7654 2018-02-05  Martin Liska  <mliska@suse.cz>
7655
7656         PR gcov-profile/83879
7657         * doc/gcov.texi: Document necessity of --dynamic-list-data when
7658         using dlopen functionality.
7659
7660 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
7661
7662         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7663         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7664         _mm_maskz_range_ss, _mm_mask_range_round_ss,
7665         _mm_maskz_range_round_ss): New intrinsics.
7666         (__builtin_ia32_rangesd128_round)
7667         (__builtin_ia32_rangess128_round): Remove.
7668         (__builtin_ia32_rangesd128_mask_round,
7669         __builtin_ia32_rangess128_mask_round): New builtins.
7670         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7671         __builtin_ia32_rangess128_round): Remove.
7672         (__builtin_ia32_rangesd128_mask_round,
7673         __builtin_ia32_rangess128_mask_round): New builtins.
7674         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7675         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7676         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7677         "<round_saeonly_constraint>")): Changed to ...
7678         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7679         "<round_saeonly_scalar_constraint>")): ... this.
7680         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7681         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7682         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7683         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7684         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7685
7686 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
7687
7688         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7689         options.
7690         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7691         Remove all values except native, 8540 and 8548.
7692
7693 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
7694
7695         * config/i386/i386.c (ix86_output_function_return): Pass
7696         INVALID_REGNUM, instead of -1, as invalid register number to
7697         indirect_thunk_name and output_indirect_thunk.
7698
7699 2018-02-02  Julia Koval  <julia.koval@intel.com>
7700
7701         * config.gcc: Add -march=icelake.
7702         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7703         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7704         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7705         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7706         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7707         (processor_target_table): Add icelake.
7708         (ix86_option_override_internal): Handle new PTAs.
7709         (get_builtin_code_for_version): Handle icelake.
7710         (M_INTEL_COREI7_ICELAKE): New.
7711         (fold_builtin_cpu): Handle icelake.
7712         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7713         * doc/invoke.texi: Add -march=icelake.
7714
7715 2018-02-02  Julia Koval  <julia.koval@intel.com>
7716
7717         * config/i386/i386.c (ix86_option_override_internal): Change flags type
7718         to wide_int_bitmask.
7719         * wide-int-bitmask.h: New.
7720
7721 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
7722
7723         PR target/84066
7724         * config/i386/i386.md: Replace Pmode with word_mode in
7725         builtin_setjmp_setup and builtin_longjmp to support x32.
7726
7727 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
7728
7729         PR target/56010
7730         PR target/83743
7731         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7732         #include "opts.h".
7733         (rs6000_supported_cpu_names): New static variable.
7734         (linux_cpu_translation_table): Likewise.
7735         (elf_platform) <cpu>: Define new static variable and use it.
7736         Translate kernel AT_PLATFORM name to canonical name if needed.
7737         Error if platform name is unknown.
7738
7739 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
7740
7741         PR target/84089
7742         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7743
7744 2018-02-01  Jeff Law  <law@redhat.com>
7745
7746         PR target/84128
7747         * config/i386/i386.c (release_scratch_register_on_entry): Add new
7748         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
7749         the scratch if RELEASE_VIA_POP is false.
7750         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7751         If we have to save a temporary register, decrement SIZE appropriately.
7752         Pass new arguments to release_scratch_register_on_entry.
7753         (ix86_adjust_stack_and_probe): Likewise.
7754         (ix86_emit_probe_stack_range): Pass new arguments to
7755         release_scratch_register_on_entry.
7756
7757 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
7758
7759         PR rtl-optimization/84157
7760         * combine.c (change_zero_ext): Use REG_P predicate in
7761         front of HARD_REGISTER_P predicate.
7762
7763 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
7764
7765         * config/avr/avr.c (avr_option_override): Move disabling of
7766         -fdelete-null-pointer-checks to...
7767         * common/config/avr/avr-common.c (avr_option_optimization_table):
7768         ...here.
7769
7770 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
7771
7772         PR tree-optimization/81635
7773         * tree-data-ref.c (split_constant_offset_1): For types that
7774         wrap on overflow, try to use range info to prove that wrapping
7775         cannot occur.
7776
7777 2018-02-01  Renlin Li  <renlin.li@arm.com>
7778
7779         PR target/83370
7780         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7781         TAILCALL_ADDR_REGS.
7782         (aarch64_register_move_cost): Likewise.
7783         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7784         TAILCALL_ADDR_REGS.
7785         (REG_CLASS_NAMES): Likewise.
7786         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7787         TAILCALL_ADDR_REGS. Remove IP registers.
7788         * config/aarch64/aarch64.md (Ucs): Update register constraint.
7789
7790 2018-02-01  Richard Biener  <rguenther@suse.de>
7791
7792         * domwalk.h (dom_walker::dom_walker): Add additional constructor
7793         for specifying RPO order and allow NULL for that.
7794         * domwalk.c (dom_walker::dom_walker): Likewise.
7795         (dom_walker::walk): Handle NULL RPO order.
7796         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7797         in RPO order.
7798         (rewrite_update_dom_walker): Likewise.
7799         (mark_def_dom_walker): Likewise.
7800
7801 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
7802
7803         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7804         (aarch64_maybe_expand_sve_subreg_move): Declare.
7805         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7806         * config/aarch64/predicates.md (aarch64_any_register_operand): New
7807         predicate.
7808         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7809         that are semantically a reverse operation.
7810         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7811         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7812         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7813         functions.
7814         (aarch64_can_change_mode_class): For big-endian, forbid changes
7815         between two SVE modes if they have different element sizes.
7816
7817 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
7818
7819         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7820         the TImode handling for big-endian targets.
7821
7822 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
7823
7824         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7825         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
7826         not just bytes.
7827         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7828         Remove BSWAP handing for big-endian targets and use the form of
7829         LD1RQ appropariate for the mode.
7830
7831 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
7832
7833         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7834         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7835         duplicated element.
7836
7837 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
7838
7839         PR tearget/83845
7840         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7841         check for operands that need to go through aarch64_sve_reload_be.
7842
7843 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
7844
7845         PR tree-optimization/81661
7846         PR tree-optimization/84117
7847         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7848         * tree-eh.c: Include gimplify.h.
7849         (find_trapping_overflow, replace_trapping_overflow,
7850         rewrite_to_non_trapping_overflow): New functions.
7851         * tree-vect-loop.c: Include tree-eh.h.
7852         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7853         * tree-data-ref.c: Include tree-eh.h.
7854         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7855
7856 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
7857
7858         PR rtl-optimization/84123
7859         * combine.c (change_zero_ext): Check if hard register satisfies
7860         can_change_dest_mode before calling gen_lowpart_SUBREG.
7861
7862 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
7863
7864         PR target/82444
7865         * ira.c (ira_init_register_move_cost): Remove assert.
7866
7867 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
7868
7869         PR rtl-optimization/84071
7870         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7871         * doc/tm.texi: Regenerate.
7872
7873 2018-01-31  Richard Biener  <rguenther@suse.de>
7874
7875         PR tree-optimization/84132
7876         * tree-data-ref.c (analyze_miv_subscript): Properly
7877         check whether evolution_function_is_affine_multivariate_p
7878         before calling gcd_of_steps_may_divide_p.
7879
7880 2018-01-31  Julia Koval  <julia.koval@intel.com>
7881
7882         PR target/83618
7883         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7884         * config/i386/i386.md (rdpid_rex64) New.
7885         (rdpid): Make 32bit only.
7886
7887 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
7888
7889         PR lto/84105
7890         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7891         an IDENTIFIER_NODE for FUNCTION_TYPE's.
7892
7893 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
7894
7895         Revert
7896         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
7897
7898         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7899
7900 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
7901
7902         PR rtl-optimization/84071
7903         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7904         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7905
7906 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
7907
7908         * config/arc/arc.c (arc_handle_aux_attribute): New function.
7909         (arc_attribute_table): Add 'aux' attribute.
7910         (arc_in_small_data_p): Consider aux like variables.
7911         (arc_is_aux_reg_p): New function.
7912         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7913         (arc_get_aux_arg): New function.
7914         (prepare_move_operands): Handle aux-register access.
7915         (arc_handle_aux_attribute): New function.
7916         * doc/extend.texi (ARC Variable attributes): Add subsection.
7917
7918 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
7919
7920         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7921         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7922         (arc_attribute_table): Add 'uncached' attribute.
7923         (arc_print_operand): Print '.di' flag for uncached memory
7924         accesses.
7925         (arc_in_small_data_p): Do not consider for small data the uncached
7926         types.
7927         (arc_is_uncached_mem_p): New function.
7928         * config/arc/predicates.md (compact_store_memory_operand): Check
7929         for uncached memory accesses.
7930         (nonvol_nonimm_operand): Likewise.
7931         * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7932
7933 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
7934
7935         PR c/84100
7936         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7937         falign-loops=): Add Optimization flag.
7938
7939 2018-01-30  Jeff Law  <law@redhat.com>
7940
7941         PR target/84064
7942         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7943         INT_REGISTERS_SAVED.  Check it prior to calling
7944         get_scratch_register_on_entry.
7945         (ix86_adjust_stack_and_probe): Similarly.
7946         (ix86_emit_probe_stack_range): Similarly.
7947         (ix86_expand_prologue): Corresponding changes.
7948
7949 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7950
7951         PR target/40411
7952         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7953         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7954
7955 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
7956
7957         PR target/84112
7958         * lra-constraints.c (curr_insn_transform): Process AND in the
7959         address.
7960
7961 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
7962
7963         PR rtl-optimization/83986
7964         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7965         dependence against last_pending_memory_flush in addition to
7966         pending_jump_insns.
7967
7968 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
7969
7970         PR tree-optimization/81611
7971         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7972         copies.
7973
7974 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
7975
7976         PR target/83758
7977         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7978         a reg rtx.
7979
7980 2018-01-30  Richard Biener  <rguenther@suse.de>
7981             Jakub Jelinek  <jakub@redhat.com>
7982
7983         PR tree-optimization/84111
7984         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7985         inner loops added during recursion, as they don't have up-to-date
7986         SSA form.
7987
7988 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
7989
7990         PR ipa/81360
7991         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7992         (can_inline_edge_by_limits_p): ... here.
7993         (can_early_inline_edge_p, check_callers,
7994         update_caller_keys, update_callee_keys, recursive_inlining,
7995         add_new_edges_to_heap, speculation_useful_p,
7996         inline_small_functions,
7997         inline_small_functions, flatten_function,
7998         inline_to_all_callers_1): Update.
7999
8000 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
8001
8002         * profile-count.c (profile_count::combine_with_ipa_count): Handle
8003         zeros correctly.
8004
8005 2018-01-30  Richard Biener  <rguenther@suse.de>
8006
8007         PR tree-optimization/83008
8008         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8009         invariant and constant vector uses in stmts when they need
8010         more than one stmt.
8011
8012 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
8013
8014         PR bootstrap/84017
8015         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8016         * configure: Regenerate.
8017
8018 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
8019
8020         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8021         pattern.
8022         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8023         Use gen_rtx_REG rather than gen_lowpart.
8024
8025 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
8026
8027         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8028         rather than 0 when creating partial subregs.
8029
8030 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
8031
8032         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8033         of usage.
8034
8035 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
8036
8037         PR target/81550
8038         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8039         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8040         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8041         flags.  This restores the settings used before the 2017-07-24.
8042         Turning off pre increment/decrement/modify allows IVOPTS to
8043         optimize DF/SF loops where the index is an int.
8044
8045 2018-01-29  Richard Biener  <rguenther@suse.de>
8046             Kelvin Nilsen  <kelvin@gcc.gnu.org>
8047
8048         PR bootstrap/80867
8049         * tree-vect-stmts.c (vectorizable_call): Don't call
8050         targetm.vectorize_builtin_md_vectorized_function if callee is
8051         NULL.
8052
8053 2018-01-22  Carl Love  <cel@us.ibm.com>
8054
8055         * doc/extend.tex: Fix typo in second arg in
8056         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8057
8058 2018-01-29  Richard Biener  <rguenther@suse.de>
8059
8060         PR tree-optimization/84086
8061         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8062         (flush_ssaname_freelist): When SSA names were released reset
8063         the SCEV hash table.
8064
8065 2018-01-29  Richard Biener  <rguenther@suse.de>
8066
8067         PR tree-optimization/84057
8068         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8069         removed paths when removing edges.
8070
8071 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
8072
8073         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8074         -mfunction-return=@var{choice}.
8075
8076 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8077
8078         PR diagnostic/84034
8079         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8080         Handle CR like TAB.
8081         (layout::print_source_line): Likewise.
8082         (test_get_line_width_without_trailing_whitespace): Add test cases.
8083
8084 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
8085
8086         PR middle-end/84040
8087         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8088         debug insns.
8089
8090 2018-01-26  Jim Wilson  <jimw@sifive.com>
8091
8092         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8093
8094         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8095         specified.
8096
8097 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8098
8099         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8100         and CMP + SUB-immediate -> SUBS.
8101
8102 2018-01-26  Martin Sebor  <msebor@redhat.com>
8103
8104         PR tree-optimization/83896
8105         * tree-ssa-strlen.c (get_string_len): Rename...
8106         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
8107         Avoid assuming length is constant.
8108         (handle_char_store): Use HOST_WIDE_INT for string length.
8109
8110 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
8111
8112         PR target/81763
8113         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8114         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8115
8116 2018-01-26  Richard Biener  <rguenther@suse.de>
8117
8118         PR rtl-optimization/84003
8119         * dse.c (record_store): Only record redundant stores when
8120         the earlier store aliases at least all accesses the later one does.
8121
8122 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
8123
8124         PR rtl-optimization/83985
8125         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8126         REG_CFA_RESTORE insns.
8127         (delete_unmarked_insns): Don't ignore separate shrink wrapping
8128         REG_CFA_RESTORE insns here.
8129
8130         PR c/83989
8131         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8132         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8133
8134 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
8135
8136         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8137         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8138         (arc_init): Likewise.
8139         (arc_override_options): Likewise.
8140         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8141         value.
8142         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8143         support.
8144         * config/arc/arc.h (TARGET_DBNZ): Define.
8145         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8146         properly set the tune attribute.
8147         (dbnz): Use TARGET_DBNZ guard.
8148         * config/arc/arc.opt (mtune): Add core3 option.
8149
8150 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
8151
8152         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8153         recognize new pic like addresses.
8154         (arc_delegitimize_address): Clean up.
8155
8156 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
8157
8158         * config/arc/arc-arches.def: Option mrf16 valid for all
8159         architectures.
8160         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8161         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8162         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8163         * config/arc/arc-tables.opt: Regenerate.
8164         * config/arc/arc.c (arc_conditional_register_usage): Handle
8165         reduced register file case.
8166         (arc_file_start): Set must have build attributes.
8167         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8168         mrf16 option value.
8169         * config/arc/arc.opt (mrf16): Add new option.
8170         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8171         * config/arc/genmultilib.awk: Handle new mrf16 option.
8172         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8173         * config/arc/t-multilib: Regenerate.
8174         * doc/invoke.texi (ARC Options): Document mrf16 option.
8175
8176 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
8177
8178         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8179         * config/arc/arc.c (arc_handle_secure_attribute): New function.
8180         (arc_attribute_table): Add 'secure_call' attribute.
8181         (arc_print_operand): Print secure call operand.
8182         (arc_function_ok_for_sibcall): Don't optimize tail calls when
8183         secure.
8184         (arc_is_secure_call_p): New function.  * config/arc/arc.md
8185         (call_i): Add support for sjli instruction.
8186         (call_value_i): Likewise.
8187         * config/arc/constraints.md (Csc): New constraint.
8188
8189 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
8190             John Eric Martin  <John.Martin@emmicro-us.com>
8191
8192         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8193         * config/arc/arc.c (_arc_jli_section): New struct.
8194         (arc_jli_section): New type.
8195         (rc_jli_sections): New static variable.
8196         (arc_handle_jli_attribute): New function.
8197         (arc_attribute_table): Add jli_always and jli_fixed attribute.
8198         (arc_file_end): New function.
8199         (TARGET_ASM_FILE_END): Define.
8200         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8201         (arc_add_jli_section): New function.
8202         (jli_call_scan): Likewise.
8203         (arc_reorg): Call jli_call_scan.
8204         (arc_output_addsi): Remove 'S' from printing asm operand.
8205         (arc_is_jli_call_p): New function.
8206         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8207         operand.
8208         (movhi_insn): Likewise.
8209         (movsi_insn): Likewise.
8210         (movsi_set_cc_insn): Likewise.
8211         (loadqi_update): Likewise.
8212         (load_zeroextendqisi_update): Likewise.
8213         (load_signextendqisi_update): Likewise.
8214         (loadhi_update): Likewise.
8215         (load_zeroextendhisi_update): Likewise.
8216         (load_signextendhisi_update): Likewise.
8217         (loadsi_update): Likewise.
8218         (loadsf_update): Likewise.
8219         (movsicc_insn): Likewise.
8220         (bset_insn): Likewise.
8221         (bxor_insn): Likewise.
8222         (bclr_insn): Likewise.
8223         (bmsk_insn): Likewise.
8224         (bicsi3_insn): Likewise.
8225         (cmpsi_cc_c_insn): Likewise.
8226         (movsi_ne): Likewise.
8227         (movsi_cond_exec): Likewise.
8228         (clrsbsi2): Likewise.
8229         (norm_f): Likewise.
8230         (normw): Likewise.
8231         (swap): Likewise.
8232         (divaw): Likewise.
8233         (flag): Likewise.
8234         (sr): Likewise.
8235         (kflag): Likewise.
8236         (ffs): Likewise.
8237         (ffs_f): Likewise.
8238         (fls): Likewise.
8239         (call_i): Remove 'S' asm letter, add jli instruction.
8240         (call_value_i): Likewise.
8241         * config/arc/arc.op (mjli-always): New option.
8242         * config/arc/constraints.md (Cji): New constraint.
8243         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8244         operand.
8245         (subsf3_fpx): Likewise.
8246         (mulsf3_fpx): Likewise.
8247         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8248         asm operand.
8249         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8250         function attrbutes.
8251         * doc/invoke.texi (ARC): Document mjli-always option.
8252
8253 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
8254
8255         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8256         avoid addition with 0 and use incw and decw where possible.
8257
8258 2018-01-26  Richard Biener  <rguenther@suse.de>
8259
8260         PR tree-optimization/81082
8261         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8262         association if it requires casting to unsigned.
8263         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8264         from fold_plusminus_mult_expr to catch important cases late when
8265         range info is available.
8266
8267 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
8268
8269         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8270         * configure.ac (hidden_linkonce): New test.
8271         * configure: Regenerate.
8272         * config.in: Regenerate.
8273
8274 2018-01-26  Julia Koval  <julia.koval@intel.com>
8275
8276         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8277         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8278         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8279         _mm_mask_bitshuffle_epi64_mask): Fix type.
8280         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8281         USI_FTYPE_V4DI_V4DI_USI): Remove.
8282         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8283         __builtin_ia32_vpshufbitqmb256_mask,
8284         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8285         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8286         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8287
8288 2018-01-26  Alan Modra  <amodra@gmail.com>
8289
8290         PR target/84033
8291         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8292         UNSPEC_VBPERMQ.  Sort other unspecs.
8293
8294 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
8295
8296         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8297
8298 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
8299
8300         PR middle-end/83055
8301         * predict.c (drop_profile): Do not push/pop cfun; update also
8302         node->count.
8303         (handle_missing_profiles): Fix logic looking for zero profiles.
8304
8305 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
8306
8307         PR middle-end/83977
8308         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8309         on functions with #pragma omp declare simd or functions with simd
8310         attribute.
8311         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8312         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8313         Remove trailing \n from warning_at calls.
8314
8315 2018-01-25  Tom de Vries  <tom@codesourcery.com>
8316
8317         PR target/84028
8318         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8319         for neutered workers.
8320
8321 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
8322
8323         PR target/68467
8324         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8325         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8326
8327 2018-01-24  Jeff Law  <law@redhat.com>
8328
8329         PR target/83994
8330         * i386.c (get_probe_interval): Move to earlier point.
8331         (ix86_compute_frame_layout): If -fstack-clash-protection and
8332         the frame is larger than the probe interval, then use pushes
8333         to save registers rather than reg->mem moves.
8334         (ix86_expand_prologue): Remove conditional for int_registers_saved
8335         assertion.
8336
8337 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
8338
8339         PR target/84014
8340         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8341         min/max for never referenced object.
8342
8343 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
8344
8345         PR middle-end/83977
8346         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8347         here.
8348         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8349         attributes from DECL_ATTRIBUTES (decl) without affecting
8350         DECL_ATTRIBUTES (current_function_decl).
8351         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8352         functions with non-NULL DECL_ABSTRACT_ORIGIN.
8353
8354 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
8355
8356         PR tree-optimization/83979
8357         * fold-const.c (fold_comparison): Use constant_boolean_node
8358         instead of boolean_{true,false}_node.
8359
8360 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
8361
8362         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8363         with zero counts.
8364
8365 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
8366
8367         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8368         Simplify the clause that sets the length attribute.
8369         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8370         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8371         clause that sets the length attribute.
8372         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8373
8374 2018-01-24  Tom de Vries  <tom@codesourcery.com>
8375
8376         PR target/83589
8377         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8378         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8379         Add strict parameter.
8380         (prevent_branch_around_nothing): Insert dummy insn between branch to
8381         label and label with no ptx insn inbetween.
8382         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8383
8384 2018-01-24  Tom de Vries  <tom@codesourcery.com>
8385
8386         PR target/81352
8387         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8388         for neutered threads in warp.
8389         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8390
8391 2018-01-24  Richard Biener  <rguenther@suse.de>
8392
8393         PR tree-optimization/83176
8394         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8395         operands.
8396
8397 2018-01-24  Richard Biener  <rguenther@suse.de>
8398
8399         PR tree-optimization/82819
8400         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8401         code generating pluses that are no-ops in the target precision.
8402
8403 2018-01-24  Richard Biener  <rguenther@suse.de>
8404
8405         PR middle-end/84000
8406         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8407
8408 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
8409
8410         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8411         to merge probabilities.
8412         * predict.c (probably_never_executed): Also mark as cold functions
8413         with global 0 profile and guessed local profile.
8414         * profile-count.c (profile_probability::combine_with_count): New
8415         member function.
8416         * profile-count.h (profile_probability::operator*,
8417         profile_probability::operator*=, profile_probability::operator/,
8418         profile_probability::operator/=): Reduce precision to adjusted
8419         and set value to guessed on contradictory divisions.
8420         (profile_probability::combine_with_freq): Remove.
8421         (profile_probability::combine_wiht_count): Declare.
8422         (profile_count::force_nonzero):: Set to adjusted.
8423         (profile_count::probability_in):: Set quality to adjusted.
8424         * tree-ssa-tail-merge.c (replace_block_by): Use
8425         combine_with_count.
8426
8427 2018-01-23  Andrew Waterman  <andrew@sifive.com>
8428             Jim Wilson  <jimw@sifive.com>
8429
8430         * config/riscv/riscv.c (riscv_stack_boundary): New.
8431         (riscv_option_override): Set riscv_stack_boundary.  Handle
8432         riscv_preferred_stack_boundary_arg.
8433         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8434         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8435         (STACK_BOUNDARY): Set to riscv_stack_boundary.
8436         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8437         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8438         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8439
8440 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
8441
8442         PR target/83905
8443         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8444         of struct ix86_frame.
8445         (ix86_expand_epilogue): Likewise.  Add a local variable for
8446         the reg_save_offset field in struct ix86_frame.
8447
8448 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
8449
8450         PR tree-optimization/82604
8451         * tree-loop-distribution.c (enum partition_kind): New enum item
8452         PKIND_PARTIAL_MEMSET.
8453         (partition_builtin_p): Support above new enum item.
8454         (generate_code_for_partition): Ditto.
8455         (compute_access_range): Differentiate cases that equality can be
8456         proven at all loops, the innermost loops or no loops.
8457         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8458         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
8459         (finalize_partitions, distribute_loop): Don't fuse partition of
8460         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8461         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8462         parloop is enabled.
8463
8464 2018-01-23  Martin Liska  <mliska@suse.cz>
8465
8466         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8467         order to ignore the predictor.
8468         (PRED_POLYMORPHIC_CALL): Likewise.
8469         (PRED_RECURSIVE_CALL): Likewise.
8470
8471 2018-01-23  Martin Liska  <mliska@suse.cz>
8472
8473         * tree-profile.c (tree_profiling): Print function header to
8474         aware reader which function we are working on.
8475         * value-prof.c (gimple_find_values_to_profile): Do not print
8476         not interesting value histograms.
8477
8478 2018-01-23  Martin Liska  <mliska@suse.cz>
8479
8480         * profile-count.h (enum profile_quality): Add
8481         profile_uninitialized as the first value. Do not number values
8482         as they are zero based.
8483         (profile_count::verify): Update sanity check.
8484         (profile_probability::verify): Likewise.
8485
8486 2018-01-23  Nathan Sidwell  <nathan@acm.org>
8487
8488         * doc/invoke.texi (ffor-scope): Deprecate.
8489
8490 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
8491
8492         PR tree-optimization/83510
8493         * domwalk.c (set_all_edges_as_executable): New function.
8494         (dom_walker::dom_walker): Convert bool param
8495         "skip_unreachable_blocks" to enum reachability.  Move setup of
8496         edge flags to set_all_edges_as_executable and only do it when
8497         reachability is REACHABLE_BLOCKS.
8498         * domwalk.h (enum dom_walker::reachability): New enum.
8499         (dom_walker::dom_walker): Convert bool param
8500         "skip_unreachable_blocks" to enum reachability.
8501         (set_all_edges_as_executable): New decl.
8502         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
8503         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8504         "reachability".
8505         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8506         but converting true to REACHABLE_BLOCKS.
8507         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8508         * tree-vrp.c
8509         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8510         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8511         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8512         REACHABLE_BLOCKS.
8513         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8514         if check_all_array_refs will be called.
8515
8516 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
8517
8518         * tree.c (selftest::test_location_wrappers): Add more test
8519         coverage.
8520
8521 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
8522
8523         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8524         (selftest::test_bit_in_range): Likewise.
8525
8526 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
8527
8528         PR testsuite/83888
8529         * doc/sourcebuild.texi (vect_float): Say that the selector
8530         only describes the situation when -funsafe-math-optimizations is on.
8531         (vect_float_strict): Document.
8532
8533 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
8534
8535         PR tree-optimization/83965
8536         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8537         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8538         instead of checking only for a reduction.
8539         (vect_recog_widen_sum_pattern): Likewise.
8540
8541 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
8542
8543         * predict.c (probably_never_executed): Only use precise profile info.
8544         (compute_function_frequency): Skip after inlining hack since we now
8545         have quality checking.
8546
8547 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
8548
8549         * profile-count.h (profile_probability::very_unlikely,
8550         profile_probability::unlikely, profile_probability::even): Set
8551         precision to guessed.
8552
8553 2018-01-23  Richard Biener  <rguenther@suse.de>
8554
8555         PR tree-optimization/83963
8556         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8557         Properly terminate dominator walk when crossing the exit edge not
8558         when visiting its source block.
8559
8560 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
8561
8562         PR c++/83918
8563         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8564         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8565
8566 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
8567
8568         PR tree-optimization/83957
8569         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
8570         semicolon after for body surrounded by braces.
8571
8572         PR tree-optimization/83081
8573         * profile-count.h (profile_probability::split): New method.
8574         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8575         Use profile_probability::split.
8576         (do_compare_rtx_and_jump): Fix adjustment of probabilities
8577         when splitting a single conditional jump into 2.
8578
8579 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
8580
8581         PR tree-optimization/69452
8582         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8583         decl.
8584
8585 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
8586
8587         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8588         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8589         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8590
8591 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
8592
8593         * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8594         * config/rl78/rl78.md (movdi): New define_expand.
8595         * config/rl78/rl78.c (rl78_split_movdi): New function.
8596
8597 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
8598
8599         PR target/83862
8600         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8601         no longer used.
8602         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8603         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8604         128-bit to produce an UNSPEC move to get the double word with the
8605         signbit and then a shift directly to do signbit.
8606         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8607         implementation with a new version that just does either a direct
8608         move or a regular move.  Move memory interface to separate insns.
8609         Move insns so they are next to the expander.
8610         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8611         with signbit move.  Split big and little endian case.
8612         (signbit<mode>2_dm_mem_le): Likewise.
8613         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8614         (signbit<mode>2_dm2): Likewise.
8615
8616 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
8617
8618         * config/rl78/rl78.md (anddi3): New define_expand.
8619
8620 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
8621
8622         * config/rl78/rl78.md (umindi3): New define_expand.
8623
8624 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
8625
8626         * config/rl78/rl78.md (smindi3): New define_expand.
8627
8628 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
8629
8630         * config/rl78/rl78.md (smaxdi3): New define_expand.
8631
8632 2018-01-22  Carl Love  <cel@us.ibm.com>
8633
8634         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8635         LVX_V1TI): Add macro expansion.
8636         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8637         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8638         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8639         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8640         Change check to determine if the instruction is a byte reversing
8641         entry.  Fix typo in comment.
8642         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8643         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8644         Add def_builtin calls for new builtins.
8645         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8646         Add define_insn expansion.
8647
8648 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
8649
8650         * config/rl78/rl78.md (umaxdi3): New define_expand.
8651
8652 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
8653
8654         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8655         for non-QImode registers.
8656
8657 2018-01-22  Richard Biener  <rguenther@suse.de>
8658
8659         PR tree-optimization/83963
8660         * graphite-scop-detection.c (scop_detection::get_sese): Delay
8661         including the loop exit block.
8662         (scop_detection::merge_sese): Likewise.
8663         (scop_detection::add_scop): Do it here instead.
8664
8665 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8666
8667         * doc/sourcebuild.texi (arm_softfloat): Document.
8668
8669 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
8670
8671         PR gcc/77734
8672         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8673         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8674         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8675
8676 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
8677             David Edelsohn  <dje.gcc@gmail.com>
8678
8679         PR target/83946
8680         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8681         Change "crset eq" to "crset 2".
8682         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8683         (*call_indirect_aix<mode>_nospec): Likewise.
8684         (*call_value_indirect_aix<mode>_nospec): Likewise.
8685         (*call_indirect_elfv2<mode>_nospec): Likewise.
8686         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8687         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8688         change assembly output from . to $.
8689         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8690         (indirect_jump<mode>_nospec): Change assembly output from . to $.
8691         (*tablejump<mode>_internal1_nospec): Likewise.
8692
8693 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
8694
8695         PR target/80870
8696         * config/sh/sh_optimize_sett_clrt.cc:
8697         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8698
8699 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
8700
8701         PR tree-optimization/83940
8702         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8703         offset_dt to vect_constant_def rather than vect_unknown_def_type.
8704         (vect_check_load_store_mask): Add a mask_dt_out parameter and
8705         use it to pass back the definition type.
8706         (vect_check_store_rhs): Likewise rhs_dt_out.
8707         (vect_build_gather_load_calls): Add a mask_dt argument and use
8708         it instead of a call to vect_is_simple_use.
8709         (vectorizable_store): Update calls to vect_check_load_store_mask
8710         and vect_check_store_rhs.  Use the dt returned by the latter instead
8711         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
8712         instead of calls to vect_is_simple_use.  Pass the scalar rather
8713         than the vector operand to vect_is_simple_use when handling
8714         second and subsequent copies of an rhs value.
8715         (vectorizable_load): Update calls to vect_check_load_store_mask
8716         and vect_build_gather_load_calls.  Use the cached mask_dt and
8717         gs_info.offset_dt instead of calls to vect_is_simple_use.
8718
8719 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
8720
8721         PR middle-end/83945
8722         * tree-emutls.c: Include gimplify.h.
8723         (lower_emutls_2): New function.
8724         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8725         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8726         it before further processing.
8727
8728         PR target/83930
8729         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8730         UINTVAL (trueop1) instead of INTVAL (op1).
8731
8732 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
8733
8734         PR debug/81570
8735         PR debug/83728
8736         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8737         INCOMING_FRAME_SP_OFFSET if not defined.
8738         (scan_trace): Add ENTRY argument.  If true and
8739         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8740         emit a note to adjust the CFA offset.
8741         (create_cfi_notes): Adjust scan_trace callers.
8742         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8743         INCOMING_FRAME_SP_OFFSET in the CIE.
8744         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8745         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8746         Likewise.
8747         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8748         * doc/tm.texi: Regenerated.
8749
8750 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
8751
8752         PR rtl-optimization/83147
8753         * lra-constraints.c (remove_inheritance_pseudos): Use
8754         lra_substitute_pseudo_within_insn.
8755
8756 2018-01-19  Tom de Vries  <tom@codesourcery.com>
8757             Cesar Philippidis  <cesar@codesourcery.com>
8758
8759         PR target/83920
8760         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8761
8762 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
8763
8764         PR target/83790
8765         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8766         spaces for function labels.
8767
8768 2018-01-19  Martin Liska  <mliska@suse.cz>
8769
8770         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8771         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8772         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8773         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8774         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8775         (PRED_CONST_RETURN): Change from 69 to 65.
8776         (PRED_NULL_RETURN): Change from 91 to 71.
8777         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8778         (PRED_LOOP_GUARD): Change from 66 to 73.
8779
8780 2018-01-19  Martin Liska  <mliska@suse.cz>
8781
8782         * predict.c (predict_insn_def): Add new assert.
8783         (struct branch_predictor): Change type to signed integer.
8784         (test_prediction_value_range): Amend test to cover
8785         PROB_UNINITIALIZED.
8786         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8787         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8788         (PRED_LOOP_ITERATIONS_MAX): Likewise.
8789         (PRED_LOOP_IV_COMPARE): Likewise.
8790         * predict.h (PROB_UNINITIALIZED): Define new constant.
8791
8792 2018-01-19  Martin Liska  <mliska@suse.cz>
8793
8794         * predict.c (dump_prediction): Add new format for
8795         analyze_brprob.py script which is enabled with -details
8796         suboption.
8797         * profile-count.h (precise_p): New function.
8798
8799 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
8800
8801         PR tree-optimization/83922
8802         * tree-vect-loop.c (vect_verify_full_masking): Return false if
8803         there are no statements that need masking.
8804         (vect_active_double_reduction_p): New function.
8805         (vect_analyze_loop_operations): Use it when handling phis that
8806         are not in the loop header.
8807
8808 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
8809
8810         PR tree-optimization/83914
8811         * tree-vect-loop.c (vectorizable_induction): Don't convert
8812         init_expr or apply the peeling adjustment for inductions
8813         that are nested within the vectorized loop.
8814
8815 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8816
8817         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8818         instead of NEG.
8819
8820 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
8821
8822         PR sanitizer/81715
8823         PR testsuite/83882
8824         * function.h (gimplify_parameters): Add gimple_seq * argument.
8825         * function.c: Include gimple.h and options.h.
8826         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8827         for the added local temporaries if needed.
8828         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8829         if there are any parameter cleanups, wrap whole body into a
8830         try/finally with the cleanups.
8831
8832 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
8833
8834         PR target/82964
8835         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8836         Use GET_MODE_CLASS for scalar floating point.
8837
8838 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
8839
8840         PR ipa/82256
8841         patch by PaX Team
8842         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8843         Fix call of call_cgraph_insertion_hooks.
8844
8845 2018-01-18  Martin Sebor  <msebor@redhat.com>
8846
8847         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8848
8849 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
8850
8851         PR ipa/83619
8852         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8853         frequencies.
8854
8855 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
8856
8857         PR other/70268
8858         * common.opt: (-ffile-prefix-map): New option.
8859         * opts.c (common_handle_option): Defer it.
8860         * opts-global.c (handle_common_deferred_options): Handle it.
8861         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8862         * file-prefix-map.h: New file.
8863         (remap_debug_filename, add_debug_prefix_map): ...here.
8864         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8865         * final.c (debug_prefix_map, add_debug_prefix_map
8866         remap_debug_filename): Move to...
8867         * file-prefix-map.c: New file.
8868         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8869         generalize, get rid of alloca(), use strrchr() instead of strchr().
8870         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8871         Implement in terms of add_prefix_map().
8872         (remap_macro_filename, remap_debug_filename): Implement in term of
8873         remap_filename().
8874         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8875         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8876         * dbxout.c: Include file-prefix-map.h.
8877         * varasm.c: Likewise.
8878         * vmsdbgout.c: Likewise.
8879         * xcoffout.c: Likewise.
8880         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8881         * doc/cppopts.texi (-fmacro-prefix-map): Document.
8882         * doc/invoke.texi (-ffile-prefix-map): Document.
8883         (-fdebug-prefix-map): Update description.
8884
8885 2018-01-18  Martin Liska  <mliska@suse.cz>
8886
8887         * config/i386/i386.c (indirect_thunk_name): Document that also
8888         lfence is emitted.
8889         (output_indirect_thunk): Document why both instructions
8890         (pause and lfence) are generated.
8891
8892 2018-01-18  Richard Biener  <rguenther@suse.de>
8893
8894         PR tree-optimization/83887
8895         * graphite-scop-detection.c
8896         (scop_detection::get_nearest_dom_with_single_entry): Remove.
8897         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8898         (scop_detection::merge_sese): Re-implement with a flood-fill
8899         algorithm that properly finds a SESE region if it exists.
8900
8901 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
8902
8903         PR c/61240
8904         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8905         pointer_diff optimizations use view_convert instead of convert.
8906
8907 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
8908
8909         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8910         Generate different code for -mno-speculate-indirect-jumps.
8911         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8912         (*call_indirect_aix<mode>): Disable for
8913         -mno-speculate-indirect-jumps.
8914         (*call_indirect_aix<mode>_nospec): New define_insn.
8915         (*call_value_indirect_aix<mode>): Disable for
8916         -mno-speculate-indirect-jumps.
8917         (*call_value_indirect_aix<mode>_nospec): New define_insn.
8918         (*sibcall_nonlocal_sysv<mode>): Generate different code for
8919         -mno-speculate-indirect-jumps.
8920         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8921
8922 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
8923
8924         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8925         long double type, set the flags for noting the default long double
8926         type, even if we don't pass or return a long double type.
8927
8928 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
8929
8930         PR ipa/83051
8931         * ipa-inline.c (flatten_function): Do not overwrite final inlining
8932         failure.
8933
8934 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
8935
8936         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8937         support for merge[hl].
8938         (fold_mergehl_helper): New helper function.
8939         (tree-vector-builder.h): New #include for tree_vector_builder usage.
8940         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8941         (altivec_vmrglw_direct): Add xxmrglw insn.
8942
8943 2018-01-17  Andrew Waterman  <andrew@sifive.com>
8944
8945         * config/riscv/riscv.c (riscv_conditional_register_usage): If
8946         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8947
8948 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
8949
8950         PR lto/83121
8951         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8952         call the lto_location_cache before reading the
8953         DECL_SOURCE_LOCATION of the types.
8954
8955 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
8956             Richard Sandiford  <richard.sandiford@linaro.org>
8957
8958         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8959         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8960         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8961         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8962         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8963         Add declaration.
8964         * config/aarch64/constraints.md (aarch64_movti_operand):
8965         Limit immediates.
8966         * config/aarch64/predicates.md (Uti): Add new constraint.
8967
8968 2018-01-17 Carl Love  <cel@us.ibm.com>
8969
8970         * config/rs6000/vsx.md (define_expand xl_len_r,
8971         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8972         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8973         lxvll.
8974         (define_expand, define_insn): Move the shift left from  the
8975         define_insn to the define_expand for lxvl and stxvl instructions.
8976         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8977         and XL_LEN_R definitions to PURE.
8978
8979 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
8980
8981         * config/i386/i386.c (indirect_thunk_name): Declare regno
8982         as unsigned int.  Compare regno with INVALID_REGNUM.
8983         (output_indirect_thunk): Ditto.
8984         (output_indirect_thunk_function): Ditto.
8985         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
8986         in the call to output_indirect_thunk_function.
8987
8988 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
8989
8990         PR middle-end/83884
8991         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8992         rather than the size of inner_type to determine the stack slot size
8993         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8994
8995 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
8996
8997         PR target/83546
8998         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8999         to PTA_SILVERMONT.
9000
9001 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
9002
9003         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9004         endian Linux systems to optionally enable multilibs for selecting
9005         the long double type if the user configured an explicit type.
9006         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9007         have no long double multilibs if not defined.
9008         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9009         warn if the user used -mabi={ieee,ibm}longdouble and we built
9010         multilibs for long double.
9011         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9012         appropriate multilib option.
9013         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9014         multilib options.
9015         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9016         for building long double multilibs.
9017         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9018
9019 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
9020
9021         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9022         copies.
9023
9024         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9025         64 bits.
9026         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9027         128 bits.
9028
9029         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9030         variables.
9031
9032         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9033         return value.
9034
9035 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
9036
9037         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9038         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9039
9040 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9041
9042         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9043         different rtl trees depending on TARGET_64BIT.
9044         (rs6000_gen_lvx): Likewise.
9045
9046 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
9047
9048         * config/visium/visium.md (nop): Tweak comment.
9049         (hazard_nop): Likewise.
9050
9051 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9052
9053         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9054         -mspeculate-indirect-jumps.
9055         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9056         for -mno-speculate-indirect-jumps.
9057         (*call_indirect_elfv2<mode>_nospec): New define_insn.
9058         (*call_value_indirect_elfv2<mode>): Disable for
9059         -mno-speculate-indirect-jumps.
9060         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9061         (indirect_jump): Emit different RTL for
9062         -mno-speculate-indirect-jumps.
9063         (*indirect_jump<mode>): Disable for
9064         -mno-speculate-indirect-jumps.
9065         (*indirect_jump<mode>_nospec): New define_insn.
9066         (tablejump): Emit different RTL for
9067         -mno-speculate-indirect-jumps.
9068         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9069         (tablejumpsi_nospec): New define_expand.
9070         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9071         (tablejumpdi_nospec): New define_expand.
9072         (*tablejump<mode>_internal1): Disable for
9073         -mno-speculate-indirect-jumps.
9074         (*tablejump<mode>_internal1_nospec): New define_insn.
9075         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9076         option.
9077
9078 2018-01-16  Artyom Skrobov tyomitch@gmail.com
9079
9080         * caller-save.c (insert_save): Drop unnecessary parameter.  All
9081         callers updated.
9082
9083 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
9084             Richard Biener  <rguenth@suse.de>
9085
9086         PR libgomp/83590
9087         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9088         return early, inline manually is_gimple_sizepos.  Make sure if we
9089         call gimplify_expr we don't end up with a gimple constant.
9090         * tree.c (variably_modified_type_p): Don't return true for
9091         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
9092         * gimplify.h (is_gimple_sizepos): Remove.
9093
9094 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
9095
9096         PR tree-optimization/83857
9097         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9098         vectorizable_live_operation for pure SLP statements.
9099         (vectorizable_live_operation): Handle PHIs.
9100
9101 2018-01-16  Richard Biener  <rguenther@suse.de>
9102
9103         PR tree-optimization/83867
9104         * tree-vect-stmts.c (vect_transform_stmt): Precompute
9105         nested_in_vect_loop_p since the scalar stmt may get invalidated.
9106
9107 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
9108
9109         PR c/83844
9110         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9111         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9112         If off is not INTEGER_CST, issue a may not be aligned warning
9113         rather than isn't aligned.  Use isn%'t rather than isn't.
9114         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9115         into MULT_EXPR.
9116         <case MULT_EXPR>: Improve the case when bottom and one of the
9117         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9118         operand, in that case check if the other operand is multiple of
9119         bottom divided by the INTEGER_CST operand.
9120
9121 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
9122
9123         PR target/83858
9124         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9125         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9126         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9127         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9128         * config/pa/pa.c (pa_function_arg_advance): Likewise.
9129         (pa_function_arg, pa_arg_partial_bytes): Likewise.
9130         (pa_function_arg_size): New function.
9131
9132 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
9133
9134         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9135         in a separate statement.
9136
9137 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
9138
9139         PR tree-optimization/83847
9140         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9141         group gathers and scatters.
9142
9143 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
9144
9145         PR rtl-optimization/86620
9146         * params.def (max-sched-ready-insns): Bump minimum value to 1.
9147
9148         PR rtl-optimization/83213
9149         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9150         to last if both are JUMP_INSNs.
9151
9152         PR tree-optimization/83843
9153         * gimple-ssa-store-merging.c
9154         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9155         store_immediate_info for bswap/nop orig_stores.
9156
9157 2018-01-15  Andrew Waterman  <andrew@sifive.com>
9158
9159         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9160         !TARGET_MUL.
9161         <UDIV>: Increase cost if !TARGET_DIV.
9162
9163 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
9164
9165         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9166         (define_attr "cr_logical_3op"): New.
9167         (cceq_ior_compare): Adjust.
9168         (cceq_ior_compare_complement): Adjust.
9169         (*cceq_rev_compare): Adjust.
9170         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9171         (is_cracked_insn): Adjust.
9172         (insn_must_be_first_in_group): Adjust.
9173         * config/rs6000/40x.md: Adjust.
9174         * config/rs6000/440.md: Adjust.
9175         * config/rs6000/476.md: Adjust.
9176         * config/rs6000/601.md: Adjust.
9177         * config/rs6000/603.md: Adjust.
9178         * config/rs6000/6xx.md: Adjust.
9179         * config/rs6000/7450.md: Adjust.
9180         * config/rs6000/7xx.md: Adjust.
9181         * config/rs6000/8540.md: Adjust.
9182         * config/rs6000/cell.md: Adjust.
9183         * config/rs6000/e300c2c3.md: Adjust.
9184         * config/rs6000/e500mc.md: Adjust.
9185         * config/rs6000/e500mc64.md: Adjust.
9186         * config/rs6000/e5500.md: Adjust.
9187         * config/rs6000/e6500.md: Adjust.
9188         * config/rs6000/mpc.md: Adjust.
9189         * config/rs6000/power4.md: Adjust.
9190         * config/rs6000/power5.md: Adjust.
9191         * config/rs6000/power6.md: Adjust.
9192         * config/rs6000/power7.md: Adjust.
9193         * config/rs6000/power8.md: Adjust.
9194         * config/rs6000/power9.md: Adjust.
9195         * config/rs6000/rs64.md: Adjust.
9196         * config/rs6000/titan.md: Adjust.
9197
9198 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
9199
9200         * config/i386/predicates.md (indirect_branch_operand): Rewrite
9201         ix86_indirect_branch_register logic.
9202
9203 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
9204
9205         * config/i386/constraints.md (Bs): Update
9206         ix86_indirect_branch_register check.  Don't check
9207         ix86_indirect_branch_register with GOT_memory_operand.
9208         (Bw): Likewise.
9209         * config/i386/predicates.md (GOT_memory_operand): Don't check
9210         ix86_indirect_branch_register here.
9211         (GOT32_symbol_operand): Likewise.
9212
9213 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
9214
9215         * config/i386/predicates.md (constant_call_address_operand):
9216         Rewrite ix86_indirect_branch_register logic.
9217         (sibcall_insn_operand): Likewise.
9218
9219 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
9220
9221         * config/i386/constraints.md (Bs): Replace
9222         ix86_indirect_branch_thunk_register with
9223         ix86_indirect_branch_register.
9224         (Bw): Likewise.
9225         * config/i386/i386.md (indirect_jump): Likewise.
9226         (tablejump): Likewise.
9227         (*sibcall_memory): Likewise.
9228         (*sibcall_value_memory): Likewise.
9229         Peepholes of indirect call and jump via memory: Likewise.
9230         * config/i386/i386.opt: Likewise.
9231         * config/i386/predicates.md (indirect_branch_operand): Likewise.
9232         (GOT_memory_operand): Likewise.
9233         (call_insn_operand): Likewise.
9234         (sibcall_insn_operand): Likewise.
9235         (GOT32_symbol_operand): Likewise.
9236
9237 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
9238
9239         PR middle-end/83837
9240         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9241         type rather than type addr's type points to.
9242         (expand_omp_atomic_mutex): Likewise.
9243         (expand_omp_atomic): Likewise.
9244
9245 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
9246
9247         PR target/83839
9248         * config/i386/i386.c (output_indirect_thunk_function): Use
9249         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9250         for  __x86_return_thunk.
9251
9252 2018-01-15  Richard Biener  <rguenther@suse.de>
9253
9254         PR middle-end/83850
9255         * expmed.c (extract_bit_field_1): Fix typo.
9256
9257 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9258
9259         PR target/83687
9260         * config/arm/iterators.md (VF): New mode iterator.
9261         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9262         Remove integer-related logic from pattern.
9263         (neon_vabd<mode>_3): Likewise.
9264
9265 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
9266
9267         PR middle-end/82694
9268         * common.opt (fstrict-overflow): No longer an alias.
9269         (fwrapv-pointer): New option.
9270         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9271         also for pointer types based on flag_wrapv_pointer.
9272         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9273         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9274         opts->x_flag_wrapv got set.
9275         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9276         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9277         POINTER_TYPE_OVERFLOW_UNDEFINED.
9278         * match.pd: Likewise in address comparison pattern.
9279         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9280
9281 2018-01-15  Richard Biener  <rguenther@suse.de>
9282
9283         PR lto/83804
9284         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9285         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
9286         Reset type names to their identifier if their TYPE_DECL doesn't
9287         have linkage (and thus is used for ODR and devirt).
9288         (save_debug_info_for_decl): Remove.
9289         (save_debug_info_for_type): Likewise.
9290         (add_tree_to_fld_list): Adjust.
9291         * tree-pretty-print.c (dump_generic_node): Make dumping of
9292         type names more robust.
9293
9294 2018-01-15  Richard Biener  <rguenther@suse.de>
9295
9296         * BASE-VER: Bump to 8.0.1.
9297
9298 2018-01-14  Martin Sebor  <msebor@redhat.com>
9299
9300         PR other/83508
9301         * builtins.c (check_access): Avoid warning when the no-warning bit
9302         is set.
9303
9304 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
9305
9306         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9307         * ira-color (allocno_hard_regs_compare): Likewise.
9308
9309 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
9310
9311         PR target/83013
9312         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9313         Use .pushsection/.popsection.
9314
9315 2018-01-14  Martin Sebor  <msebor@redhat.com>
9316
9317         PR c++/81327
9318         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9319
9320 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
9321
9322         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9323         entry from extra_headers.
9324         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9325         extra_headers, make the list bitwise identical to the i?86-*-* one.
9326
9327 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
9328
9329         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9330         -mcmodel=large with -mindirect-branch=thunk,
9331         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9332         -mfunction-return=thunk-extern.
9333         * doc/invoke.texi: Document -mcmodel=large is incompatible with
9334         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9335         -mfunction-return=thunk and -mfunction-return=thunk-extern.
9336
9337 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
9338
9339         * config/i386/i386.c (print_reg): Print the name of the full
9340         integer register without '%'.
9341         (ix86_print_operand): Handle 'V'.
9342          * doc/extend.texi: Document 'V' modifier.
9343
9344 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
9345
9346         * config/i386/constraints.md (Bs): Disallow memory operand for
9347         -mindirect-branch-register.
9348         (Bw): Likewise.
9349         * config/i386/predicates.md (indirect_branch_operand): Likewise.
9350         (GOT_memory_operand): Likewise.
9351         (call_insn_operand): Likewise.
9352         (sibcall_insn_operand): Likewise.
9353         (GOT32_symbol_operand): Likewise.
9354         * config/i386/i386.md (indirect_jump): Call convert_memory_address
9355         for -mindirect-branch-register.
9356         (tablejump): Likewise.
9357         (*sibcall_memory): Likewise.
9358         (*sibcall_value_memory): Likewise.
9359         Disallow peepholes of indirect call and jump via memory for
9360         -mindirect-branch-register.
9361         (*call_pop): Replace m with Bw.
9362         (*call_value_pop): Likewise.
9363         (*sibcall_pop_memory): Replace m with Bs.
9364         * config/i386/i386.opt (mindirect-branch-register): New option.
9365         * doc/invoke.texi: Document -mindirect-branch-register option.
9366
9367 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
9368
9369         * config/i386/i386-protos.h (ix86_output_function_return): New.
9370         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9371         set function_return_type.
9372         (indirect_thunk_name): Add ret_p to indicate thunk for function
9373         return.
9374         (output_indirect_thunk_function): Pass false to
9375         indirect_thunk_name.
9376         (ix86_output_indirect_branch_via_reg): Likewise.
9377         (ix86_output_indirect_branch_via_push): Likewise.
9378         (output_indirect_thunk_function): Create alias for function
9379         return thunk if regno < 0.
9380         (ix86_output_function_return): New function.
9381         (ix86_handle_fndecl_attribute): Handle function_return.
9382         (ix86_attribute_table): Add function_return.
9383         * config/i386/i386.h (machine_function): Add
9384         function_return_type.
9385         * config/i386/i386.md (simple_return_internal): Use
9386         ix86_output_function_return.
9387         (simple_return_internal_long): Likewise.
9388         * config/i386/i386.opt (mfunction-return=): New option.
9389         (indirect_branch): Mention -mfunction-return=.
9390         * doc/extend.texi: Document function_return function attribute.
9391         * doc/invoke.texi: Document -mfunction-return= option.
9392
9393 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
9394
9395         * config/i386/i386-opts.h (indirect_branch): New.
9396         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9397         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9398         with local indirect jump when converting indirect call and jump.
9399         (ix86_set_indirect_branch_type): New.
9400         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9401         (indirectlabelno): New.
9402         (indirect_thunk_needed): Likewise.
9403         (indirect_thunk_bnd_needed): Likewise.
9404         (indirect_thunks_used): Likewise.
9405         (indirect_thunks_bnd_used): Likewise.
9406         (INDIRECT_LABEL): Likewise.
9407         (indirect_thunk_name): Likewise.
9408         (output_indirect_thunk): Likewise.
9409         (output_indirect_thunk_function): Likewise.
9410         (ix86_output_indirect_branch_via_reg): Likewise.
9411         (ix86_output_indirect_branch_via_push): Likewise.
9412         (ix86_output_indirect_branch): Likewise.
9413         (ix86_output_indirect_jmp): Likewise.
9414         (ix86_code_end): Call output_indirect_thunk_function if needed.
9415         (ix86_output_call_insn): Call ix86_output_indirect_branch if
9416         needed.
9417         (ix86_handle_fndecl_attribute): Handle indirect_branch.
9418         (ix86_attribute_table): Add indirect_branch.
9419         * config/i386/i386.h (machine_function): Add indirect_branch_type
9420         and has_local_indirect_jump.
9421         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9422         to true.
9423         (tablejump): Likewise.
9424         (*indirect_jump): Use ix86_output_indirect_jmp.
9425         (*tablejump_1): Likewise.
9426         (simple_return_indirect_internal): Likewise.
9427         * config/i386/i386.opt (mindirect-branch=): New option.
9428         (indirect_branch): New.
9429         (keep): Likewise.
9430         (thunk): Likewise.
9431         (thunk-inline): Likewise.
9432         (thunk-extern): Likewise.
9433         * doc/extend.texi: Document indirect_branch function attribute.
9434         * doc/invoke.texi: Document -mindirect-branch= option.
9435
9436 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
9437
9438         PR ipa/83051
9439         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9440
9441 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
9442
9443         * ipa-inline.c (want_inline_small_function_p): Return false if
9444         inlining has already failed with CIF_FINAL_ERROR.
9445         (update_caller_keys): Call want_inline_small_function_p before
9446         can_inline_edge_p.
9447         (update_callee_keys): Likewise.
9448
9449 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9450
9451         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9452         New function.
9453         (rs6000_quadword_masked_address_p): Likewise.
9454         (quad_aligned_load_p): Likewise.
9455         (quad_aligned_store_p): Likewise.
9456         (const_load_sequence_p): Add comment to describe the outer-most loop.
9457         (mimic_memory_attributes_and_flags): New function.
9458         (rs6000_gen_stvx): Likewise.
9459         (replace_swapped_aligned_store): Likewise.
9460         (rs6000_gen_lvx): Likewise.
9461         (replace_swapped_aligned_load): Likewise.
9462         (replace_swapped_load_constant): Capitalize argument name in
9463         comment describing this function.
9464         (rs6000_analyze_swaps): Add a third pass to search for vector loads
9465         and stores that access quad-word aligned addresses and replace
9466         with stvx or lvx instructions when appropriate.
9467         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9468         New function prototype.
9469         (rs6000_quadword_masked_address_p): Likewise.
9470         (rs6000_gen_lvx): Likewise.
9471         (rs6000_gen_stvx): Likewise.
9472         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9473         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9474         when memory address is aligned.
9475         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9476         this split to select lvx instruction when memory address is aligned.
9477         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9478         instruction when memory address is aligned.
9479         (*vsx_le_perm_load_v16qi): Likewise.
9480         (four unnamed splitters): Modify to select the stvx instruction
9481         when memory is aligned.
9482
9483 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
9484
9485         * predict.c (determine_unlikely_bbs): Handle correctly BBs
9486         which appears in the queue multiple times.
9487
9488 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9489             Alan Hayward  <alan.hayward@arm.com>
9490             David Sherwood  <david.sherwood@arm.com>
9491
9492         * tree-vectorizer.h (vec_lower_bound): New structure.
9493         (_loop_vec_info): Add check_nonzero and lower_bounds.
9494         (LOOP_VINFO_CHECK_NONZERO): New macro.
9495         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9496         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9497         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9498         fields.  Make seg_len the distance travelled, not including the
9499         access size.
9500         (dr_direction_indicator): Declare.
9501         (dr_zero_step_indicator): Likewise.
9502         (dr_known_forward_stride_p): Likewise.
9503         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9504         tree-ssanames.h.
9505         (runtime_alias_check_p): Allow runtime alias checks with
9506         variable strides.
9507         (operator ==): Compare access_size and align.
9508         (prune_runtime_alias_test_list): Rework for new distinction between
9509         the access_size and seg_len.
9510         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
9511         segment lengths.
9512         (get_segment_min_max): New function.
9513         (create_intersect_range_checks): Use it.
9514         (dr_step_indicator): New function.
9515         (dr_direction_indicator): Likewise.
9516         (dr_zero_step_indicator): Likewise.
9517         (dr_known_forward_stride_p): Likewise.
9518         * tree-loop-distribution.c (data_ref_segment_size): Return
9519         DR_STEP * (niters - 1).
9520         (compute_alias_check_pairs): Update call to the dr_with_seg_len
9521         constructor.
9522         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9523         (vect_preserves_scalar_order_p): New function, split out from...
9524         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
9525         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9526         (vect_vfa_access_size): New function.
9527         (vect_vfa_align): Likewise.
9528         (vect_compile_time_alias): Take access_size_a and access_b arguments.
9529         (dump_lower_bound): New function.
9530         (vect_check_lower_bound): Likewise.
9531         (vect_small_gap_p): Likewise.
9532         (vectorizable_with_step_bound_p): Likewise.
9533         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9534         depencies if the vectorization factor is 1.  Convert the checks
9535         for nonzero steps into checks on the bounds of DR_STEP.  Try using
9536         a bunds check for variable steps if the minimum required step is
9537         relatively small. Update calls to the dr_with_seg_len
9538         constructor and to vect_compile_time_alias.
9539         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9540         function.
9541         (vect_loop_versioning): Call it.
9542         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9543         when retrying.
9544         (vect_estimate_min_profitable_iters): Account for any bounds checks.
9545
9546 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9547             Alan Hayward  <alan.hayward@arm.com>
9548             David Sherwood  <david.sherwood@arm.com>
9549
9550         * doc/sourcebuild.texi (vect_scatter_store): Document.
9551         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9552         optabs.
9553         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9554         Document.
9555         * genopinit.c (main): Add supports_vec_scatter_store and
9556         supports_vec_scatter_store_cached to target_optabs.
9557         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9558         IFN_MASK_SCATTER_STORE.
9559         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9560         functions.
9561         * internal-fn.h (internal_store_fn_p): Declare.
9562         (internal_fn_stored_value_index): Likewise.
9563         * internal-fn.c (scatter_store_direct): New macro.
9564         (expand_scatter_store_optab_fn): New function.
9565         (direct_scatter_store_optab_supported_p): New macro.
9566         (internal_store_fn_p): New function.
9567         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9568         IFN_MASK_SCATTER_STORE.
9569         (internal_fn_mask_index): Likewise.
9570         (internal_fn_stored_value_index): New function.
9571         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9572         for scatter stores.
9573         * optabs-query.h (supports_vec_scatter_store_p): Declare.
9574         * optabs-query.c (supports_vec_scatter_store_p): New function.
9575         * tree-vectorizer.h (vect_get_store_rhs): Declare.
9576         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9577         true for scatter stores.
9578         (vect_gather_scatter_fn_p): Handle scatter stores too.
9579         (vect_check_gather_scatter): Consider using scatter stores if
9580         supports_vec_scatter_store_p.
9581         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9582         scatter stores too.
9583         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9584         internal_fn_stored_value_index.
9585         (check_load_store_masking): Handle scatter stores too.
9586         (vect_get_store_rhs): Make public.
9587         (vectorizable_call): Use internal_store_fn_p.
9588         (vectorizable_store): Handle scatter store internal functions.
9589         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9590         when deciding whether the end of the group has been reached.
9591         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9592         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9593         (mask_scatter_store<mode>): New insns.
9594
9595 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9596             Alan Hayward  <alan.hayward@arm.com>
9597             David Sherwood  <david.sherwood@arm.com>
9598
9599         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9600         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9601         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9602         function.
9603         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9604         Use vect_truncate_gather_scatter_offset if we can't treat the
9605         operation as a normal gather load or scatter store.
9606         (get_group_load_store_type): Take the gather_scatter_info
9607         as argument.  Try using a gather load or scatter store for
9608         single-element groups.
9609         (get_load_store_type): Update calls to get_group_load_store_type
9610         and vect_use_strided_gather_scatters_p.
9611
9612 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9613             Alan Hayward  <alan.hayward@arm.com>
9614             David Sherwood  <david.sherwood@arm.com>
9615
9616         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9617         optional tree argument.
9618         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9619         null target hooks.
9620         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9621         but continue to use the current value as a fallback.
9622         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9623         to compare the updates.
9624         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9625         (get_load_store_type): Use it when handling a strided access.
9626         (vect_get_strided_load_store_ops): New function.
9627         (vect_get_data_ptr_increment): Likewise.
9628         (vectorizable_load): Handle strided gather loads.  Always pass
9629         a step to vect_create_data_ref_ptr and bump_vector_ptr.
9630
9631 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9632             Alan Hayward  <alan.hayward@arm.com>
9633             David Sherwood  <david.sherwood@arm.com>
9634
9635         * doc/md.texi (gather_load@var{m}): Document.
9636         (mask_gather_load@var{m}): Likewise.
9637         * genopinit.c (main): Add supports_vec_gather_load and
9638         supports_vec_gather_load_cached to target_optabs.
9639         * optabs-tree.c (init_tree_optimization_optabs): Use
9640         ggc_cleared_alloc to allocate target_optabs.
9641         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9642         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9643         functions.
9644         * internal-fn.h (internal_load_fn_p): Declare.
9645         (internal_gather_scatter_fn_p): Likewise.
9646         (internal_fn_mask_index): Likewise.
9647         (internal_gather_scatter_fn_supported_p): Likewise.
9648         * internal-fn.c (gather_load_direct): New macro.
9649         (expand_gather_load_optab_fn): New function.
9650         (direct_gather_load_optab_supported_p): New macro.
9651         (direct_internal_fn_optab): New function.
9652         (internal_load_fn_p): Likewise.
9653         (internal_gather_scatter_fn_p): Likewise.
9654         (internal_fn_mask_index): Likewise.
9655         (internal_gather_scatter_fn_supported_p): Likewise.
9656         * optabs-query.c (supports_at_least_one_mode_p): New function.
9657         (supports_vec_gather_load_p): Likewise.
9658         * optabs-query.h (supports_vec_gather_load_p): Declare.
9659         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9660         and memory_type field.
9661         (NUM_PATTERNS): Bump to 15.
9662         * tree-vect-data-refs.c: Include internal-fn.h.
9663         (vect_gather_scatter_fn_p): New function.
9664         (vect_describe_gather_scatter_call): Likewise.
9665         (vect_check_gather_scatter): Try using internal functions for
9666         gather loads.  Recognize existing calls to a gather load function.
9667         (vect_analyze_data_refs): Consider using gather loads if
9668         supports_vec_gather_load_p.
9669         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9670         (vect_get_gather_scatter_offset_type): Likewise.
9671         (vect_convert_mask_for_vectype): Likewise.
9672         (vect_add_conversion_to_patterm): Likewise.
9673         (vect_try_gather_scatter_pattern): Likewise.
9674         (vect_recog_gather_scatter_pattern): New pattern recognizer.
9675         (vect_vect_recog_func_ptrs): Add it.
9676         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9677         internal_fn_mask_index and internal_gather_scatter_fn_p.
9678         (check_load_store_masking): Take the gather_scatter_info as an
9679         argument and handle gather loads.
9680         (vect_get_gather_scatter_ops): New function.
9681         (vectorizable_call): Check internal_load_fn_p.
9682         (vectorizable_load): Likewise.  Handle gather load internal
9683         functions.
9684         (vectorizable_store): Update call to check_load_store_masking.
9685         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9686         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9687         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9688         (aarch64_gather_scale_operand_d): New predicates.
9689         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9690         (mask_gather_load<mode>): New insns.
9691
9692 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9693             Alan Hayward  <alan.hayward@arm.com>
9694             David Sherwood  <david.sherwood@arm.com>
9695
9696         * optabs.def (fold_left_plus_optab): New optab.
9697         * doc/md.texi (fold_left_plus_@var{m}): Document.
9698         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9699         * internal-fn.c (fold_left_direct): Define.
9700         (expand_fold_left_optab_fn): Likewise.
9701         (direct_fold_left_optab_supported_p): Likewise.
9702         * fold-const-call.c (fold_const_fold_left): New function.
9703         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9704         * tree-parloops.c (valid_reduction_p): New function.
9705         (gather_scalar_reductions): Use it.
9706         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9707         (vect_finish_replace_stmt): Declare.
9708         * tree-vect-loop.c (fold_left_reduction_fn): New function.
9709         (needs_fold_left_reduction_p): New function, split out from...
9710         (vect_is_simple_reduction): ...here.  Accept reductions that
9711         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9712         (vect_force_simple_reduction): Also store the reduction type in
9713         the assignment's STMT_VINFO_REDUC_TYPE.
9714         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9715         (merge_with_identity): New function.
9716         (vect_expand_fold_left): Likewise.
9717         (vectorize_fold_left_reduction): Likewise.
9718         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
9719         scalar phi in place for it.  Check for target support and reject
9720         cases that would reassociate the operation.  Defer the transform
9721         phase to vectorize_fold_left_reduction.
9722         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9723         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9724         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9725
9726 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9727
9728         * tree-if-conv.c (predicate_mem_writes): Remove redundant
9729         call to ifc_temp_var.
9730
9731 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9732             Alan Hayward  <alan.hayward@arm.com>
9733             David Sherwood  <david.sherwood@arm.com>
9734
9735         * target.def (legitimize_address_displacement): Take the original
9736         offset as a poly_int.
9737         * targhooks.h (default_legitimize_address_displacement): Update
9738         accordingly.
9739         * targhooks.c (default_legitimize_address_displacement): Likewise.
9740         * doc/tm.texi: Regenerate.
9741         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9742         as an argument, moving assert of ad->disp == ad->disp_term to...
9743         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
9744         Try calling targetm.legitimize_address_displacement before expanding
9745         the address rather than afterwards, and adjust for the new interface.
9746         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9747         Match the new hook interface.  Handle SVE addresses.
9748         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9749         new hook interface.
9750
9751 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9752
9753         * Makefile.in (OBJS): Add early-remat.o.
9754         * target.def (select_early_remat_modes): New hook.
9755         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9756         * doc/tm.texi: Regenerate.
9757         * targhooks.h (default_select_early_remat_modes): Declare.
9758         * targhooks.c (default_select_early_remat_modes): New function.
9759         * timevar.def (TV_EARLY_REMAT): New timevar.
9760         * passes.def (pass_early_remat): New pass.
9761         * tree-pass.h (make_pass_early_remat): Declare.
9762         * early-remat.c: New file.
9763         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9764         function.
9765         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9766
9767 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9768             Alan Hayward  <alan.hayward@arm.com>
9769             David Sherwood  <david.sherwood@arm.com>
9770
9771         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9772         vfm1 with a bound_epilog parameter.
9773         (vect_do_peeling): Update calls accordingly, and move the prologue
9774         call earlier in the function.  Treat the base bound_epilog as 0 for
9775         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
9776         this base when peeling for gaps.
9777         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9778         with fully-masked loops.
9779         (vect_estimate_min_profitable_iters): Handle the single peeled
9780         iteration in that case.
9781
9782 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9783             Alan Hayward  <alan.hayward@arm.com>
9784             David Sherwood  <david.sherwood@arm.com>
9785
9786         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9787         single-element interleaving even if the size is not a power of 2.
9788         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9789         accesses for single-element interleaving if the group size is
9790         not a power of 2.
9791
9792 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9793             Alan Hayward  <alan.hayward@arm.com>
9794             David Sherwood  <david.sherwood@arm.com>
9795
9796         * doc/md.texi (fold_extract_last_@var{m}): Document.
9797         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9798         * optabs.def (fold_extract_last_optab): New optab.
9799         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9800         * internal-fn.c (fold_extract_direct): New macro.
9801         (expand_fold_extract_optab_fn): Likewise.
9802         (direct_fold_extract_optab_supported_p): Likewise.
9803         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9804         * tree-vect-loop.c (vect_model_reduction_cost): Handle
9805         EXTRACT_LAST_REDUCTION.
9806         (get_initial_def_for_reduction): Do not create an initial vector
9807         for EXTRACT_LAST_REDUCTION reductions.
9808         (vectorizable_reduction): Leave the scalar phi in place for
9809         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
9810         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
9811         epilogue code for EXTRACT_LAST_REDUCTION and defer the
9812         transform phase to vectorizable_condition.
9813         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9814         split out from...
9815         (vect_finish_stmt_generation): ...here.
9816         (vect_finish_replace_stmt): New function.
9817         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9818         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9819         pattern.
9820         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9821
9822 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9823             Alan Hayward  <alan.hayward@arm.com>
9824             David Sherwood  <david.sherwood@arm.com>
9825
9826         * doc/md.texi (extract_last_@var{m}): Document.
9827         * optabs.def (extract_last_optab): New optab.
9828         * internal-fn.def (EXTRACT_LAST): New internal function.
9829         * internal-fn.c (cond_unary_direct): New macro.
9830         (expand_cond_unary_optab_fn): Likewise.
9831         (direct_cond_unary_optab_supported_p): Likewise.
9832         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9833         loops using EXTRACT_LAST.
9834         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9835         (extract_last_<mode>): ...this optab.
9836         (vec_extract<mode><Vel>): Update accordingly.
9837
9838 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9839             Alan Hayward  <alan.hayward@arm.com>
9840             David Sherwood  <david.sherwood@arm.com>
9841
9842         * target.def (empty_mask_is_expensive): New hook.
9843         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9844         * doc/tm.texi: Regenerate.
9845         * targhooks.h (default_empty_mask_is_expensive): Declare.
9846         * targhooks.c (default_empty_mask_is_expensive): New function.
9847         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9848         if the target says that empty masks are expensive.
9849         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9850         New function.
9851         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9852
9853 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9854             Alan Hayward  <alan.hayward@arm.com>
9855             David Sherwood  <david.sherwood@arm.com>
9856
9857         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9858         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9859         (vect_use_loop_mask_for_alignment_p): New function.
9860         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9861         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9862         niters_skip argument.  Make sure that the first niters_skip elements
9863         of the first iteration are inactive.
9864         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9865         Update call to vect_set_loop_masks_directly.
9866         (get_misalign_in_elems): New function, split out from...
9867         (vect_gen_prolog_loop_niters): ...here.
9868         (vect_update_init_of_dr): Take a code argument that specifies whether
9869         the adjustment should be added or subtracted.
9870         (vect_update_init_of_drs): Likewise.
9871         (vect_prepare_for_masked_peels): New function.
9872         (vect_do_peeling): Skip prologue peeling if we're using a mask
9873         instead.  Update call to vect_update_inits_of_drs.
9874         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9875         mask_skip_niters.
9876         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9877         alignment.  Do not include the number of peeled iterations in
9878         the minimum threshold in that case.
9879         (vectorizable_induction): Adjust the start value down by
9880         LOOP_VINFO_MASK_SKIP_NITERS iterations.
9881         (vect_transform_loop): Call vect_prepare_for_masked_peels.
9882         Take the number of skipped iterations into account when calculating
9883         the loop bounds.
9884         * tree-vect-stmts.c (vect_gen_while_not): New function.
9885
9886 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9887             Alan Hayward  <alan.hayward@arm.com>
9888             David Sherwood  <david.sherwood@arm.com>
9889
9890         * doc/sourcebuild.texi (vect_fully_masked): Document.
9891         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9892         default value to 0.
9893         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9894         split out from...
9895         (vect_analyze_loop_2): ...here. Don't check the vectorization
9896         factor against the number of loop iterations if the loop is
9897         fully-masked.
9898
9899 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9900             Alan Hayward  <alan.hayward@arm.com>
9901             David Sherwood  <david.sherwood@arm.com>
9902
9903         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9904         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9905         (dump_groups): Update accordingly.
9906         (iv_use::mem_type): New member variable.
9907         (address_p): New function.
9908         (record_use): Add a mem_type argument and initialize the new
9909         mem_type field.
9910         (record_group_use): Add a mem_type argument.  Use address_p.
9911         Remove obsolete null checks of base_object.  Update call to record_use.
9912         (find_interesting_uses_op): Update call to record_group_use.
9913         (find_interesting_uses_cond): Likewise.
9914         (find_interesting_uses_address): Likewise.
9915         (get_mem_type_for_internal_fn): New function.
9916         (find_address_like_use): Likewise.
9917         (find_interesting_uses_stmt): Try find_address_like_use before
9918         calling find_interesting_uses_op.
9919         (addr_offset_valid_p): Use the iv mem_type field as the type
9920         of the addressed memory.
9921         (add_autoinc_candidates): Likewise.
9922         (get_address_cost): Likewise.
9923         (split_small_address_groups_p): Use address_p.
9924         (split_address_groups): Likewise.
9925         (add_iv_candidate_for_use): Likewise.
9926         (autoinc_possible_for_pair): Likewise.
9927         (rewrite_groups): Likewise.
9928         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9929         (determine_group_iv_cost): Update after split of USE_ADDRESS.
9930         (get_alias_ptr_type_for_ptr_address): New function.
9931         (rewrite_use_address): Rewrite address uses in calls that were
9932         identified by find_address_like_use.
9933
9934 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9935             Alan Hayward  <alan.hayward@arm.com>
9936             David Sherwood  <david.sherwood@arm.com>
9937
9938         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9939         TARGET_MEM_REFs.
9940         * gimple-expr.h (is_gimple_addressable: Likewise.
9941         * gimple-expr.c (is_gimple_address): Likewise.
9942         * internal-fn.c (expand_call_mem_ref): New function.
9943         (expand_mask_load_optab_fn): Use it.
9944         (expand_mask_store_optab_fn): Likewise.
9945
9946 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9947             Alan Hayward  <alan.hayward@arm.com>
9948             David Sherwood  <david.sherwood@arm.com>
9949
9950         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9951         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9952         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9953         (cond_umax@var{mode}): Document.
9954         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9955         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9956         (cond_umin_optab, cond_umax_optab): New optabs.
9957         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9958         (COND_IOR, COND_XOR): New internal functions.
9959         * internal-fn.h (get_conditional_internal_fn): Declare.
9960         * internal-fn.c (cond_binary_direct): New macro.
9961         (expand_cond_binary_optab_fn): Likewise.
9962         (direct_cond_binary_optab_supported_p): Likewise.
9963         (get_conditional_internal_fn): New function.
9964         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9965         Cope with reduction statements that are vectorized as calls rather
9966         than assignments.
9967         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9968         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9969         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9970         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9971         (UNSPEC_COND_EOR): New unspecs.
9972         (optab): Add mappings for them.
9973         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9974         (sve_int_op, sve_fp_op): New int attributes.
9975
9976 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
9977             Alan Hayward  <alan.hayward@arm.com>
9978             David Sherwood  <david.sherwood@arm.com>
9979
9980         * optabs.def (while_ult_optab): New optab.
9981         * doc/md.texi (while_ult@var{m}@var{n}): Document.
9982         * internal-fn.def (WHILE_ULT): New internal function.
9983         * internal-fn.h (direct_internal_fn_supported_p): New override
9984         that takes two types as argument.
9985         * internal-fn.c (while_direct): New macro.
9986         (expand_while_optab_fn): New function.
9987         (convert_optab_supported_p): Likewise.
9988         (direct_while_optab_supported_p): New macro.
9989         * wide-int.h (wi::udiv_ceil): New function.
9990         * tree-vectorizer.h (rgroup_masks): New structure.
9991         (vec_loop_masks): New typedef.
9992         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9993         and fully_masked_p.
9994         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9995         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9996         (vect_max_vf): New function.
9997         (slpeel_make_loop_iterate_ntimes): Delete.
9998         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9999         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10000         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10001         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10002         internal-fn.h, stor-layout.h and optabs-query.h.
10003         (vect_set_loop_mask): New function.
10004         (add_preheader_seq): Likewise.
10005         (add_header_seq): Likewise.
10006         (interleave_supported_p): Likewise.
10007         (vect_maybe_permute_loop_masks): Likewise.
10008         (vect_set_loop_masks_directly): Likewise.
10009         (vect_set_loop_condition_masked): Likewise.
10010         (vect_set_loop_condition_unmasked): New function, split out from
10011         slpeel_make_loop_iterate_ntimes.
10012         (slpeel_make_loop_iterate_ntimes): Rename to..
10013         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
10014         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10015         (vect_do_peeling): Update call accordingly.
10016         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10017         loops.
10018         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10019         mask_compare_type, can_fully_mask_p and fully_masked_p.
10020         (release_vec_loop_masks): New function.
10021         (_loop_vec_info): Use it to free the loop masks.
10022         (can_produce_all_loop_masks_p): New function.
10023         (vect_get_max_nscalars_per_iter): Likewise.
10024         (vect_verify_full_masking): Likewise.
10025         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10026         retries, and free the mask rgroups before retrying.  Check loop-wide
10027         reasons for disallowing fully-masked loops.  Make the final decision
10028         about whether use a fully-masked loop or not.
10029         (vect_estimate_min_profitable_iters): Do not assume that peeling
10030         for the number of iterations will be needed for fully-masked loops.
10031         (vectorizable_reduction): Disable fully-masked loops.
10032         (vectorizable_live_operation): Likewise.
10033         (vect_halve_mask_nunits): New function.
10034         (vect_double_mask_nunits): Likewise.
10035         (vect_record_loop_mask): Likewise.
10036         (vect_get_loop_mask): Likewise.
10037         (vect_transform_loop): Handle the case in which the final loop
10038         iteration might handle a partial vector.  Call vect_set_loop_condition
10039         instead of slpeel_make_loop_iterate_ntimes.
10040         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10041         (check_load_store_masking): New function.
10042         (prepare_load_store_mask): Likewise.
10043         (vectorizable_store): Handle fully-masked loops.
10044         (vectorizable_load): Likewise.
10045         (supportable_widening_operation): Use vect_halve_mask_nunits for
10046         booleans.
10047         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10048         (vect_gen_while): New function.
10049         * config/aarch64/aarch64.md (umax<mode>3): New expander.
10050         (aarch64_uqdec<mode>): New insn.
10051
10052 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10053             Alan Hayward  <alan.hayward@arm.com>
10054             David Sherwood  <david.sherwood@arm.com>
10055
10056         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10057         (reduc_xor_scal_optab): New optabs.
10058         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10059         (reduc_xor_scal_@var{m}): Document.
10060         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10061         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10062         internal functions.
10063         * fold-const-call.c (fold_const_call): Handle them.
10064         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10065         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10066         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10067         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10068         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10069         (UNSPEC_XORV): New unspecs.
10070         (optab): Add entries for them.
10071         (BITWISEV): New int iterator.
10072         (bit_reduc_op): New int attributes.
10073
10074 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10075             Alan Hayward  <alan.hayward@arm.com>
10076             David Sherwood  <david.sherwood@arm.com>
10077
10078         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10079         * internal-fn.def (VEC_SHL_INSERT): New internal function.
10080         * optabs.def (vec_shl_insert_optab): New optab.
10081         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10082         (duplicate_and_interleave): Likewise.
10083         * tree-vect-loop.c: Include internal-fn.h.
10084         (neutral_op_for_slp_reduction): New function, split out from
10085         get_initial_defs_for_reduction.
10086         (get_initial_def_for_reduction): Handle option 2 for variable-length
10087         vectors by loading the neutral value into a vector and then shifting
10088         the initial value into element 0.
10089         (get_initial_defs_for_reduction): Replace the code argument with
10090         the neutral value calculated by neutral_op_for_slp_reduction.
10091         Use gimple_build_vector for constant-length vectors.
10092         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10093         but the first group_size elements have a neutral value.
10094         Use duplicate_and_interleave otherwise.
10095         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10096         Update call to get_initial_defs_for_reduction.  Handle SLP
10097         reductions for variable-length vectors by creating one vector
10098         result for each scalar result, with the elements associated
10099         with other scalar results stubbed out with the neutral value.
10100         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10101         Require IFN_VEC_SHL_INSERT for double reductions on
10102         variable-length vectors, or SLP reductions that have
10103         a neutral value.  Require can_duplicate_and_interleave_p
10104         support for variable-length unchained SLP reductions if there
10105         is no neutral value, such as for MIN/MAX reductions.  Also require
10106         the number of vector elements to be a multiple of the number of
10107         SLP statements when doing variable-length unchained SLP reductions.
10108         Update call to vect_create_epilog_for_reduction.
10109         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10110         and remove initial values.
10111         (duplicate_and_interleave): Make public.
10112         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10113         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10114
10115 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10116             Alan Hayward  <alan.hayward@arm.com>
10117             David Sherwood  <david.sherwood@arm.com>
10118
10119         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10120         (can_duplicate_and_interleave_p): New function.
10121         (vect_get_and_check_slp_defs): Take the vector of statements
10122         rather than just the current one.  Remove excess parentheses.
10123         Restriction rejectinon of vect_constant_def and vect_external_def
10124         for variable-length vectors to boolean types, or types for which
10125         can_duplicate_and_interleave_p is false.
10126         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10127         (duplicate_and_interleave): New function.
10128         (vect_get_constant_vectors): Use gimple_build_vector for
10129         constant-length vectors and suitable variable-length constant
10130         vectors.  Use duplicate_and_interleave for other variable-length
10131         vectors.  Don't defer the update when inserting new statements.
10132
10133 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10134             Alan Hayward  <alan.hayward@arm.com>
10135             David Sherwood  <david.sherwood@arm.com>
10136
10137         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10138         min_profitable_iters doesn't go negative.
10139
10140 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10141             Alan Hayward  <alan.hayward@arm.com>
10142             David Sherwood  <david.sherwood@arm.com>
10143
10144         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10145         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10146         * optabs.def (vec_mask_load_lanes_optab): New optab.
10147         (vec_mask_store_lanes_optab): Likewise.
10148         * internal-fn.def (MASK_LOAD_LANES): New internal function.
10149         (MASK_STORE_LANES): Likewise.
10150         * internal-fn.c (mask_load_lanes_direct): New macro.
10151         (mask_store_lanes_direct): Likewise.
10152         (expand_mask_load_optab_fn): Handle masked operations.
10153         (expand_mask_load_lanes_optab_fn): New macro.
10154         (expand_mask_store_optab_fn): Handle masked operations.
10155         (expand_mask_store_lanes_optab_fn): New macro.
10156         (direct_mask_load_lanes_optab_supported_p): Likewise.
10157         (direct_mask_store_lanes_optab_supported_p): Likewise.
10158         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10159         parameter.
10160         (vect_load_lanes_supported): Likewise.
10161         * tree-vect-data-refs.c (strip_conversion): New function.
10162         (can_group_stmts_p): Likewise.
10163         (vect_analyze_data_ref_accesses): Use it instead of checking
10164         for a pair of assignments.
10165         (vect_store_lanes_supported): Take a masked_p parameter.
10166         (vect_load_lanes_supported): Likewise.
10167         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10168         vect_store_lanes_supported and vect_load_lanes_supported.
10169         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10170         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10171         parameter.  Don't allow gaps for masked accesses.
10172         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
10173         and vect_load_lanes_supported.
10174         (get_load_store_type): Take a masked_p parameter and update
10175         call to get_group_load_store_type.
10176         (vectorizable_store): Update call to get_load_store_type.
10177         Handle IFN_MASK_STORE_LANES.
10178         (vectorizable_load): Update call to get_load_store_type.
10179         Handle IFN_MASK_LOAD_LANES.
10180
10181 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10182             Alan Hayward  <alan.hayward@arm.com>
10183             David Sherwood  <david.sherwood@arm.com>
10184
10185         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10186         modes for SVE.
10187         * config/aarch64/aarch64-protos.h
10188         (aarch64_sve_struct_memory_operand_p): Declare.
10189         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10190         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10191         (VPRED, vpred): Handle SVE structure modes.
10192         * config/aarch64/constraints.md (Utx): New constraint.
10193         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10194         (aarch64_sve_struct_nonimmediate_operand): New predicates.
10195         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10196         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10197         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10198         structure modes.  Split into pieces after RA.
10199         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10200         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10201         New patterns.
10202         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10203         SVE structure modes.
10204         (aarch64_classify_address): Likewise.
10205         (sizetochar): Move earlier in file.
10206         (aarch64_print_operand): Handle SVE register lists.
10207         (aarch64_array_mode): New function.
10208         (aarch64_sve_struct_memory_operand_p): Likewise.
10209         (TARGET_ARRAY_MODE): Redefine.
10210
10211 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10212             Alan Hayward  <alan.hayward@arm.com>
10213             David Sherwood  <david.sherwood@arm.com>
10214
10215         * target.def (array_mode): New target hook.
10216         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10217         * doc/tm.texi: Regenerate.
10218         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10219         * hooks.c (hook_optmode_mode_uhwi_none): New function.
10220         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10221         targetm.array_mode.
10222         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
10223         type sizes.
10224
10225 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10226             Alan Hayward  <alan.hayward@arm.com>
10227             David Sherwood  <david.sherwood@arm.com>
10228
10229         * fold-const.c (fold_binary_loc): Check the argument types
10230         rather than the result type when testing for a vector operation.
10231
10232 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10233
10234         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10235         * doc/tm.texi: Regenerate.
10236
10237 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10238             Alan Hayward  <alan.hayward@arm.com>
10239             David Sherwood  <david.sherwood@arm.com>
10240
10241         * doc/invoke.texi (-msve-vector-bits=): Document new option.
10242         (sve): Document new AArch64 extension.
10243         * doc/md.texi (w): Extend the description of the AArch64
10244         constraint to include SVE vectors.
10245         (Upl, Upa): Document new AArch64 predicate constraints.
10246         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10247         enum.
10248         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10249         (msve-vector-bits=): New option.
10250         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10251         SVE when these are disabled.
10252         (sve): New extension.
10253         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10254         modes.  Adjust their number of units based on aarch64_sve_vg.
10255         (MAX_BITSIZE_MODE_ANY_MODE): Define.
10256         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10257         aarch64_addr_query_type.
10258         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10259         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10260         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10261         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10262         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10263         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10264         (aarch64_simd_imm_zero_p): Delete.
10265         (aarch64_check_zero_based_sve_index_immediate): Declare.
10266         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10267         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10268         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10269         (aarch64_sve_float_mul_immediate_p): Likewise.
10270         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10271         rather than an rtx.
10272         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10273         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10274         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10275         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10276         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10277         (aarch64_regmode_natural_size): Likewise.
10278         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10279         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10280         left one place.
10281         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10282         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10283         for VG and the SVE predicate registers.
10284         (V_ALIASES): Add a "z"-prefixed alias.
10285         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10286         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10287         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10288         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10289         (REG_CLASS_NAMES): Add entries for them.
10290         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
10291         and the predicate registers.
10292         (aarch64_sve_vg): Declare.
10293         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10294         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10295         (REGMODE_NATURAL_SIZE): Define.
10296         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10297         SVE macros.
10298         * config/aarch64/aarch64.c: Include cfgrtl.h.
10299         (simd_immediate_info): Add a constructor for series vectors,
10300         and an associated step field.
10301         (aarch64_sve_vg): New variable.
10302         (aarch64_dbx_register_number): Handle VG and the predicate registers.
10303         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10304         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10305         (VEC_ANY_DATA, VEC_STRUCT): New constants.
10306         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10307         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10308         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10309         (aarch64_get_mask_mode): New functions.
10310         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10311         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10312         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
10313         predicate modes and predicate registers.  Explicitly restrict
10314         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
10315         to store a vector mode if it is recognized by
10316         aarch64_classify_vector_mode.
10317         (aarch64_regmode_natural_size): New function.
10318         (aarch64_hard_regno_caller_save_mode): Return the original mode
10319         for predicates.
10320         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10321         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10322         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10323         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10324         functions.
10325         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
10326         does not overlap dest if the function is frame-related.  Handle
10327         SVE constants.
10328         (aarch64_split_add_offset): New function.
10329         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10330         them aarch64_add_offset.
10331         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10332         and update call to aarch64_sub_sp.
10333         (aarch64_add_cfa_expression): New function.
10334         (aarch64_expand_prologue): Pass extra temporary registers to the
10335         functions above.  Handle the case in which we need to emit new
10336         DW_CFA_expressions for registers that were originally saved
10337         relative to the stack pointer, but now have to be expressed
10338         relative to the frame pointer.
10339         (aarch64_output_mi_thunk): Pass extra temporary registers to the
10340         functions above.
10341         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
10342         IP0 and IP1 values for SVE frames.
10343         (aarch64_expand_vec_series): New function.
10344         (aarch64_expand_sve_widened_duplicate): Likewise.
10345         (aarch64_expand_sve_const_vector): Likewise.
10346         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10347         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
10348         into the register, rather than emitting a SET directly.
10349         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10350         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10351         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10352         (offset_9bit_signed_scaled_p): New functions.
10353         (aarch64_replicate_bitmask_imm): New function.
10354         (aarch64_bitmask_imm): Use it.
10355         (aarch64_cannot_force_const_mem): Reject expressions involving
10356         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
10357         (aarch64_classify_index): Handle SVE indices, by requiring
10358         a plain register index with a scale that matches the element size.
10359         (aarch64_classify_address): Handle SVE addresses.  Assert that
10360         the mode of the address is VOIDmode or an integer mode.
10361         Update call to aarch64_classify_symbol.
10362         (aarch64_classify_symbolic_expression): Update call to
10363         aarch64_classify_symbol.
10364         (aarch64_const_vec_all_in_range_p): New function.
10365         (aarch64_print_vector_float_operand): Likewise.
10366         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
10367         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
10368         and the FP immediates 1.0 and 0.5.
10369         (aarch64_print_address_internal): Handle SVE addresses.
10370         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10371         (aarch64_regno_regclass): Handle predicate registers.
10372         (aarch64_secondary_reload): Handle big-endian reloads of SVE
10373         data modes.
10374         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10375         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10376         (aarch64_convert_sve_vector_bits): New function.
10377         (aarch64_override_options): Use it to handle -msve-vector-bits=.
10378         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10379         rather than an rtx.
10380         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10381         Handle SVE vector and predicate modes.  Accept VL-based constants
10382         that need only one temporary register, and VL offsets that require
10383         no temporary registers.
10384         (aarch64_conditional_register_usage): Mark the predicate registers
10385         as fixed if SVE isn't available.
10386         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10387         Return true for SVE vector and predicate modes.
10388         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10389         rather than an unsigned int.  Handle SVE modes.
10390         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
10391         SVE modes.
10392         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10393         if SVE is enabled.
10394         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10395         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10396         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10397         (aarch64_sve_float_mul_immediate_p): New functions.
10398         (aarch64_sve_valid_immediate): New function.
10399         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10400         Explicitly reject structure modes.  Check for INDEX constants.
10401         Handle PTRUE and PFALSE constants.
10402         (aarch64_check_zero_based_sve_index_immediate): New function.
10403         (aarch64_simd_imm_zero_p): Delete.
10404         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10405         vector modes.  Accept constants in the range of CNT[BHWD].
10406         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10407         ask for an Advanced SIMD mode.
10408         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10409         (aarch64_simd_vector_alignment): Handle SVE predicates.
10410         (aarch64_vectorize_preferred_vector_alignment): New function.
10411         (aarch64_simd_vector_alignment_reachable): Use it instead of
10412         the vector size.
10413         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10414         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10415         functions.
10416         (MAX_VECT_LEN): Delete.
10417         (expand_vec_perm_d): Add a vec_flags field.
10418         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10419         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10420         (aarch64_evpc_ext): Don't apply a big-endian lane correction
10421         for SVE modes.
10422         (aarch64_evpc_rev): Rename to...
10423         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
10424         (aarch64_evpc_rev_global): New function.
10425         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10426         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10427         MAX_VECT_LEN.
10428         (aarch64_evpc_sve_tbl): New function.
10429         (aarch64_expand_vec_perm_const_1): Update after rename of
10430         aarch64_evpc_rev.  Handle SVE permutes too, trying
10431         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10432         than aarch64_evpc_tbl.
10433         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10434         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10435         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10436         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10437         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10438         (aarch64_expand_sve_vcond): New functions.
10439         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10440         of aarch64_vector_mode_p.
10441         (aarch64_dwarf_poly_indeterminate_value): New function.
10442         (aarch64_compute_pressure_classes): Likewise.
10443         (aarch64_can_change_mode_class): Likewise.
10444         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10445         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10446         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10447         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10448         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10449         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10450         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10451         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10452         constraints.
10453         (Dn, Dl, Dr): Accept const as well as const_vector.
10454         (Dz): Likewise.  Compare against CONST0_RTX.
10455         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10456         of "vector" where appropriate.
10457         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10458         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10459         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10460         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10461         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10462         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10463         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10464         (v_int_equiv): Extend to SVE modes.
10465         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10466         mode attributes.
10467         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10468         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10469         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10470         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10471         (SVE_COND_FP_CMP): New int iterators.
10472         (perm_hilo): Handle the new unpack unspecs.
10473         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10474         attributes.
10475         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10476         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10477         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10478         (aarch64_equality_operator, aarch64_constant_vector_operand)
10479         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10480         (aarch64_sve_nonimmediate_operand): Likewise.
10481         (aarch64_sve_general_operand): Likewise.
10482         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10483         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10484         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10485         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10486         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10487         (aarch64_sve_float_arith_immediate): Likewise.
10488         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10489         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10490         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10491         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10492         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10493         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10494         (aarch64_sve_float_arith_operand): Likewise.
10495         (aarch64_sve_float_arith_with_sub_operand): Likewise.
10496         (aarch64_sve_float_mul_operand): Likewise.
10497         (aarch64_sve_vec_perm_operand): Likewise.
10498         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10499         (aarch64_mov_operand): Accept const_poly_int and const_vector.
10500         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10501         as well as const_vector.
10502         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10503         in file.  Use CONST0_RTX and CONSTM1_RTX.
10504         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
10505         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10506         Use aarch64_simd_imm_zero.
10507         * config/aarch64/aarch64-sve.md: New file.
10508         * config/aarch64/aarch64.md: Include it.
10509         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10510         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10511         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10512         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10513         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10514         (sve): New attribute.
10515         (enabled): Disable instructions with the sve attribute unless
10516         TARGET_SVE.
10517         (movqi, movhi): Pass CONST_POLY_INT operaneds through
10518         aarch64_expand_mov_immediate.
10519         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10520         CNT[BHSD] immediates.
10521         (movti): Split CONST_POLY_INT moves into two halves.
10522         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10523         Split additions that need a temporary here if the destination
10524         is the stack pointer.
10525         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10526         (*add<mode>3_poly_1): New instruction.
10527         (set_clobber_cc): New expander.
10528
10529 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10530
10531         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10532         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
10533         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10534         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10535         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
10536         Change innermode from fixed_mode_size to machine_mode.
10537         (simplify_subreg): Update call accordingly.  Handle a constant-sized
10538         subreg of a variable-length CONST_VECTOR.
10539
10540 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10541             Alan Hayward  <alan.hayward@arm.com>
10542             David Sherwood  <david.sherwood@arm.com>
10543
10544         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10545         (add_offset_to_base): New function, split out from...
10546         (create_mem_ref): ...here.  When handling a scale other than 1,
10547         check first whether the address is valid without the offset.
10548         Add it into the base if so, leaving the index and scale as-is.
10549
10550 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
10551
10552         PR c++/83778
10553         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10554         fold_for_warn before checking if arg2 is INTEGER_CST.
10555
10556 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
10557
10558         * config/rs6000/predicates.md (load_multiple_operation): Delete.
10559         (store_multiple_operation): Delete.
10560         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10561         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10562         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10563         guarded by TARGET_STRING.
10564         (rs6000_output_load_multiple): Delete.
10565         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10566         OPTION_MASK_STRING / TARGET_STRING handling.
10567         (print_operand) <'N', 'O'>: Add comment that these are unused now.
10568         (const rs6000_opt_masks) <"string">: Change mask to 0.
10569         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10570         (MASK_STRING): Delete.
10571         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10572         parts.  Simplify.
10573         (load_multiple): Delete.
10574         (*ldmsi8): Delete.
10575         (*ldmsi7): Delete.
10576         (*ldmsi6): Delete.
10577         (*ldmsi5): Delete.
10578         (*ldmsi4): Delete.
10579         (*ldmsi3): Delete.
10580         (store_multiple): Delete.
10581         (*stmsi8): Delete.
10582         (*stmsi7): Delete.
10583         (*stmsi6): Delete.
10584         (*stmsi5): Delete.
10585         (*stmsi4): Delete.
10586         (*stmsi3): Delete.
10587         (movmemsi_8reg): Delete.
10588         (corresponding unnamed define_insn): Delete.
10589         (movmemsi_6reg): Delete.
10590         (corresponding unnamed define_insn): Delete.
10591         (movmemsi_4reg): Delete.
10592         (corresponding unnamed define_insn): Delete.
10593         (movmemsi_2reg): Delete.
10594         (corresponding unnamed define_insn): Delete.
10595         (movmemsi_1reg): Delete.
10596         (corresponding unnamed define_insn): Delete.
10597         * config/rs6000/rs6000.opt (mno-string): New.
10598         (mstring): Replace by deprecation warning stub.
10599         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10600
10601 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
10602
10603         * regrename.c (regrename_do_replace): If replacing the same
10604         reg multiple times, try to reuse last created gen_raw_REG.
10605
10606         PR debug/81155
10607         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10608         main to workaround a bug in GDB.
10609
10610 2018-01-12  Tom de Vries  <tom@codesourcery.com>
10611
10612         PR target/83737
10613         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10614
10615 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
10616
10617         PR rtl-optimization/80481
10618         * ira-color.c (get_cap_member): New function.
10619         (allocnos_conflict_by_live_ranges_p): Use it.
10620         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10621         (setup_slot_coalesced_allocno_live_ranges): Ditto.
10622
10623 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
10624
10625         PR target/83628
10626         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10627         (*saddl_se_1): Ditto.
10628         (*ssubsi_1): Ditto.
10629         (*ssubl_se_1): Ditto.
10630
10631 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
10632
10633         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10634         rather than wi::to_widest for DR_INITs.
10635         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10636         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10637         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10638         INTEGER_CSTs.
10639         (vect_analyze_group_access_1): Note that here.
10640
10641 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
10642
10643         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10644         polynomial type sizes.
10645
10646 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
10647
10648         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10649         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10650         (gimple_add_tmp_var): Likewise.
10651
10652 2018-01-12  Martin Liska  <mliska@suse.cz>
10653
10654         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10655         (gimple_alloc_sizes): Likewise.
10656         (dump_gimple_statistics): Use PRIu64 in printf format.
10657         * gimple.h: Change uint64_t to int.
10658
10659 2018-01-12  Martin Liska  <mliska@suse.cz>
10660
10661         * tree-core.h: Use uint64_t instead of int.
10662         * tree.c (tree_node_counts): Likewise.
10663         (tree_node_sizes): Likewise.
10664         (dump_tree_statistics): Use PRIu64 in printf format.
10665
10666 2018-01-12  Martin Liska  <mliska@suse.cz>
10667
10668         * Makefile.in: As qsort_chk is implemented in vec.c, add
10669         vec.o to linkage of gencfn-macros.
10670         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10671         passing the info to record_node_allocation_statistics.
10672         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10673         and pass the info.
10674         * ggc-common.c (struct ggc_usage): Add operator== and use
10675         it in operator< and compare function.
10676         * mem-stats.h (struct mem_usage): Likewise.
10677         * vec.c (struct vec_usage): Remove operator< and compare
10678         function. Can be simply inherited.
10679
10680 2018-01-12  Martin Jambor  <mjambor@suse.cz>
10681
10682         PR target/81616
10683         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10684         * tree-ssa-math-opts.c: Include domwalk.h.
10685         (convert_mult_to_fma_1): New function.
10686         (fma_transformation_info): New type.
10687         (fma_deferring_state): Likewise.
10688         (cancel_fma_deferring): New function.
10689         (result_of_phi): Likewise.
10690         (last_fma_candidate_feeds_initial_phi): Likewise.
10691         (convert_mult_to_fma): Added deferring logic, split actual
10692         transformation to convert_mult_to_fma_1.
10693         (math_opts_dom_walker): New type.
10694         (math_opts_dom_walker::after_dom_children): New method, body moved
10695         here from pass_optimize_widening_mul::execute, added deferring logic
10696         bits.
10697         (pass_optimize_widening_mul::execute): Moved most of code to
10698         math_opts_dom_walker::after_dom_children.
10699         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10700         * config/i386/i386.c (ix86_option_override_internal): Added
10701         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10702
10703 2018-01-12  Richard Biener  <rguenther@suse.de>
10704
10705         PR debug/83157
10706         * dwarf2out.c (gen_variable_die): Do not reset old_die for
10707         inline instance vars.
10708
10709 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
10710
10711         PR target/81819
10712         * config/rx/rx.c (rx_is_restricted_memory_address):
10713         Handle SUBREG case.
10714
10715 2018-01-12  Richard Biener  <rguenther@suse.de>
10716
10717         PR tree-optimization/80846
10718         * target.def (split_reduction): New target hook.
10719         * targhooks.c (default_split_reduction): New function.
10720         * targhooks.h (default_split_reduction): Declare.
10721         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10722         target requests first reduce vectors by combining low and high
10723         parts.
10724         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10725         (get_vectype_for_scalar_type_and_size): Export.
10726         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10727         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10728         * doc/tm.texi: Regenerate.
10729         * config/i386/i386.c (ix86_split_reduction): Implement
10730         TARGET_VECTORIZE_SPLIT_REDUCTION.
10731
10732 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
10733
10734         PR target/83368
10735         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10736         in PIC mode except for TARGET_VXWORKS_RTP.
10737         * config/sparc/sparc.c: Include cfgrtl.h.
10738         (TARGET_INIT_PIC_REG): Define.
10739         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10740         (sparc_pic_register_p): New predicate.
10741         (sparc_legitimate_address_p): Use it.
10742         (sparc_legitimize_pic_address): Likewise.
10743         (sparc_delegitimize_address): Likewise.
10744         (sparc_mode_dependent_address_p): Likewise.
10745         (gen_load_pcrel_sym): Remove 4th parameter.
10746         (load_got_register): Adjust call to above.  Remove obsolete stuff.
10747         (sparc_expand_prologue): Do not call load_got_register here.
10748         (sparc_flat_expand_prologue): Likewise.
10749         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10750         (sparc_use_pseudo_pic_reg): New function.
10751         (sparc_init_pic_reg): Likewise.
10752         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10753         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10754
10755 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
10756
10757         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10758         Add item for branch_cost.
10759
10760 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
10761
10762         PR rtl-optimization/83565
10763         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10764         not extend the result to a larger mode for rotate operations.
10765         (num_sign_bit_copies1): Likewise.
10766
10767 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10768
10769         PR target/40411
10770         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10771         -symbolic.
10772         Use values-Xc.o for -pedantic.
10773         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10774
10775 2018-01-12  Martin Liska  <mliska@suse.cz>
10776
10777         PR ipa/83054
10778         * ipa-devirt.c (final_warning_record::grow_type_warnings):
10779         New function.
10780         (possible_polymorphic_call_targets): Use it.
10781         (ipa_devirt): Likewise.
10782
10783 2018-01-12  Martin Liska  <mliska@suse.cz>
10784
10785         * profile-count.h (enum profile_quality): Use 0 as invalid
10786         enum value of profile_quality.
10787
10788 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
10789
10790         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10791         -mext-string options.
10792
10793 2018-01-12  Richard Biener  <rguenther@suse.de>
10794
10795         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10796         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10797         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10798         Likewise.
10799         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10800
10801 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
10802
10803         * configure.ac (--with-long-double-format): Add support for the
10804         configuration option to change the default long double format on
10805         PowerPC systems.
10806         * config.gcc (powerpc*-linux*-*): Likewise.
10807         * configure: Regenerate.
10808         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10809         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10810         used without modification.
10811
10812 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10813
10814         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10815         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10816         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10817         MISC_BUILTIN_SPEC_BARRIER.
10818         (rs6000_init_builtins): Likewise.
10819         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10820         enum value.
10821         (speculation_barrier): New define_insn.
10822         * doc/extend.texi: Document __builtin_speculation_barrier.
10823
10824 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
10825
10826         PR target/83203
10827         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10828         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10829         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10830         iterators.
10831         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
10832         integral modes instead of "ss" and "sd".
10833         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10834         vectors with 32-bit and 64-bit elements.
10835         (vecdupssescalarmodesuffix): New mode attribute.
10836         (vec_dup<mode>): Use it.
10837
10838 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
10839
10840         PR target/83330
10841         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10842         frame if argument is passed on stack.
10843
10844 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
10845
10846         PR target/82682
10847         * ree.c (combine_reaching_defs): Optimize also
10848         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10849         reg2=any_extend(exp); reg1=reg2;, formatting fix.
10850
10851 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
10852
10853         PR middle-end/83189
10854         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10855
10856 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
10857
10858         PR middle-end/83718
10859         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10860         after they are computed.
10861
10862 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
10863
10864         PR tree-optimization/83695
10865         * gimple-loop-linterchange.cc
10866         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10867         reset cached scev information after interchange.
10868         (pass_linterchange::execute): Remove call to scev_reset_htab.
10869
10870 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10871
10872         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10873         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10874         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10875         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10876         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10877         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10878         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10879         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10880         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10881         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10882         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10883         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10884         (V_lane_reg): Likewise.
10885         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10886         New define_expand.
10887         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10888         (vfmal_lane_low<mode>_intrinsic,
10889         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10890         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10891         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10892         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10893         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10894         vfmsl_lane_high<mode>_intrinsic): New define_insns.
10895
10896 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10897
10898         * config/arm/arm-cpus.in (fp16fml): New feature.
10899         (ALL_SIMD): Add fp16fml.
10900         (armv8.2-a): Add fp16fml as an option.
10901         (armv8.3-a): Likewise.
10902         (armv8.4-a): Add fp16fml as part of fp16.
10903         * config/arm/arm.h (TARGET_FP16FML): Define.
10904         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10905         when appropriate.
10906         * config/arm/arm-modes.def (V2HF): Define.
10907         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10908         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10909         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10910         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10911         vfmsl_low, vfmsl_high): New set of builtins.
10912         * config/arm/iterators.md (PLUSMINUS): New code iterator.
10913         (vfml_op): New code attribute.
10914         (VFMLHALVES): New int iterator.
10915         (VFML, VFMLSEL): New mode attributes.
10916         (V_reg): Define mapping for V2HF.
10917         (V_hi, V_lo): New mode attributes.
10918         (VF_constraint): Likewise.
10919         (vfml_half, vfml_half_selector): New int attributes.
10920         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10921         define_expand.
10922         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10923         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10924         New define_insn.
10925         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10926         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10927         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10928         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
10929         documentation.
10930         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10931         Document new effective target and option set.
10932
10933 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10934
10935         * config/arm/arm-cpus.in (armv8_4): New feature.
10936         (ARMv8_4a): New fgroup.
10937         (armv8.4-a): New arch.
10938         * config/arm/arm-tables.opt: Regenerate.
10939         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10940         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10941         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10942         Add matching rules for -march=armv8.4-a and extensions.
10943         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10944
10945 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
10946
10947         PR target/81821
10948         * config/rx/rx.md (BW): New mode attribute.
10949         (sync_lock_test_and_setsi): Add mode suffix to insn output.
10950
10951 2018-01-11  Richard Biener  <rguenther@suse.de>
10952
10953         PR tree-optimization/83435
10954         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10955         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10956         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10957
10958 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
10959             Alan Hayward  <alan.hayward@arm.com>
10960             David Sherwood  <david.sherwood@arm.com>
10961
10962         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10963         field.
10964         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
10965         (aarch64_print_address_internal): Use it to check for a zero offset.
10966
10967 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
10968             Alan Hayward  <alan.hayward@arm.com>
10969             David Sherwood  <david.sherwood@arm.com>
10970
10971         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10972         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10973         Return a poly_int64 rather than a HOST_WIDE_INT.
10974         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10975         rather than a HOST_WIDE_INT.
10976         * config/aarch64/aarch64.h (aarch64_frame): Protect with
10977         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
10978         hard_fp_offset, frame_size, initial_adjust, callee_offset and
10979         final_offset from HOST_WIDE_INT to poly_int64.
10980         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10981         to_constant when getting the number of units in an Advanced SIMD
10982         mode.
10983         (aarch64_builtin_vectorized_function): Check for a constant number
10984         of units.
10985         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10986         GET_MODE_SIZE.
10987         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10988         attribute instead of GET_MODE_NUNITS.
10989         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10990         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10991         GET_MODE_SIZE for fixed-size registers.
10992         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10993         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10994         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10995         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10996         (aarch64_print_operand, aarch64_print_address_internal)
10997         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10998         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10999         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11000         Handle polynomial GET_MODE_SIZE.
11001         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
11002         wider than SImode without modification.
11003         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11004         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11005         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11006         passing and returning SVE modes.
11007         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11008         rather than GEN_INT.
11009         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11010         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11011         (aarch64_allocate_and_probe_stack_space): Likewise.
11012         (aarch64_layout_frame): Cope with polynomial offsets.
11013         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11014         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
11015         polynomial offsets.
11016         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11017         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11018         poly_int64 rather than a HOST_WIDE_INT.
11019         (aarch64_get_separate_components, aarch64_process_components)
11020         (aarch64_expand_prologue, aarch64_expand_epilogue)
11021         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11022         (aarch64_anchor_offset): New function, split out from...
11023         (aarch64_legitimize_address): ...here.
11024         (aarch64_builtin_vectorization_cost): Handle polynomial
11025         TYPE_VECTOR_SUBPARTS.
11026         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11027         GET_MODE_NUNITS.
11028         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11029         number of elements from the PARALLEL rather than the mode.
11030         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11031         rather than GET_MODE_BITSIZE.
11032         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11033         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11034         (aarch64_expand_vec_perm_const_1): Handle polynomial
11035         d->perm.length () and d->perm elements.
11036         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
11037         Apply to_constant to d->perm elements.
11038         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11039         polynomial CONST_VECTOR_NUNITS.
11040         (aarch64_move_pointer): Take amount as a poly_int64 rather
11041         than an int.
11042         (aarch64_progress_pointer): Avoid temporary variable.
11043         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11044         the mode attribute instead of GET_MODE.
11045
11046 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
11047             Alan Hayward  <alan.hayward@arm.com>
11048             David Sherwood  <david.sherwood@arm.com>
11049
11050         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11051         x exists before using it.
11052         (aarch64_add_constant_internal): Rename to...
11053         (aarch64_add_offset_1): ...this.  Replace regnum with separate
11054         src and dest rtxes.  Handle the case in which they're different,
11055         including when the offset is zero.  Replace scratchreg with an rtx.
11056         Use 2 additions if there is no spare register into which we can
11057         move a 16-bit constant.
11058         (aarch64_add_constant): Delete.
11059         (aarch64_add_offset): Replace reg with separate src and dest
11060         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
11061         Use aarch64_add_offset_1.
11062         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11063         an rtx rather than an int.  Take the delta as a poly_int64
11064         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
11065         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11066         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11067         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11068         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11069         and aarch64_add_sp.
11070         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11071         aarch64_add_constant.
11072
11073 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
11074
11075         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11076         Use scalar_float_mode.
11077
11078 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
11079
11080         * config/aarch64/aarch64-simd.md
11081         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11082         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11083         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11084         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11085         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11086         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11087         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11088         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11089         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11090         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11091
11092 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
11093
11094         PR target/83514
11095         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11096         targ_options->x_arm_arch_string is non NULL.
11097
11098 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
11099
11100         * config/aarch64/aarch64.h
11101         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
11102
11103 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
11104
11105         PR target/82096
11106         * expmed.c (emit_store_flag_force): Swap if const op0
11107         and change VOIDmode to mode of op0.
11108
11109 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
11110
11111         PR rtl-optimization/83761
11112         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11113         than bytes to mode_for_size.
11114
11115 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
11116
11117         PR middle-end/83189
11118         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11119         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11120         profile.
11121
11122 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
11123
11124         PR middle-end/83575
11125         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11126         when in layout mode.
11127         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11128         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11129         partition fixup.
11130
11131 2018-01-10  Michael Collison  <michael.collison@arm.com>
11132
11133         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11134         * config/aarch64/aarch64-option-extension.def: Add
11135         AARCH64_OPT_EXTENSION of 'fp16fml'.
11136         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11137         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11138         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11139         * config/aarch64/constraints.md (Ui7): New constraint.
11140         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11141         (VFMLA_SEL_W): Ditto.
11142         (f16quad): Ditto.
11143         (f16mac1): Ditto.
11144         (VFMLA16_LOW): New int iterator.
11145         (VFMLA16_HIGH): Ditto.
11146         (UNSPEC_FMLAL): New unspec.
11147         (UNSPEC_FMLSL): Ditto.
11148         (UNSPEC_FMLAL2): Ditto.
11149         (UNSPEC_FMLSL2): Ditto.
11150         (f16mac): New code attribute.
11151         * config/aarch64/aarch64-simd-builtins.def
11152         (aarch64_fmlal_lowv2sf): Ditto.
11153         (aarch64_fmlsl_lowv2sf): Ditto.
11154         (aarch64_fmlalq_lowv4sf): Ditto.
11155         (aarch64_fmlslq_lowv4sf): Ditto.
11156         (aarch64_fmlal_highv2sf): Ditto.
11157         (aarch64_fmlsl_highv2sf): Ditto.
11158         (aarch64_fmlalq_highv4sf): Ditto.
11159         (aarch64_fmlslq_highv4sf): Ditto.
11160         (aarch64_fmlal_lane_lowv2sf): Ditto.
11161         (aarch64_fmlsl_lane_lowv2sf): Ditto.
11162         (aarch64_fmlal_laneq_lowv2sf): Ditto.
11163         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11164         (aarch64_fmlalq_lane_lowv4sf): Ditto.
11165         (aarch64_fmlsl_lane_lowv4sf): Ditto.
11166         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11167         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11168         (aarch64_fmlal_lane_highv2sf): Ditto.
11169         (aarch64_fmlsl_lane_highv2sf): Ditto.
11170         (aarch64_fmlal_laneq_highv2sf): Ditto.
11171         (aarch64_fmlsl_laneq_highv2sf): Ditto.
11172         (aarch64_fmlalq_lane_highv4sf): Ditto.
11173         (aarch64_fmlsl_lane_highv4sf): Ditto.
11174         (aarch64_fmlalq_laneq_highv4sf): Ditto.
11175         (aarch64_fmlsl_laneq_highv4sf): Ditto.
11176         * config/aarch64/aarch64-simd.md:
11177         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11178         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11179         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11180         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11181         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11182         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11183         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11184         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11185         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11186         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11187         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11188         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11189         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11190         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11191         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11192         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11193         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11194         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11195         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11196         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11197         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11198         (vfmlsl_low_u32): Ditto.
11199         (vfmlalq_low_u32): Ditto.
11200         (vfmlslq_low_u32): Ditto.
11201         (vfmlal_high_u32): Ditto.
11202         (vfmlsl_high_u32): Ditto.
11203         (vfmlalq_high_u32): Ditto.
11204         (vfmlslq_high_u32): Ditto.
11205         (vfmlal_lane_low_u32): Ditto.
11206         (vfmlsl_lane_low_u32): Ditto.
11207         (vfmlal_laneq_low_u32): Ditto.
11208         (vfmlsl_laneq_low_u32): Ditto.
11209         (vfmlalq_lane_low_u32): Ditto.
11210         (vfmlslq_lane_low_u32): Ditto.
11211         (vfmlalq_laneq_low_u32): Ditto.
11212         (vfmlslq_laneq_low_u32): Ditto.
11213         (vfmlal_lane_high_u32): Ditto.
11214         (vfmlsl_lane_high_u32): Ditto.
11215         (vfmlal_laneq_high_u32): Ditto.
11216         (vfmlsl_laneq_high_u32): Ditto.
11217         (vfmlalq_lane_high_u32): Ditto.
11218         (vfmlslq_lane_high_u32): Ditto.
11219         (vfmlalq_laneq_high_u32): Ditto.
11220         (vfmlslq_laneq_high_u32): Ditto.
11221         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11222         (AARCH64_FL_FOR_ARCH8_4): New.
11223         (AARCH64_ISA_F16FML): New ISA flag.
11224         (TARGET_F16FML): New feature flag for fp16fml.
11225         (doc/invoke.texi): Document new fp16fml option.
11226
11227 2018-01-10  Michael Collison  <michael.collison@arm.com>
11228
11229         * config/aarch64/aarch64-builtins.c:
11230         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11231         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11232         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11233         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11234         (AARCH64_ISA_SHA3): New ISA flag.
11235         (TARGET_SHA3): New feature flag for sha3.
11236         * config/aarch64/iterators.md (sha512_op): New int attribute.
11237         (CRYPTO_SHA512): New int iterator.
11238         (UNSPEC_SHA512H): New unspec.
11239         (UNSPEC_SHA512H2): Ditto.
11240         (UNSPEC_SHA512SU0): Ditto.
11241         (UNSPEC_SHA512SU1): Ditto.
11242         * config/aarch64/aarch64-simd-builtins.def
11243         (aarch64_crypto_sha512hqv2di): New builtin.
11244         (aarch64_crypto_sha512h2qv2di): Ditto.
11245         (aarch64_crypto_sha512su0qv2di): Ditto.
11246         (aarch64_crypto_sha512su1qv2di): Ditto.
11247         (aarch64_eor3qv8hi): Ditto.
11248         (aarch64_rax1qv2di): Ditto.
11249         (aarch64_xarqv2di): Ditto.
11250         (aarch64_bcaxqv8hi): Ditto.
11251         * config/aarch64/aarch64-simd.md:
11252         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11253         (aarch64_crypto_sha512su0qv2di): Ditto.
11254         (aarch64_crypto_sha512su1qv2di): Ditto.
11255         (aarch64_eor3qv8hi): Ditto.
11256         (aarch64_rax1qv2di): Ditto.
11257         (aarch64_xarqv2di): Ditto.
11258         (aarch64_bcaxqv8hi): Ditto.
11259         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11260         (vsha512h2q_u64): Ditto.
11261         (vsha512su0q_u64): Ditto.
11262         (vsha512su1q_u64): Ditto.
11263         (veor3q_u16): Ditto.
11264         (vrax1q_u64): Ditto.
11265         (vxarq_u64): Ditto.
11266         (vbcaxq_u16): Ditto.
11267         * config/arm/types.md (crypto_sha512): New type attribute.
11268         (crypto_sha3): Ditto.
11269         (doc/invoke.texi): Document new sha3 option.
11270
11271 2018-01-10  Michael Collison  <michael.collison@arm.com>
11272
11273         * config/aarch64/aarch64-builtins.c:
11274         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11275         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11276         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11277         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11278         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11279         (AARCH64_ISA_SM4): New ISA flag.
11280         (TARGET_SM4): New feature flag for sm4.
11281         * config/aarch64/aarch64-simd-builtins.def
11282         (aarch64_sm3ss1qv4si): Ditto.
11283         (aarch64_sm3tt1aq4si): Ditto.
11284         (aarch64_sm3tt1bq4si): Ditto.
11285         (aarch64_sm3tt2aq4si): Ditto.
11286         (aarch64_sm3tt2bq4si): Ditto.
11287         (aarch64_sm3partw1qv4si): Ditto.
11288         (aarch64_sm3partw2qv4si): Ditto.
11289         (aarch64_sm4eqv4si): Ditto.
11290         (aarch64_sm4ekeyqv4si): Ditto.
11291         * config/aarch64/aarch64-simd.md:
11292         (aarch64_sm3ss1qv4si): Ditto.
11293         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11294         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11295         (aarch64_sm4eqv4si): Ditto.
11296         (aarch64_sm4ekeyqv4si): Ditto.
11297         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11298         (sm3part_op): Ditto.
11299         (CRYPTO_SM3TT): Ditto.
11300         (CRYPTO_SM3PART): Ditto.
11301         (UNSPEC_SM3SS1): New unspec.
11302         (UNSPEC_SM3TT1A): Ditto.
11303         (UNSPEC_SM3TT1B): Ditto.
11304         (UNSPEC_SM3TT2A): Ditto.
11305         (UNSPEC_SM3TT2B): Ditto.
11306         (UNSPEC_SM3PARTW1): Ditto.
11307         (UNSPEC_SM3PARTW2): Ditto.
11308         (UNSPEC_SM4E): Ditto.
11309         (UNSPEC_SM4EKEY): Ditto.
11310         * config/aarch64/constraints.md (Ui2): New constraint.
11311         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11312         * config/arm/types.md (crypto_sm3): New type attribute.
11313         (crypto_sm4): Ditto.
11314         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11315         (vsm3tt1aq_u32): Ditto.
11316         (vsm3tt1bq_u32): Ditto.
11317         (vsm3tt2aq_u32): Ditto.
11318         (vsm3tt2bq_u32): Ditto.
11319         (vsm3partw1q_u32): Ditto.
11320         (vsm3partw2q_u32): Ditto.
11321         (vsm4eq_u32): Ditto.
11322         (vsm4ekeyq_u32): Ditto.
11323         (doc/invoke.texi): Document new sm4 option.
11324
11325 2018-01-10  Michael Collison  <michael.collison@arm.com>
11326
11327         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11328         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11329         (AARCH64_FL_FOR_ARCH8_4): New.
11330         (AARCH64_FL_V8_4): New flag.
11331         (doc/invoke.texi): Document new armv8.4-a option.
11332
11333 2018-01-10  Michael Collison  <michael.collison@arm.com>
11334
11335         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11336         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11337         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11338         * config/aarch64/aarch64-option-extension.def: Add
11339         AARCH64_OPT_EXTENSION of 'sha2'.
11340         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11341         (crypto): Disable sha2 and aes if crypto disabled.
11342         (crypto): Enable aes and sha2 if enabled.
11343         (simd): Disable sha2 and aes if simd disabled.
11344         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11345         New flags.
11346         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11347         (TARGET_SHA2): New feature flag for sha2.
11348         (TARGET_AES): New feature flag for aes.
11349         * config/aarch64/aarch64-simd.md:
11350         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11351         conditional on TARGET_AES.
11352         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11353         (aarch64_crypto_sha1hsi): Make pattern conditional
11354         on TARGET_SHA2.
11355         (aarch64_crypto_sha1hv4si): Ditto.
11356         (aarch64_be_crypto_sha1hv4si): Ditto.
11357         (aarch64_crypto_sha1su1v4si): Ditto.
11358         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11359         (aarch64_crypto_sha1su0v4si): Ditto.
11360         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11361         (aarch64_crypto_sha256su0v4si): Ditto.
11362         (aarch64_crypto_sha256su1v4si): Ditto.
11363         (doc/invoke.texi): Document new aes and sha2 options.
11364
11365 2018-01-10  Martin Sebor  <msebor@redhat.com>
11366
11367         PR tree-optimization/83781
11368         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11369         as string arrays.
11370
11371 2018-01-11  Martin Sebor  <msebor@gmail.com>
11372             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
11373
11374         PR tree-optimization/83501
11375         PR tree-optimization/81703
11376
11377         * tree-ssa-strlen.c (get_string_cst): Rename...
11378         (get_string_len): ...to this.  Handle global constants.
11379         (handle_char_store): Adjust.
11380
11381 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
11382             Jim Wilson  <jimw@sifive.com>
11383
11384         * config/riscv/riscv-protos.h (riscv_output_return): New.
11385         * config/riscv/riscv.c (struct machine_function): New naked_p field.
11386         (riscv_attribute_table, riscv_output_return),
11387         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11388         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11389         (riscv_compute_frame_info): Only compute frame->mask if not a naked
11390         function.
11391         (riscv_expand_prologue): Add early return for naked function.
11392         (riscv_expand_epilogue): Likewise.
11393         (riscv_function_ok_for_sibcall): Return false for naked function.
11394         (riscv_set_current_function): New.
11395         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11396         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11397         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11398         * doc/extend.texi (RISC-V Function Attributes): New.
11399
11400 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
11401
11402         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11403         check for 128-bit long double before checking TCmode.
11404         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11405         128-bit long doubles before checking TFmode or TCmode.
11406         (FLOAT128_IBM_P): Likewise.
11407
11408 2018-01-10  Martin Sebor  <msebor@redhat.com>
11409
11410         PR tree-optimization/83671
11411         * builtins.c (c_strlen): Unconditionally return zero for the empty
11412         string.
11413         Use -Warray-bounds for warnings.
11414         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11415         for non-constant array indices with COMPONENT_REF, arrays of
11416         arrays, and pointers to arrays.
11417         (gimple_fold_builtin_strlen): Determine and set length range for
11418         non-constant character arrays.
11419
11420 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
11421
11422         PR middle-end/81897
11423         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11424         empty blocks.
11425
11426 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
11427
11428         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11429
11430 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
11431
11432         PR target/83399
11433         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11434         VECTOR_MEM_ALTIVEC_OR_VSX_P.
11435         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11436         indexed_or_indirect_operand predicate.
11437         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11438         (*vsx_le_perm_load_v8hi): Likewise.
11439         (*vsx_le_perm_load_v16qi): Likewise.
11440         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11441         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11442         (*vsx_le_perm_store_v8hi): Likewise.
11443         (*vsx_le_perm_store_v16qi): Likewise.
11444         (eight unnamed splitters): Likewise.
11445
11446 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
11447
11448         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11449         * config/rs6000/emmintrin.h: Likewise.
11450         * config/rs6000/mmintrin.h: Likewise.
11451         * config/rs6000/xmmintrin.h: Likewise.
11452
11453 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
11454
11455         PR c++/43486
11456         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11457         "public_flag".
11458         * tree.c (tree_nop_conversion): Return true for location wrapper
11459         nodes.
11460         (maybe_wrap_with_location): New function.
11461         (selftest::check_strip_nops): New function.
11462         (selftest::test_location_wrappers): New function.
11463         (selftest::tree_c_tests): Call it.
11464         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11465         (maybe_wrap_with_location): New decl.
11466         (EXPR_LOCATION_WRAPPER_P): New macro.
11467         (location_wrapper_p): New inline function.
11468         (tree_strip_any_location_wrapper): New inline function.
11469
11470 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
11471
11472         PR target/83735
11473         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11474         stack_realign_offset for the largest alignment of stack slot
11475         actually used.
11476         (ix86_find_max_used_stack_alignment): New function.
11477         (ix86_finalize_stack_frame_flags): Use it.  Set
11478         max_used_stack_alignment if we don't realign stack.
11479         * config/i386/i386.h (machine_function): Add
11480         max_used_stack_alignment.
11481
11482 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
11483
11484         * config/arm/arm.opt (-mbranch-cost): New option.
11485         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11486         account.
11487
11488 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
11489
11490         PR target/83629
11491         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11492         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11493
11494 2018-01-10  Richard Biener  <rguenther@suse.de>
11495
11496         PR debug/83765
11497         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11498         early out so it also covers the case where we have a non-NULL
11499         origin.
11500
11501 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
11502
11503         PR tree-optimization/83753
11504         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11505         for non-strided grouped accesses if the number of elements is 1.
11506
11507 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
11508
11509         PR target/81616
11510         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11511         * i386.h (TARGET_USE_GATHER): Define.
11512         * x86-tune.def (X86_TUNE_USE_GATHER): New.
11513
11514 2018-01-10  Martin Liska  <mliska@suse.cz>
11515
11516         PR bootstrap/82831
11517         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11518         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11519         partitioning.
11520         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11521         CLEANUP_NO_PARTITIONING is not set.
11522
11523 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
11524
11525         * doc/rtl.texi: Remove documentation of (const ...) wrappers
11526         for vectors, as a partial revert of r254296.
11527         * rtl.h (const_vec_p): Delete.
11528         (const_vec_duplicate_p): Don't test for vector CONSTs.
11529         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11530         * expmed.c (make_tree): Likewise.
11531
11532         Revert:
11533         * common.md (E, F): Use CONSTANT_P instead of checking for
11534         CONST_VECTOR.
11535         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11536         checking for CONST_VECTOR.
11537
11538 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
11539
11540         PR middle-end/83575
11541         * predict.c (force_edge_cold): Handle in more sane way edges
11542         with no prediction.
11543
11544 2018-01-09  Carl Love  <cel@us.ibm.com>
11545
11546         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11547         V4SI, V4SF types.
11548         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11549         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11550         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11551         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
11552         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11553         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
11554         * config/rs6000/rs6000-protos.h: Add extern defition for
11555         rs6000_generate_float2_double_code.
11556         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11557         function.
11558         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11559         (float2_v2df): Add define_expand.
11560
11561 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
11562
11563         PR target/83628
11564         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11565         op_mode in the force_to_mode call.
11566
11567 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
11568
11569         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11570         instead of checking each element individually.
11571         (aarch64_evpc_uzp): Likewise.
11572         (aarch64_evpc_zip): Likewise.
11573         (aarch64_evpc_ext): Likewise.
11574         (aarch64_evpc_rev): Likewise.
11575         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11576         instead of checking each element individually.  Return true without
11577         generating rtl if
11578         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11579         whether all selected elements come from the same input, instead of
11580         checking each element individually.  Remove calls to gen_rtx_REG,
11581         start_sequence and end_sequence and instead assert that no rtl is
11582         generated.
11583
11584 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
11585
11586         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11587         order of HIGH and CONST checks.
11588
11589 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
11590
11591         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11592         if the destination isn't an SSA_NAME.
11593
11594 2018-01-09  Richard Biener  <rguenther@suse.de>
11595
11596         PR tree-optimization/83668
11597         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11598         move prologue...
11599         (canonicalize_loop_form): ... here, renamed from ...
11600         (canonicalize_loop_closed_ssa_form): ... this and amended to
11601         swap successor edges for loop exit blocks to make us use
11602         the RPO order we need for initial schedule generation.
11603
11604 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
11605
11606         PR tree-optimization/64811
11607         * match.pd: When optimizing comparisons with Inf, avoid
11608         introducing or losing exceptions from comparisons with NaN.
11609
11610 2018-01-09  Martin Liska  <mliska@suse.cz>
11611
11612         PR sanitizer/82517
11613         * asan.c (shadow_mem_size): Add gcc_assert.
11614
11615 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
11616
11617         Don't save registers in main().
11618
11619         PR target/83738
11620         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11621         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11622         * config/avr/avr.c (avr_set_current_function): Don't error if
11623         naked, OS_task or OS_main are specified at the same time.
11624         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11625         OS_main.
11626         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11627         attribute.
11628         * common/config/avr/avr-common.c (avr_option_optimization_table):
11629         Switch on -mmain-is-OS_task for optimizing compilations.
11630
11631 2018-01-09  Richard Biener  <rguenther@suse.de>
11632
11633         PR tree-optimization/83572
11634         * graphite.c: Include cfganal.h.
11635         (graphite_transform_loops): Connect infinite loops to exit
11636         and remove fake edges at the end.
11637
11638 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
11639
11640         * ipa-inline.c (edge_badness): Revert accidental checkin.
11641
11642 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
11643
11644         PR ipa/80763
11645         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11646         symbols; not inline clones.
11647
11648 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
11649
11650         PR target/83507
11651         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11652         hard registers.  Formatting fixes.
11653
11654         PR preprocessor/83722
11655         * gcc.c (try_generate_repro): Pass
11656         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11657         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11658         do_report_bug.
11659
11660 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
11661             Kito Cheng  <kito.cheng@gmail.com>
11662
11663         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11664         (riscv_leaf_function_p): Delete.
11665         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11666
11667 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
11668
11669         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11670         function.
11671         (do_ifelse): New function.
11672         (do_isel): New function.
11673         (do_sub3): New function.
11674         (do_add3): New function.
11675         (do_load_mask_compare): New function.
11676         (do_overlap_load_compare): New function.
11677         (expand_compare_loop): New function.
11678         (expand_block_compare): Call expand_compare_loop() when appropriate.
11679         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11680         option description.
11681         (-mblock-compare-inline-loop-limit): New option.
11682
11683 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
11684
11685         PR target/83677
11686         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11687         Reverse order of second and third operands in first alternative.
11688         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11689         of first and second elements in UNSPEC_VPERMR vector.
11690         (altivec_expand_vec_perm_le): Likewise.
11691
11692 2018-01-08  Jeff Law  <law@redhat.com>
11693
11694         PR rtl-optimizatin/81308
11695         * tree-switch-conversion.c (cfg_altered): New file scoped static.
11696         (process_switch): If group_case_labels makes a change, then set
11697         cfg_altered.
11698         (pass_convert_switch::execute): If a switch is converted, then
11699         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
11700
11701         PR rtl-optimization/81308
11702         * recog.c (split_all_insns): Conditionally cleanup the CFG after
11703         splitting insns.
11704
11705 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
11706
11707         PR target/83663 - Revert r255946
11708         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11709         generation for cases where splatting a value is not useful.
11710         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11711         across a vec_duplicate and a paradoxical subreg forming a vector
11712         mode to a vec_concat.
11713
11714 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11715
11716         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11717         -march=armv8.3-a variants.
11718         * config/arm/t-multilib: Likewise.
11719         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
11720
11721 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
11722
11723         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11724         to generate rtl.
11725         (cceq_ior_compare_complement): Give it a name so I can use it, and
11726         change boolean_or_operator predicate to boolean_operator so it can
11727         be used to generate a crand.
11728         (eqne): New code iterator.
11729         (bd/bd_neg): New code_attrs.
11730         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11731         a single define_insn.
11732         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11733         decrement (bdnzt/bdnzf/bdzt/bdzf).
11734         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11735         with the new names of the branch decrement patterns, and added the
11736         names of the branch decrement conditional patterns.
11737
11738 2018-01-08  Richard Biener  <rguenther@suse.de>
11739
11740         PR tree-optimization/83563
11741         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11742         cache.
11743
11744 2018-01-08  Richard Biener  <rguenther@suse.de>
11745
11746         PR middle-end/83713
11747         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11748
11749 2018-01-08  Richard Biener  <rguenther@suse.de>
11750
11751         PR tree-optimization/83685
11752         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11753         references to abnormals.
11754
11755 2018-01-08  Richard Biener  <rguenther@suse.de>
11756
11757         PR lto/83719
11758         * dwarf2out.c (output_indirect_strings): Handle empty
11759         skeleton_debug_str_hash.
11760         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11761
11762 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
11763
11764         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11765         (emit_store_direct): Likewise.
11766         (arc_trampoline_adjust_address): Likewise.
11767         (arc_asm_trampoline_template): New function.
11768         (arc_initialize_trampoline): Use asm_trampoline_template.
11769         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11770         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11771         * config/arc/arc.md (flush_icache): Delete pattern.
11772
11773 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
11774
11775         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11776         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11777         munaligned-access.
11778
11779 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
11780
11781         PR target/83681
11782         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11783         by not USED_FOR_TARGET.
11784         (make_pass_resolve_sw_modes): Likewise.
11785
11786 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
11787
11788         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11789         USED_FOR_TARGET.
11790
11791 2018-01-08  Richard Biener  <rguenther@suse.de>
11792
11793         PR middle-end/83580
11794         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11795
11796 2018-01-08  Richard Biener  <rguenther@suse.de>
11797
11798         PR middle-end/83517
11799         * match.pd ((t * 2) / 2) -> t): Add missing :c.
11800
11801 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
11802
11803         PR middle-end/81897
11804         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11805         basic blocks with a small number of successors.
11806         (convert_control_dep_chain_into_preds): Improve handling of
11807         forwarder blocks.
11808         (dump_predicates): Split apart into...
11809         (dump_pred_chain): ...here...
11810         (dump_pred_info): ...and here.
11811         (can_one_predicate_be_invalidated_p): Add debugging printfs.
11812         (can_chain_union_be_invalidated_p): Improve check for invalidation
11813         of paths.
11814         (uninit_uses_cannot_happen): Avoid unnecessary if
11815         convert_control_dep_chain_into_preds yielded nothing.
11816
11817 2018-01-06  Martin Sebor  <msebor@redhat.com>
11818
11819         PR tree-optimization/83640
11820         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11821         subtracting negative offset from size.
11822         (builtin_access::overlap): Adjust offset bounds of the access to fall
11823         within the size of the object if possible.
11824
11825 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
11826
11827         PR rtl-optimization/83699
11828         * expmed.c (extract_bit_field_1): Restrict the vector usage of
11829         extract_bit_field_as_subreg to cases in which the extracted
11830         value is also a vector.
11831
11832         * lra-constraints.c (process_alt_operands): Test for the equivalence
11833         substitutions when detecting a possible reload cycle.
11834
11835 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
11836
11837         PR debug/83480
11838         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11839         by default if flag_selective_schedling{,2}.  Formatting fixes.
11840
11841         PR rtl-optimization/83682
11842         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11843         if it has non-VECTOR_MODE element mode.
11844         (vec_duplicate_p): Likewise.
11845
11846         PR middle-end/83694
11847         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11848         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11849
11850 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
11851
11852         PR target/83604
11853         * config/i386/i386-builtin.def
11854         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11855         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11856         Require also OPTION_MASK_ISA_AVX512F in addition to
11857         OPTION_MASK_ISA_GFNI.
11858         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11859         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11860         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11861         to OPTION_MASK_ISA_GFNI.
11862         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11863         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11864         OPTION_MASK_ISA_AVX512BW.
11865         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11866         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11867         addition to OPTION_MASK_ISA_GFNI.
11868         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11869         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11870         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11871         to OPTION_MASK_ISA_GFNI.
11872         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11873         a requirement for all ISAs rather than any of them with a few
11874         exceptions.
11875         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11876         processing.
11877         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11878         bitmasks to be enabled with 3 exceptions, instead of requiring any
11879         enabled ISA with lots of exceptions.
11880         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11881         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11882         Change avx512bw in isa attribute to avx512f.
11883         * config/i386/sgxintrin.h: Add license boilerplate.
11884         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
11885         to __AVX512F__ and __AVX512VL to __AVX512VL__.
11886         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11887         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11888         defined.
11889         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11890         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11891         temporarily sse2 rather than sse if not enabled already.
11892
11893         PR target/83604
11894         * config/i386/sse.md (VI248_VLBW): Rename to ...
11895         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
11896         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11897         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11898         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11899         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11900         mode iterator instead of VI248_VLBW.
11901
11902 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
11903
11904         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11905         (record_modified): Skip clobbers; add debug output.
11906         (param_change_prob): Use sreal frequencies.
11907
11908 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
11909
11910         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11911         punt for user-aligned variables.
11912
11913 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
11914
11915         * tree-chrec.c (chrec_contains_symbols): Return true for
11916         POLY_INT_CST.
11917
11918 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
11919
11920         PR target/82439
11921         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11922         of (x|y) == x for BICS pattern.
11923
11924 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
11925
11926         PR tree-optimization/83605
11927         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11928         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11929         can throw.
11930
11931 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
11932
11933         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11934         * config/epiphany/rtems.h: New file.
11935
11936 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
11937             Uros Bizjak  <ubizjak@gmail.com>
11938
11939         PR target/83554
11940         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11941         QIreg_operand instead of register_operand predicate.
11942         * config/i386/i386.c (ix86_rop_should_change_byte_p,
11943         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11944         comments instead of -fmitigate[-_]rop.
11945
11946 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
11947
11948         PR bootstrap/81926
11949         * cgraphunit.c (symbol_table::compile): Switch to text_section
11950         before calling assembly_start debug hook.
11951         * run-rtl-passes.c (run_rtl_passes): Likewise.
11952         Include output.h.
11953
11954 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
11955
11956         * tree-vrp.c (extract_range_from_binary_expr_1): Check
11957         range_int_cst_p rather than !symbolic_range_p before calling
11958         extract_range_from_multiplicative_op_1.
11959
11960 2018-01-04  Jeff Law  <law@redhat.com>
11961
11962         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11963         redundant test in assertion.
11964
11965 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
11966
11967         * doc/rtl.texi: Document machine_mode wrapper classes.
11968
11969 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
11970
11971         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11972         using tree_to_uhwi.
11973
11974 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
11975
11976         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11977         the VEC_PERM_EXPR fold to fail.
11978
11979 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
11980
11981         PR debug/83585
11982         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11983         to switched_sections.
11984
11985 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
11986
11987         PR target/83680
11988         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11989         test for d.testing.
11990
11991 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
11992
11993         PR target/83387
11994         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11995         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11996
11997 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
11998
11999         PR debug/83666
12000         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12001         is BLKmode and bitpos not zero or mode change is needed.
12002
12003 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
12004
12005         PR target/83675
12006         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12007         TARGET_VIS2.
12008
12009 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
12010
12011         PR target/83628
12012         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12013         instead of MULT rtx.  Update all corresponding splitters.
12014         (*saddl_se): Ditto.
12015         (*ssub<modesuffix>): Ditto.
12016         (*ssubl_se): Ditto.
12017         (*cmp_sadd_di): Update split patterns.
12018         (*cmp_sadd_si): Ditto.
12019         (*cmp_sadd_sidi): Ditto.
12020         (*cmp_ssub_di): Ditto.
12021         (*cmp_ssub_si): Ditto.
12022         (*cmp_ssub_sidi): Ditto.
12023         * config/alpha/predicates.md (const23_operand): New predicate.
12024         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12025         Look for ASHIFT, not MULT inner operand.
12026         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12027
12028 2018-01-04  Martin Liska  <mliska@suse.cz>
12029
12030         PR gcov-profile/83669
12031         * gcov.c (output_intermediate_file): Add version to intermediate
12032         gcov file.
12033         * doc/gcov.texi: Document new field 'version' in intermediate
12034         file format. Fix location of '-k' option of gcov command.
12035
12036 2018-01-04  Martin Liska  <mliska@suse.cz>
12037
12038         PR ipa/82352
12039         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12040
12041 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
12042
12043         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12044
12045 2018-01-03  Martin Sebor  <msebor@redhat.com>
12046
12047         PR tree-optimization/83655
12048         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12049         checking calls with invalid arguments.
12050
12051 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12052
12053         * tree-vect-stmts.c (vect_get_store_rhs): New function.
12054         (vectorizable_mask_load_store): Delete.
12055         (vectorizable_call): Return false for masked loads and stores.
12056         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
12057         instead of gimple_assign_rhs1.
12058         (vectorizable_load): Handle IFN_MASK_LOAD.
12059         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12060
12061 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12062
12063         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12064         split out from..,
12065         (vectorizable_mask_load_store): ...here.
12066         (vectorizable_load): ...and here.
12067
12068 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12069
12070         * tree-vect-stmts.c (vect_build_all_ones_mask)
12071         (vect_build_zero_merge_argument): New functions, split out from...
12072         (vectorizable_load): ...here.
12073
12074 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12075
12076         * tree-vect-stmts.c (vect_check_store_rhs): New function,
12077         split out from...
12078         (vectorizable_mask_load_store): ...here.
12079         (vectorizable_store): ...and here.
12080
12081 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12082
12083         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12084         split out from...
12085         (vectorizable_mask_load_store): ...here.
12086
12087 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12088
12089         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12090         (vect_model_store_cost): Take a vec_load_store_type instead of a
12091         vect_def_type.
12092         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12093         (vect_model_store_cost): Take a vec_load_store_type instead of a
12094         vect_def_type.
12095         (vectorizable_mask_load_store): Update accordingly.
12096         (vectorizable_store): Likewise.
12097         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12098
12099 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12100
12101         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12102         IFN_MASK_LOAD calls here rather than...
12103         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12104
12105 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12106             Alan Hayward  <alan.hayward@arm.com>
12107             David Sherwood  <david.sherwood@arm.com>
12108
12109         * expmed.c (extract_bit_field_1): For vector extracts,
12110         fall back to extract_bit_field_as_subreg if vec_extract
12111         isn't available.
12112
12113 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12114             Alan Hayward  <alan.hayward@arm.com>
12115             David Sherwood  <david.sherwood@arm.com>
12116
12117         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12118         they are variable or constant sized.
12119         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12120         slots for constant-sized data.
12121
12122 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12123             Alan Hayward  <alan.hayward@arm.com>
12124             David Sherwood  <david.sherwood@arm.com>
12125
12126         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12127         handling COND_EXPRs with boolean comparisons, try to find a better
12128         basis for the mask type than the boolean itself.
12129
12130 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12131
12132         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12133         is calculated and how it can be overridden.
12134         * genmodes.c (max_bitsize_mode_any_mode): New variable.
12135         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12136         if defined.
12137         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12138         if nonzero.
12139
12140 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12141             Alan Hayward  <alan.hayward@arm.com>
12142             David Sherwood  <david.sherwood@arm.com>
12143
12144         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12145         Remove the mode argument.
12146         (aarch64_simd_valid_immediate): Remove the mode and inverse
12147         arguments.
12148         * config/aarch64/iterators.md (bitsize): New iterator.
12149         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12150         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12151         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12152         aarch64_simd_valid_immediate.
12153         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12154         (aarch64_reg_or_bic_imm): Likewise.
12155         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12156         with an insn_type enum and msl with a modifier_type enum.
12157         Replace element_width with a scalar_mode.  Change the shift
12158         to unsigned int.  Add constructors for scalar_float_mode and
12159         scalar_int_mode elements.
12160         (aarch64_vect_float_const_representable_p): Delete.
12161         (aarch64_can_const_movi_rtx_p)
12162         (aarch64_simd_scalar_immediate_valid_for_move)
12163         (aarch64_simd_make_constant): Update call to
12164         aarch64_simd_valid_immediate.
12165         (aarch64_advsimd_valid_immediate_hs): New function.
12166         (aarch64_advsimd_valid_immediate): Likewise.
12167         (aarch64_simd_valid_immediate): Remove mode and inverse
12168         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
12169         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12170         and aarch64_float_const_representable_p on the result.
12171         (aarch64_output_simd_mov_immediate): Remove mode argument.
12172         Update call to aarch64_simd_valid_immediate and use of
12173         simd_immediate_info.
12174         (aarch64_output_scalar_simd_mov_immediate): Update call
12175         accordingly.
12176
12177 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12178             Alan Hayward  <alan.hayward@arm.com>
12179             David Sherwood  <david.sherwood@arm.com>
12180
12181         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12182         (mode_nunits): Likewise CONST_MODE_NUNITS.
12183         * machmode.def (ADJUST_NUNITS): Document.
12184         * genmodes.c (mode_data::need_nunits_adj): New field.
12185         (blank_mode): Update accordingly.
12186         (adj_nunits): New variable.
12187         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12188         parameter.
12189         (emit_mode_size_inline): Set need_bytesize_adj for all modes
12190         listed in adj_nunits.
12191         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12192         listed in adj_nunits.  Don't emit case statements for such modes.
12193         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12194         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
12195         nothing if adj_nunits is nonnull.
12196         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12197         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12198         (emit_mode_fbit): Update use of print_maybe_const_decl.
12199         (emit_move_size): Likewise.  Treat the array as non-const
12200         if adj_nunits.
12201         (emit_mode_adjustments): Handle adj_nunits.
12202
12203 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12204
12205         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12206         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12207         (VECTOR_MODES): Use it.
12208         (make_vector_modes): Take the prefix as an argument.
12209
12210 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12211             Alan Hayward  <alan.hayward@arm.com>
12212             David Sherwood  <david.sherwood@arm.com>
12213
12214         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12215         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12216         for MODE_VECTOR_BOOL.
12217         * machmode.def (VECTOR_BOOL_MODE): Document.
12218         * genmodes.c (VECTOR_BOOL_MODE): New macro.
12219         (make_vector_bool_mode): New function.
12220         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12221         MODE_VECTOR_BOOL.
12222         * lto-streamer-in.c (lto_input_mode_table): Likewise.
12223         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12224         Likewise.
12225         * stor-layout.c (int_mode_for_mode): Likewise.
12226         * tree.c (build_vector_type_for_mode): Likewise.
12227         * varasm.c (output_constant_pool_2): Likewise.
12228         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12229         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
12230         for MODE_VECTOR_BOOL.
12231         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12232         of mode class checks.
12233         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12234         instead of a list of mode class checks.
12235         (expand_vector_scalar_condition): Likewise.
12236         (type_for_widest_vector_mode): Handle BImode as an inner mode.
12237
12238 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12239             Alan Hayward  <alan.hayward@arm.com>
12240             David Sherwood  <david.sherwood@arm.com>
12241
12242         * machmode.h (mode_size): Change from unsigned short to
12243         poly_uint16_pod.
12244         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12245         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12246         or if measurement_type is not polynomial.
12247         (fixed_size_mode::includes_p): Check for constant-sized modes.
12248         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12249         return a poly_uint16 rather than an unsigned short.
12250         (emit_mode_size): Change the type of mode_size from unsigned short
12251         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
12252         (emit_mode_adjustments): Cope with polynomial vector sizes.
12253         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12254         for GET_MODE_SIZE.
12255         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12256         for GET_MODE_SIZE.
12257         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12258         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12259         * caller-save.c (setup_save_areas): Likewise.
12260         (replace_reg_with_saved_mem): Likewise.
12261         * calls.c (emit_library_call_value_1): Likewise.
12262         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12263         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12264         (gen_lowpart_for_combine): Likewise.
12265         * convert.c (convert_to_integer_1): Likewise.
12266         * cse.c (equiv_constant, cse_insn): Likewise.
12267         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12268         (cselib_subst_to_values): Likewise.
12269         * dce.c (word_dce_process_block): Likewise.
12270         * df-problems.c (df_word_lr_mark_ref): Likewise.
12271         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12272         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12273         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12274         (rtl_for_decl_location): Likewise.
12275         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12276         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12277         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12278         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12279         (expand_expr_real_1): Likewise.
12280         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12281         (pad_below): Likewise.
12282         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12283         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12284         * ira.c (get_subreg_tracking_sizes): Likewise.
12285         * ira-build.c (ira_create_allocno_objects): Likewise.
12286         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12287         (ira_sort_regnos_for_alter_reg): Likewise.
12288         * ira-costs.c (record_operand_costs): Likewise.
12289         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12290         (resolve_simple_move): Likewise.
12291         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12292         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12293         (lra_constraints): Likewise.
12294         (CONST_POOL_OK_P): Reject variable-sized modes.
12295         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12296         (add_pseudo_to_slot, lra_spill): Likewise.
12297         * omp-low.c (omp_clause_aligned_alignment): Likewise.
12298         * optabs-query.c (get_best_extraction_insn): Likewise.
12299         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12300         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12301         (expand_mult_highpart, valid_multiword_target_p): Likewise.
12302         * recog.c (offsettable_address_addr_space_p): Likewise.
12303         * regcprop.c (maybe_mode_change): Likewise.
12304         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12305         * regrename.c (build_def_use): Likewise.
12306         * regstat.c (dump_reg_info): Likewise.
12307         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12308         (find_reloads, find_reloads_subreg_address): Likewise.
12309         * reload1.c (eliminate_regs_1): Likewise.
12310         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12311         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12312         (simplify_binary_operation_1, simplify_subreg): Likewise.
12313         * targhooks.c (default_function_arg_padding): Likewise.
12314         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12315         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12316         (verify_gimple_assign_ternary): Likewise.
12317         * tree-inline.c (estimate_move_cost): Likewise.
12318         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12319         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12320         (get_address_cost_ainc): Likewise.
12321         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12322         (vect_supportable_dr_alignment): Likewise.
12323         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12324         (vectorizable_reduction): Likewise.
12325         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12326         (vectorizable_operation, vectorizable_load): Likewise.
12327         * tree.c (build_same_sized_truth_vector_type): Likewise.
12328         * valtrack.c (cleanup_auto_inc_dec): Likewise.
12329         * var-tracking.c (emit_note_insn_var_location): Likewise.
12330         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12331         (ADDR_VEC_ALIGN): Likewise.
12332
12333 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12334             Alan Hayward  <alan.hayward@arm.com>
12335             David Sherwood  <david.sherwood@arm.com>
12336
12337         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12338         unsigned short.
12339         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12340         or if measurement_type is polynomial.
12341         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12342         * combine.c (make_extraction): Likewise.
12343         * dse.c (find_shift_sequence): Likewise.
12344         * dwarf2out.c (mem_loc_descriptor): Likewise.
12345         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12346         (extract_bit_field, extract_low_bits): Likewise.
12347         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12348         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12349         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12350         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12351         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12352         * reload.c (find_reloads): Likewise.
12353         * reload1.c (alter_reg): Likewise.
12354         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12355         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12356         * tree-if-conv.c (predicate_mem_writes): Likewise.
12357         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12358         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12359         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12360         * valtrack.c (dead_debug_insert_temp): Likewise.
12361         * varasm.c (mergeable_constant_section): Likewise.
12362         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12363
12364 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12365             Alan Hayward  <alan.hayward@arm.com>
12366             David Sherwood  <david.sherwood@arm.com>
12367
12368         * expr.c (expand_assignment): Cope with polynomial mode sizes
12369         when assigning to a CONCAT.
12370
12371 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12372             Alan Hayward  <alan.hayward@arm.com>
12373             David Sherwood  <david.sherwood@arm.com>
12374
12375         * machmode.h (mode_precision): Change from unsigned short to
12376         poly_uint16_pod.
12377         (mode_to_precision): Return a poly_uint16 rather than an unsigned
12378         short.
12379         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12380         or if measurement_type is not polynomial.
12381         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
12382         in which the mode is already known to be a scalar_int_mode.
12383         * genmodes.c (emit_mode_precision): Change the type of mode_precision
12384         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
12385         initializer.
12386         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12387         for GET_MODE_PRECISION.
12388         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12389         for GET_MODE_PRECISION.
12390         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12391         as polynomial.
12392         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12393         (expand_field_assignment, make_extraction): Likewise.
12394         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12395         (get_last_value): Likewise.
12396         * convert.c (convert_to_integer_1): Likewise.
12397         * cse.c (cse_insn): Likewise.
12398         * expr.c (expand_expr_real_1): Likewise.
12399         * lra-constraints.c (simplify_operand_subreg): Likewise.
12400         * optabs-query.c (can_atomic_load_p): Likewise.
12401         * optabs.c (expand_atomic_load): Likewise.
12402         (expand_atomic_store): Likewise.
12403         * ree.c (combine_reaching_defs): Likewise.
12404         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12405         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12406         * tree.h (type_has_mode_precision_p): Likewise.
12407         * ubsan.c (instrument_si_overflow): Likewise.
12408
12409 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12410             Alan Hayward  <alan.hayward@arm.com>
12411             David Sherwood  <david.sherwood@arm.com>
12412
12413         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12414         polynomial numbers of units.
12415         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12416         (valid_vector_subparts_p): New function.
12417         (build_vector_type): Remove temporary shim and take the number
12418         of units as a poly_uint64 rather than an int.
12419         (build_opaque_vector_type): Take the number of units as a
12420         poly_uint64 rather than an int.
12421         * tree.c (build_vector_from_ctor): Handle polynomial
12422         TYPE_VECTOR_SUBPARTS.
12423         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12424         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12425         (build_vector_from_val): If the number of units is variable,
12426         use build_vec_duplicate_cst for constant operands and
12427         VEC_DUPLICATE_EXPR otherwise.
12428         (make_vector_type): Remove temporary is_constant ().
12429         (build_vector_type, build_opaque_vector_type): Take the number of
12430         units as a poly_uint64 rather than an int.
12431         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12432         VECTOR_CST_NELTS.
12433         * cfgexpand.c (expand_debug_expr): Likewise.
12434         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12435         (store_constructor, expand_expr_real_1): Likewise.
12436         (const_scalar_mask_from_tree): Likewise.
12437         * fold-const-call.c (fold_const_reduction): Likewise.
12438         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12439         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12440         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12441         (fold_relational_const): Likewise.
12442         (native_interpret_vector): Likewise.  Change the size from an
12443         int to an unsigned int.
12444         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12445         TYPE_VECTOR_SUBPARTS.
12446         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12447         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12448         duplicating a non-constant operand into a variable-length vector.
12449         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12450         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12451         * ipa-icf.c (sem_variable::equals): Likewise.
12452         * match.pd: Likewise.
12453         * omp-simd-clone.c (simd_clone_subparts): Likewise.
12454         * print-tree.c (print_node): Likewise.
12455         * stor-layout.c (layout_type): Likewise.
12456         * targhooks.c (default_builtin_vectorization_cost): Likewise.
12457         * tree-cfg.c (verify_gimple_comparison): Likewise.
12458         (verify_gimple_assign_binary): Likewise.
12459         (verify_gimple_assign_ternary): Likewise.
12460         (verify_gimple_assign_single): Likewise.
12461         * tree-pretty-print.c (dump_generic_node): Likewise.
12462         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12463         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12464         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12465         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12466         (vect_shift_permute_load_chain): Likewise.
12467         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12468         (expand_vector_condition, optimize_vector_constructor): Likewise.
12469         (lower_vec_perm, get_compute_type): Likewise.
12470         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12471         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12472         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12473         (vect_recog_mask_conversion_pattern): Likewise.
12474         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12475         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12476         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12477         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12478         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12479         (vectorizable_shift, vectorizable_operation, vectorizable_store)
12480         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12481         (supportable_widening_operation): Likewise.
12482         (supportable_narrowing_operation): Likewise.
12483         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12484         Likewise.
12485         * varasm.c (output_constant): Likewise.
12486
12487 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12488             Alan Hayward  <alan.hayward@arm.com>
12489             David Sherwood  <david.sherwood@arm.com>
12490
12491         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12492         so that both the length == 3 and length != 3 cases set up their
12493         own permute vectors.  Add comments explaining why we know the
12494         number of elements is constant.
12495         (vect_permute_load_chain): Likewise.
12496
12497 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12498             Alan Hayward  <alan.hayward@arm.com>
12499             David Sherwood  <david.sherwood@arm.com>
12500
12501         * machmode.h (mode_nunits): Change from unsigned char to
12502         poly_uint16_pod.
12503         (ONLY_FIXED_SIZE_MODES): New macro.
12504         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12505         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12506         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12507         New typedefs.
12508         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12509         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12510         or if measurement_type is not polynomial.
12511         * genmodes.c (ZERO_COEFFS): New macro.
12512         (emit_mode_nunits_inline): Make mode_nunits_inline return a
12513         poly_uint16.
12514         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12515         Use ZERO_COEFFS when emitting initializers.
12516         * data-streamer.h (bp_pack_poly_value): New function.
12517         (bp_unpack_poly_value): Likewise.
12518         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12519         for GET_MODE_NUNITS.
12520         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12521         for GET_MODE_NUNITS.
12522         * tree.c (make_vector_type): Remove temporary shim and make
12523         the real function take the number of units as a poly_uint64
12524         rather than an int.
12525         (build_vector_type_for_mode): Handle polynomial nunits.
12526         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12527         * emit-rtl.c (const_vec_series_p_1): Likewise.
12528         (gen_rtx_CONST_VECTOR): Likewise.
12529         * fold-const.c (test_vec_duplicate_folding): Likewise.
12530         * genrecog.c (validate_pattern): Likewise.
12531         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12532         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12533         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12534         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12535         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12536         * rtlanal.c (subreg_get_info): Likewise.
12537         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12538         (vect_grouped_load_supported): Likewise.
12539         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12540         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12541         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12542         (simplify_const_unary_operation, simplify_binary_operation_1)
12543         (simplify_const_binary_operation, simplify_ternary_operation)
12544         (test_vector_ops_duplicate, test_vector_ops): Likewise.
12545         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12546         instead of CONST_VECTOR_NUNITS.
12547         * varasm.c (output_constant_pool_2): Likewise.
12548         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12549         explicit-encoded elements in the XVEC for variable-length vectors.
12550
12551 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12552
12553         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12554
12555 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12556             Alan Hayward  <alan.hayward@arm.com>
12557             David Sherwood  <david.sherwood@arm.com>
12558
12559         * coretypes.h (fixed_size_mode): Declare.
12560         (fixed_size_mode_pod): New typedef.
12561         * builtins.h (target_builtins::x_apply_args_mode)
12562         (target_builtins::x_apply_result_mode): Change type to
12563         fixed_size_mode_pod.
12564         * builtins.c (apply_args_size, apply_result_size, result_vector)
12565         (expand_builtin_apply_args_1, expand_builtin_apply)
12566         (expand_builtin_return): Update accordingly.
12567
12568 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12569
12570         * cse.c (hash_rtx_cb): Hash only the encoded elements.
12571         * cselib.c (cselib_hash_rtx): Likewise.
12572         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12573         CONST_VECTOR encoding.
12574
12575 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
12576             Jeff Law  <law@redhat.com>
12577
12578         PR target/83641
12579         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12580         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12581         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12582         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12583
12584         PR target/83641
12585         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12586         explicitly probe *sp in a noreturn function if there were any callee
12587         register saves or frame pointer is needed.
12588
12589 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
12590
12591         PR debug/83621
12592         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12593         BLKmode for ternary, binary or unary expressions.
12594
12595         PR debug/83645
12596         * var-tracking.c (delete_vta_debug_insn): New inline function.
12597         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12598         insns from get_insns () to NULL instead of each bb separately.
12599         Use delete_vta_debug_insn.  No longer static.
12600         (vt_debug_insns_local, variable_tracking_main_1): Adjust
12601         delete_vta_debug_insns callers.
12602         * rtl.h (delete_vta_debug_insns): Declare.
12603         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12604         instead of variable_tracking_main.
12605
12606 2018-01-03  Martin Sebor  <msebor@redhat.com>
12607
12608         PR tree-optimization/83603
12609         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12610         arguments past the endof the argument list in functions declared
12611         without a prototype.
12612         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12613         Avoid checking when arguments are null.
12614
12615 2018-01-03  Martin Sebor  <msebor@redhat.com>
12616
12617         PR c/83559
12618         * doc/extend.texi (attribute const): Fix a typo.
12619         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12620         issuing -Wsuggest-attribute for void functions.
12621
12622 2018-01-03  Martin Sebor  <msebor@redhat.com>
12623
12624         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12625         offset_int::from instead of wide_int::to_shwi.
12626         (maybe_diag_overlap): Remove assertion.
12627         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12628         * gimple-ssa-sprintf.c (format_directive): Same.
12629         (parse_directive): Same.
12630         (sprintf_dom_walker::compute_format_length): Same.
12631         (try_substitute_return_value): Same.
12632
12633 2018-01-03  Jeff Law  <law@redhat.com>
12634
12635         PR middle-end/83654
12636         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12637         non-constant residual for zero at runtime and avoid probing in
12638         that case.  Reorganize code for trailing problem to mirror handling
12639         of the residual.
12640
12641 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
12642
12643         PR tree-optimization/83501
12644         * tree-ssa-strlen.c (get_string_cst): New.
12645         (handle_char_store): Call get_string_cst.
12646
12647 2018-01-03  Martin Liska  <mliska@suse.cz>
12648
12649         PR tree-optimization/83593
12650         * tree-ssa-strlen.c: Include tree-cfg.h.
12651         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12652         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12653         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12654         to false.
12655         (strlen_dom_walker::before_dom_children): Call
12656         gimple_purge_dead_eh_edges. Dump tranformation with details
12657         dump flags.
12658         (strlen_dom_walker::before_dom_children): Update call by adding
12659         new argument cleanup_eh.
12660         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12661
12662 2018-01-03  Martin Liska  <mliska@suse.cz>
12663
12664         PR ipa/83549
12665         * cif-code.def (VARIADIC_THUNK): New enum value.
12666         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12667         thunks.
12668
12669 2018-01-03  Jan Beulich  <jbeulich@suse.com>
12670
12671         * sse.md (mov<mode>_internal): Tighten condition for when to use
12672         vmovdqu<ssescalarsize> for TI and OI modes.
12673
12674 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
12675
12676         Update copyright years.
12677
12678 2018-01-03  Martin Liska  <mliska@suse.cz>
12679
12680         PR ipa/83594
12681         * ipa-visibility.c (function_and_variable_visibility): Skip
12682         functions with noipa attribure.
12683
12684 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
12685
12686         * gcc.c (process_command): Update copyright notice dates.
12687         * gcov-dump.c (print_version): Ditto.
12688         * gcov.c (print_version): Ditto.
12689         * gcov-tool.c (print_version): Ditto.
12690         * gengtype.c (create_file): Ditto.
12691         * doc/cpp.texi: Bump @copying's copyright year.
12692         * doc/cppinternals.texi: Ditto.
12693         * doc/gcc.texi: Ditto.
12694         * doc/gccint.texi: Ditto.
12695         * doc/gcov.texi: Ditto.
12696         * doc/install.texi: Ditto.
12697         * doc/invoke.texi: Ditto.
12698
12699 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12700
12701         * vector-builder.h (vector_builder::m_full_nelts): Change from
12702         unsigned int to poly_uint64.
12703         (vector_builder::full_nelts): Update prototype accordingly.
12704         (vector_builder::new_vector): Likewise.
12705         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12706         (vector_builder::operator ==): Likewise.
12707         (vector_builder::finalize): Likewise.
12708         * int-vector-builder.h (int_vector_builder::int_vector_builder):
12709         Take the number of elements as a poly_uint64 rather than an
12710         unsigned int.
12711         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12712         from unsigned int to poly_uint64.
12713         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12714         (vec_perm_indices::new_vector): Likewise.
12715         (vec_perm_indices::length): Likewise.
12716         (vec_perm_indices::nelts_per_input): Likewise.
12717         (vec_perm_indices::input_nelts): Likewise.
12718         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12719         number of elements per input as a poly_uint64 rather than an
12720         unsigned int.  Use the original encoding for variable-length
12721         vectors, rather than clamping each individual element.
12722         For the second and subsequent elements in each pattern,
12723         clamp the step and base before clamping their sum.
12724         (vec_perm_indices::series_p): Handle polynomial element counts.
12725         (vec_perm_indices::all_in_range_p): Likewise.
12726         (vec_perm_indices_to_tree): Likewise.
12727         (vec_perm_indices_to_rtx): Likewise.
12728         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12729         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12730         (tree_vector_builder::new_binary_operation): Handle polynomial
12731         element counts.  Return false if we need to know the number
12732         of elements at compile time.
12733         * fold-const.c (fold_vec_perm): Punt if the number of elements
12734         isn't known at compile time.
12735
12736 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12737
12738         * vec-perm-indices.h (vec_perm_builder): Change element type
12739         from HOST_WIDE_INT to poly_int64.
12740         (vec_perm_indices::element_type): Update accordingly.
12741         (vec_perm_indices::clamp): Handle polynomial element_types.
12742         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12743         (vec_perm_indices::all_in_range_p): Likewise.
12744         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12745         than shwi trees.
12746         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12747         polynomial vec_perm_indices element types.
12748         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12749         * fold-const.c (fold_vec_perm): Likewise.
12750         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12751         * tree-vect-generic.c (lower_vec_perm): Likewise.
12752         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12753         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12754         element type to HOST_WIDE_INT.
12755
12756 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12757             Alan Hayward  <alan.hayward@arm.com>
12758             David Sherwood  <david.sherwood@arm.com>
12759
12760         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12761         rather than an int.  Use plus_constant.
12762         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12763         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12764
12765 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12766             Alan Hayward  <alan.hayward@arm.com>
12767             David Sherwood  <david.sherwood@arm.com>
12768
12769         * calls.c (emit_call_1, expand_call): Change struct_value_size from
12770         a HOST_WIDE_INT to a poly_int64.
12771
12772 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12773             Alan Hayward  <alan.hayward@arm.com>
12774             David Sherwood  <david.sherwood@arm.com>
12775
12776         * calls.c (load_register_parameters): Cope with polynomial
12777         mode sizes.  Require a constant size for BLKmode parameters
12778         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
12779         forces a parameter to be padded at the lsb end in order to
12780         fill a complete number of words, require the parameter size
12781         to be ordered wrt UNITS_PER_WORD.
12782
12783 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12784             Alan Hayward  <alan.hayward@arm.com>
12785             David Sherwood  <david.sherwood@arm.com>
12786
12787         * reload1.c (spill_stack_slot_width): Change element type
12788         from unsigned int to poly_uint64_pod.
12789         (alter_reg): Treat mode sizes as polynomial.
12790
12791 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12792             Alan Hayward  <alan.hayward@arm.com>
12793             David Sherwood  <david.sherwood@arm.com>
12794
12795         * reload.c (complex_word_subreg_p): New function.
12796         (reload_inner_reg_of_subreg, push_reload): Use it.
12797
12798 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12799             Alan Hayward  <alan.hayward@arm.com>
12800             David Sherwood  <david.sherwood@arm.com>
12801
12802         * lra-constraints.c (process_alt_operands): Reject matched
12803         operands whose sizes aren't ordered.
12804         (match_reload): Refer to this check here.
12805
12806 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12807             Alan Hayward  <alan.hayward@arm.com>
12808             David Sherwood  <david.sherwood@arm.com>
12809
12810         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12811         that the mode size is in the set {1, 2, 4, 8, 16}.
12812
12813 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12814             Alan Hayward  <alan.hayward@arm.com>
12815             David Sherwood  <david.sherwood@arm.com>
12816
12817         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12818         Use plus_constant instead of gen_rtx_PLUS.
12819
12820 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12821             Alan Hayward  <alan.hayward@arm.com>
12822             David Sherwood  <david.sherwood@arm.com>
12823
12824         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12825         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12826         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12827         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12828         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12829         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12830         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12831         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12832         * config/i386/i386.c (ix86_push_rounding): ...this new function.
12833         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12834         a poly_int64.
12835         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12836         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12837         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12838         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12839         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12840         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12841         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12842         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12843         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12844         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12845         function.
12846         * expr.c (emit_move_resolve_push): Treat the input and result
12847         of PUSH_ROUNDING as a poly_int64.
12848         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12849         (emit_push_insn): Likewise.
12850         * lra-eliminations.c (mark_not_eliminable): Likewise.
12851         * recog.c (push_operand): Likewise.
12852         * reload1.c (elimination_effects): Likewise.
12853         * rtlanal.c (nonzero_bits1): Likewise.
12854         * calls.c (store_one_arg): Likewise.  Require the padding to be
12855         known at compile time.
12856
12857 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12858             Alan Hayward  <alan.hayward@arm.com>
12859             David Sherwood  <david.sherwood@arm.com>
12860
12861         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12862         Use plus_constant instead of gen_rtx_PLUS.
12863
12864 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12865             Alan Hayward  <alan.hayward@arm.com>
12866             David Sherwood  <david.sherwood@arm.com>
12867
12868         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12869         rather than an int.
12870
12871 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12872             Alan Hayward  <alan.hayward@arm.com>
12873             David Sherwood  <david.sherwood@arm.com>
12874
12875         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12876         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12877         via stack temporaries.  Treat the mode size as polynomial too.
12878
12879 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12880             Alan Hayward  <alan.hayward@arm.com>
12881             David Sherwood  <david.sherwood@arm.com>
12882
12883         * expr.c (expand_expr_real_2): When handling conversions involving
12884         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12885         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
12886         as a poly_uint64 too.
12887
12888 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12889             Alan Hayward  <alan.hayward@arm.com>
12890             David Sherwood  <david.sherwood@arm.com>
12891
12892         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12893
12894 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12895             Alan Hayward  <alan.hayward@arm.com>
12896             David Sherwood  <david.sherwood@arm.com>
12897
12898         * combine.c (can_change_dest_mode): Handle polynomial
12899         REGMODE_NATURAL_SIZE.
12900         * expmed.c (store_bit_field_1): Likewise.
12901         * expr.c (store_constructor): Likewise.
12902         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12903         and polynomial REGMODE_NATURAL_SIZE.
12904         (gen_lowpart_common): Likewise.
12905         * reginfo.c (record_subregs_of_mode): Likewise.
12906         * rtlanal.c (read_modify_subreg_p): Likewise.
12907
12908 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12909             Alan Hayward  <alan.hayward@arm.com>
12910             David Sherwood  <david.sherwood@arm.com>
12911
12912         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12913         numbers of elements.
12914
12915 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12916             Alan Hayward  <alan.hayward@arm.com>
12917             David Sherwood  <david.sherwood@arm.com>
12918
12919         * match.pd: Cope with polynomial numbers of vector elements.
12920
12921 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12922             Alan Hayward  <alan.hayward@arm.com>
12923             David Sherwood  <david.sherwood@arm.com>
12924
12925         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12926         in a POINTER_PLUS_EXPR.
12927
12928 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12929             Alan Hayward  <alan.hayward@arm.com>
12930             David Sherwood  <david.sherwood@arm.com>
12931
12932         * omp-simd-clone.c (simd_clone_subparts): New function.
12933         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12934         (ipa_simd_modify_function_body): Likewise.
12935
12936 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12937             Alan Hayward  <alan.hayward@arm.com>
12938             David Sherwood  <david.sherwood@arm.com>
12939
12940         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12941         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12942         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12943         (expand_vector_condition, vector_element): Likewise.
12944         (subparts_gt): New function.
12945         (get_compute_type): Use subparts_gt.
12946         (count_type_subparts): Delete.
12947         (expand_vector_operations_1): Use subparts_gt instead of
12948         count_type_subparts.
12949
12950 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12951             Alan Hayward  <alan.hayward@arm.com>
12952             David Sherwood  <david.sherwood@arm.com>
12953
12954         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12955         (vect_compile_time_alias): ...this new function.  Do the calculation
12956         on poly_ints rather than trees.
12957         (vect_prune_runtime_alias_test_list): Update call accordingly.
12958
12959 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12960             Alan Hayward  <alan.hayward@arm.com>
12961             David Sherwood  <david.sherwood@arm.com>
12962
12963         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12964         numbers of units.
12965         (vect_schedule_slp_instance): Likewise.
12966
12967 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12968             Alan Hayward  <alan.hayward@arm.com>
12969             David Sherwood  <david.sherwood@arm.com>
12970
12971         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12972         constant and extern definitions for variable-length vectors.
12973         (vect_get_constant_vectors): Note that the number of units
12974         is known to be constant.
12975
12976 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12977             Alan Hayward  <alan.hayward@arm.com>
12978             David Sherwood  <david.sherwood@arm.com>
12979
12980         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12981         of units as polynomial.  Choose between WIDE and NARROW based
12982         on multiple_p.
12983
12984 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12985             Alan Hayward  <alan.hayward@arm.com>
12986             David Sherwood  <david.sherwood@arm.com>
12987
12988         * tree-vect-stmts.c (simd_clone_subparts): New function.
12989         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12990
12991 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12992             Alan Hayward  <alan.hayward@arm.com>
12993             David Sherwood  <david.sherwood@arm.com>
12994
12995         * tree-vect-stmts.c (vectorizable_call): Treat the number of
12996         vectors as polynomial.  Use build_index_vector for
12997         IFN_GOMP_SIMD_LANE.
12998
12999 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13000             Alan Hayward  <alan.hayward@arm.com>
13001             David Sherwood  <david.sherwood@arm.com>
13002
13003         * tree-vect-stmts.c (get_load_store_type): Treat the number of
13004         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13005         for variable-length vectors.
13006         (vectorizable_mask_load_store): Treat the number of units as
13007         polynomial, asserting that it is constant if the condition has
13008         already been enforced.
13009         (vectorizable_store, vectorizable_load): Likewise.
13010
13011 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13012             Alan Hayward  <alan.hayward@arm.com>
13013             David Sherwood  <david.sherwood@arm.com>
13014
13015         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13016         of units as polynomial.  Punt if we can't tell at compile time
13017         which vector contains the final result.
13018
13019 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13020             Alan Hayward  <alan.hayward@arm.com>
13021             David Sherwood  <david.sherwood@arm.com>
13022
13023         * tree-vect-loop.c (vectorizable_induction): Treat the number
13024         of units as polynomial.  Punt on SLP inductions.  Use an integer
13025         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
13026         cast of such a series for variable-length floating-point
13027         reductions.
13028
13029 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13030             Alan Hayward  <alan.hayward@arm.com>
13031             David Sherwood  <david.sherwood@arm.com>
13032
13033         * tree.h (build_index_vector): Declare.
13034         * tree.c (build_index_vector): New function.
13035         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13036         of units as polynomial, forcibly converting it to a constant if
13037         vectorizable_reduction has already enforced the condition.
13038         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
13039         to create a {1,2,3,...} vector.
13040         (vectorizable_reduction): Treat the number of units as polynomial.
13041         Choose vectype_in based on the largest scalar element size rather
13042         than the smallest number of units.  Enforce the restrictions
13043         relied on above.
13044
13045 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13046             Alan Hayward  <alan.hayward@arm.com>
13047             David Sherwood  <david.sherwood@arm.com>
13048
13049         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13050         number of units as polynomial.
13051
13052 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13053             Alan Hayward  <alan.hayward@arm.com>
13054             David Sherwood  <david.sherwood@arm.com>
13055
13056         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13057         * target.def (autovectorize_vector_sizes): Return the vector sizes
13058         by pointer, using vector_sizes rather than a bitmask.
13059         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13060         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13061         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13062         Likewise.
13063         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13064         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13065         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13066         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13067         * omp-general.c (omp_max_vf): Likewise.
13068         * omp-low.c (omp_clause_aligned_alignment): Likewise.
13069         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13070         * tree-vect-loop.c (vect_analyze_loop): Likewise.
13071         * tree-vect-slp.c (vect_slp_bb): Likewise.
13072         * doc/tm.texi: Regenerate.
13073         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13074         to a poly_uint64.
13075         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13076         the vector size as a poly_uint64 rather than an unsigned int.
13077         (current_vector_size): Change from an unsigned int to a poly_uint64.
13078         (get_vectype_for_scalar_type): Update accordingly.
13079         * tree.h (build_truth_vector_type): Take the size and number of
13080         units as a poly_uint64 rather than an unsigned int.
13081         (build_vector_type): Add a temporary overload that takes
13082         the number of units as a poly_uint64 rather than an unsigned int.
13083         * tree.c (make_vector_type): Likewise.
13084         (build_truth_vector_type): Take the number of units as a poly_uint64
13085         rather than an unsigned int.
13086
13087 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13088             Alan Hayward  <alan.hayward@arm.com>
13089             David Sherwood  <david.sherwood@arm.com>
13090
13091         * target.def (get_mask_mode): Take the number of units and length
13092         as poly_uint64s rather than unsigned ints.
13093         * targhooks.h (default_get_mask_mode): Update accordingly.
13094         * targhooks.c (default_get_mask_mode): Likewise.
13095         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13096         * doc/tm.texi: Regenerate.
13097
13098 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13099             Alan Hayward  <alan.hayward@arm.com>
13100             David Sherwood  <david.sherwood@arm.com>
13101
13102         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13103         * omp-general.c (omp_max_vf): Likewise.
13104         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13105         (expand_omp_simd): Handle polynomial safelen.
13106         * omp-low.c (omplow_simd_context): Add a default constructor.
13107         (omplow_simd_context::max_vf): Change from int to poly_uint64.
13108         (lower_rec_simd_input_clauses): Update accordingly.
13109         (lower_rec_input_clauses): Likewise.
13110
13111 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13112             Alan Hayward  <alan.hayward@arm.com>
13113             David Sherwood  <david.sherwood@arm.com>
13114
13115         * tree-vectorizer.h (vect_nunits_for_cost): New function.
13116         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13117         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13118         (vect_analyze_slp_cost): Likewise.
13119         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13120         (vect_model_load_cost): Likewise.
13121
13122 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13123             Alan Hayward  <alan.hayward@arm.com>
13124             David Sherwood  <david.sherwood@arm.com>
13125
13126         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13127         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13128         from an unsigned int * to a poly_uint64_pod *.
13129         (calculate_unrolling_factor): New function.
13130         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
13131
13132 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13133             Alan Hayward  <alan.hayward@arm.com>
13134             David Sherwood  <david.sherwood@arm.com>
13135
13136         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13137         from an unsigned int to a poly_uint64.
13138         (_loop_vec_info::slp_unrolling_factor): Likewise.
13139         (_loop_vec_info::vectorization_factor): Change from an int
13140         to a poly_uint64.
13141         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13142         (vect_get_num_vectors): New function.
13143         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13144         (vect_get_num_copies): Use vect_get_num_vectors.
13145         (vect_analyze_data_ref_dependences): Change max_vf from an int *
13146         to an unsigned int *.
13147         (vect_analyze_data_refs): Change min_vf from an int * to a
13148         poly_uint64 *.
13149         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13150         than an unsigned HOST_WIDE_INT.
13151         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13152         (vect_analyze_data_ref_dependence): Change max_vf from an int *
13153         to an unsigned int *.
13154         (vect_analyze_data_ref_dependences): Likewise.
13155         (vect_compute_data_ref_alignment): Handle polynomial vf.
13156         (vect_enhance_data_refs_alignment): Likewise.
13157         (vect_prune_runtime_alias_test_list): Likewise.
13158         (vect_shift_permute_load_chain): Likewise.
13159         (vect_supportable_dr_alignment): Likewise.
13160         (dependence_distance_ge_vf): Take the vectorization factor as a
13161         poly_uint64 rather than an unsigned HOST_WIDE_INT.
13162         (vect_analyze_data_refs): Change min_vf from an int * to a
13163         poly_uint64 *.
13164         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13165         vfm1 as a poly_uint64 rather than an int.  Make the same change
13166         for the returned bound_scalar.
13167         (vect_gen_vector_loop_niters): Handle polynomial vf.
13168         (vect_do_peeling): Likewise.  Update call to
13169         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13170         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13171         be constant.
13172         * tree-vect-loop.c (vect_determine_vectorization_factor)
13173         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13174         (vect_get_known_peeling_cost): Likewise.
13175         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13176         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13177         (vect_transform_loop): Likewise.  Use the lowest possible VF when
13178         updating the upper bounds of the loop.
13179         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
13180         rather than an int.
13181         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13182         polynomial unroll factors.
13183         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13184         (vect_make_slp_decision): Likewise.
13185         (vect_supported_load_permutation_p): Likewise, and polynomial
13186         vf too.
13187         (vect_analyze_slp_cost): Handle polynomial vf.
13188         (vect_slp_analyze_node_operations): Likewise.
13189         (vect_slp_analyze_bb_1): Likewise.
13190         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13191         than an unsigned HOST_WIDE_INT.
13192         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13193         (vectorizable_load): Handle polynomial vf.
13194         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13195         a poly_uint64.
13196         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13197
13198 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13199             Alan Hayward  <alan.hayward@arm.com>
13200             David Sherwood  <david.sherwood@arm.com>
13201
13202         * match.pd: Handle bit operations involving three constants
13203         and try to fold one pair.
13204
13205 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13206
13207         * tree-vect-loop-manip.c: Include gimple-fold.h.
13208         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13209         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
13210         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13211         Add a path that uses a step of VF instead of 1, but disable it
13212         for now.
13213         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13214         and niters_no_overflow parameters.  Update calls to
13215         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13216         Create a new SSA name if the latter choses to use a ste other
13217         than zero, and return it via niters_vector_mult_vf_var.
13218         * tree-vect-loop.c (vect_transform_loop): Update calls to
13219         vect_do_peeling, vect_gen_vector_loop_niters and
13220         slpeel_make_loop_iterate_ntimes.
13221         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13222         (vect_gen_vector_loop_niters): Update declarations after above changes.
13223
13224 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
13225
13226         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13227         128-bit round to integer instructions.
13228         (ceil<mode>2): Likewise.
13229         (btrunc<mode>2): Likewise.
13230         (round<mode>2): Likewise.
13231
13232 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
13233
13234         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13235         unaligned VSX load/store on P8/P9.
13236         (expand_block_clear): Allow the use of unaligned VSX
13237         load/store on P8/P9.
13238
13239 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13240
13241         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13242         New function.
13243         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13244         swap associated with both a load and a store.
13245
13246 2018-01-02  Andrew Waterman  <andrew@sifive.com>
13247
13248         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13249         * config/riscv/riscv.md (clear_cache): Use it.
13250
13251 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
13252
13253         * web.c: Remove out-of-date comment.
13254
13255 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
13256
13257         * expr.c (fixup_args_size_notes): Check that any existing
13258         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13259         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13260         (emit_single_push_insn): ...here.
13261
13262 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
13263
13264         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13265         (const_vector_encoded_nelts): New function.
13266         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13267         (const_vector_int_elt, const_vector_elt): Declare.
13268         * emit-rtl.c (const_vector_int_elt_1): New function.
13269         (const_vector_elt): Likewise.
13270         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13271         of CONST_VECTOR_ELT.
13272
13273 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
13274
13275         * expr.c: Include rtx-vector-builder.h.
13276         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13277         directly on the tree encoding.
13278         (const_vector_from_tree): Likewise.
13279         * optabs.c: Include rtx-vector-builder.h.
13280         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13281         sequence of "u" values.
13282         * vec-perm-indices.c: Include rtx-vector-builder.h.
13283         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13284         directly on the vec_perm_indices encoding.
13285
13286 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
13287
13288         * doc/rtl.texi (const_vector): Describe new encoding scheme.
13289         * Makefile.in (OBJS): Add rtx-vector-builder.o.
13290         * rtx-vector-builder.h: New file.
13291         * rtx-vector-builder.c: Likewise.
13292         * rtl.h (rtx_def::u2): Add a const_vector field.
13293         (CONST_VECTOR_NPATTERNS): New macro.
13294         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13295         (CONST_VECTOR_DUPLICATE_P): Likewise.
13296         (CONST_VECTOR_STEPPED_P): Likewise.
13297         (CONST_VECTOR_ENCODED_ELT): Likewise.
13298         (const_vec_duplicate_p): Check for a duplicated vector encoding.
13299         (unwrap_const_vec_duplicate): Likewise.
13300         (const_vec_series_p): Check for a non-duplicated vector encoding.
13301         Say that the function only returns true for integer vectors.
13302         * emit-rtl.c: Include rtx-vector-builder.h.
13303         (gen_const_vec_duplicate_1): Delete.
13304         (gen_const_vector): Call gen_const_vec_duplicate instead of
13305         gen_const_vec_duplicate_1.
13306         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13307         (gen_const_vec_duplicate): Use rtx_vector_builder.
13308         (gen_const_vec_series): Likewise.
13309         (gen_rtx_CONST_VECTOR): Likewise.
13310         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13311         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13312         Build a new vector rather than modifying a CONST_VECTOR in-place.
13313         (handle_special_swappables): Update call accordingly.
13314         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13315         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13316         Build a new vector rather than modifying a CONST_VECTOR in-place.
13317         (handle_special_swappables): Update call accordingly.
13318
13319 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
13320
13321         * simplify-rtx.c (simplify_const_binary_operation): Use
13322         CONST_VECTOR_ELT instead of XVECEXP.
13323
13324 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
13325
13326         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13327         the selector elements to be different from the data elements
13328         if the selector is a VECTOR_CST.
13329         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13330         ssizetype for the selector.
13331
13332 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
13333
13334         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13335         before testing each element individually.
13336         * tree-vect-generic.c (lower_vec_perm): Likewise.
13337
13338 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
13339
13340         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13341         * selftest-run-tests.c (selftest::run_tests): Call it.
13342         * vector-builder.h (vector_builder::operator ==): New function.
13343         (vector_builder::operator !=): Likewise.
13344         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13345         (vec_perm_indices::all_from_input_p): New function.
13346         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13347         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13348         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13349         instead of reading the VECTOR_CST directly.  Detect whether both
13350         vector inputs are the same before constructing the vec_perm_indices,
13351         and update the number of inputs argument accordingly.  Use the
13352         utility functions added above.  Only construct sel2 if we need to.
13353
13354 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
13355
13356         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13357         the broadcast of the low byte.
13358         (expand_mult_highpart): Use an explicit encoding for the permutes.
13359         * optabs-query.c (can_mult_highpart_p): Likewise.
13360         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13361         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13362         (vectorizable_bswap): Likewise.
13363         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13364         explicit encoding for the power-of-2 permutes.
13365         (vect_permute_store_chain): Likewise.
13366         (vect_grouped_load_supported): Likewise.
13367         (vect_permute_load_chain): Likewise.
13368
13369 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
13370
13371         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13372         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13373         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13374         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13375         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13376         (vect_gen_perm_mask_any): Likewise.
13377
13378 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
13379
13380         * int-vector-builder.h: New file.
13381         * vec-perm-indices.h: Include int-vector-builder.h.
13382         (vec_perm_indices): Redefine as an int_vector_builder.
13383         (auto_vec_perm_indices): Delete.
13384         (vec_perm_builder): Redefine as a stand-alone class.
13385         (vec_perm_indices::vec_perm_indices): New function.
13386         (vec_perm_indices::clamp): Likewise.
13387         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13388         (vec_perm_indices::new_vector): New function.
13389         (vec_perm_indices::new_expanded_vector): Update for new
13390         vec_perm_indices class.
13391         (vec_perm_indices::rotate_inputs): New function.
13392         (vec_perm_indices::all_in_range_p): Operate directly on the
13393         encoded form, without computing elided elements.
13394         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13395         encoding.  Update for new vec_perm_indices class.
13396         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13397         the given vec_perm_builder.
13398         (expand_vec_perm_var): Update vec_perm_builder constructor.
13399         (expand_mult_highpart): Use vec_perm_builder instead of
13400         auto_vec_perm_indices.
13401         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13402         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
13403         or double series encoding as appropriate.
13404         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13405         vec_perm_indices instead of auto_vec_perm_indices.
13406         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13407         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13408         (vect_permute_store_chain): Likewise.
13409         (vect_grouped_load_supported): Likewise.
13410         (vect_permute_load_chain): Likewise.
13411         (vect_shift_permute_load_chain): Likewise.
13412         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13413         (vect_transform_slp_perm_load): Likewise.
13414         (vect_schedule_slp_instance): Likewise.
13415         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13416         (vectorizable_mask_load_store): Likewise.
13417         (vectorizable_bswap): Likewise.
13418         (vectorizable_store): Likewise.
13419         (vectorizable_load): Likewise.
13420         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13421         vec_perm_indices instead of auto_vec_perm_indices.  Use
13422         tree_to_vec_perm_builder to read the vector from a tree.
13423         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13424         vec_perm_builder instead of a vec_perm_indices.
13425         (have_whole_vector_shift): Use vec_perm_builder and
13426         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
13427         truncation to calc_vec_perm_mask_for_shift.
13428         (vect_create_epilog_for_reduction): Likewise.
13429         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13430         from auto_vec_perm_indices to vec_perm_indices.
13431         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13432         instead of changing individual elements.
13433         (aarch64_vectorize_vec_perm_const): Use new_vector to install
13434         the vector in d.perm.
13435         * config/arm/arm.c (expand_vec_perm_d::perm): Change
13436         from auto_vec_perm_indices to vec_perm_indices.
13437         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13438         instead of changing individual elements.
13439         (arm_vectorize_vec_perm_const): Use new_vector to install
13440         the vector in d.perm.
13441         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13442         Update vec_perm_builder constructor.
13443         (rs6000_expand_interleave): Likewise.
13444         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13445         (rs6000_expand_interleave): Likewise.
13446
13447 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
13448
13449         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13450         to qimode could truncate the indices.
13451         * optabs.c (expand_vec_perm_var): Likewise.
13452
13453 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
13454
13455         * Makefile.in (OBJS): Add vec-perm-indices.o.
13456         * vec-perm-indices.h: New file.
13457         * vec-perm-indices.c: Likewise.
13458         * target.h (vec_perm_indices): Replace with a forward class
13459         declaration.
13460         (auto_vec_perm_indices): Move to vec-perm-indices.h.
13461         * optabs.h: Include vec-perm-indices.h.
13462         (expand_vec_perm): Delete.
13463         (selector_fits_mode_p, expand_vec_perm_var): Declare.
13464         (expand_vec_perm_const): Declare.
13465         * target.def (vec_perm_const_ok): Replace with...
13466         (vec_perm_const): ...this new hook.
13467         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13468         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13469         * doc/tm.texi: Regenerate.
13470         * optabs.def (vec_perm_const): Delete.
13471         * doc/md.texi (vec_perm_const): Likewise.
13472         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13473         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13474         expand_vec_perm for constant permutation vectors.  Assert that
13475         the mode of variable permutation vectors is the integer equivalent
13476         of the mode that is being permuted.
13477         * optabs-query.h (selector_fits_mode_p): Declare.
13478         * optabs-query.c: Include vec-perm-indices.h.
13479         (selector_fits_mode_p): New function.
13480         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13481         is defined, instead of checking whether the vec_perm_const_optab
13482         exists.  Use targetm.vectorize.vec_perm_const instead of
13483         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
13484         fit in the vector mode before using a variable permute.
13485         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13486         vec_perm_indices instead of an rtx.
13487         (expand_vec_perm): Replace with...
13488         (expand_vec_perm_const): ...this new function.  Take the selector
13489         as a vec_perm_indices rather than an rtx.  Also take the mode of
13490         the selector.  Update call to shift_amt_for_vec_perm_mask.
13491         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13492         Use vec_perm_indices::new_expanded_vector to expand the original
13493         selector into bytes.  Check whether the indices fit in the vector
13494         mode before using a variable permute.
13495         (expand_vec_perm_var): Make global.
13496         (expand_mult_highpart): Use expand_vec_perm_const.
13497         * fold-const.c: Includes vec-perm-indices.h.
13498         * tree-ssa-forwprop.c: Likewise.
13499         * tree-vect-data-refs.c: Likewise.
13500         * tree-vect-generic.c: Likewise.
13501         * tree-vect-loop.c: Likewise.
13502         * tree-vect-slp.c: Likewise.
13503         * tree-vect-stmts.c: Likewise.
13504         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13505         Delete.
13506         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13507         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13508         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13509         (aarch64_vectorize_vec_perm_const): ...this new function.
13510         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13511         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13512         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13513         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13514         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13515         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13516         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13517         into...
13518         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
13519         check for NEON modes.
13520         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13521         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13522         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13523         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13524         into...
13525         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
13526         the old VEC_PERM_CONST conditions.
13527         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13528         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13529         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13530         (ia64_vectorize_vec_perm_const_ok): Merge into...
13531         (ia64_vectorize_vec_perm_const): ...this new function.
13532         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13533         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13534         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13535         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13536         * config/mips/mips.c (mips_expand_vec_perm_const)
13537         (mips_vectorize_vec_perm_const_ok): Merge into...
13538         (mips_vectorize_vec_perm_const): ...this new function.
13539         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13540         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13541         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13542         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13543         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13544         (rs6000_expand_vec_perm_const): Delete.
13545         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13546         Delete.
13547         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13548         (altivec_expand_vec_perm_const_le): Take each operand individually.
13549         Operate on constant selectors rather than rtxes.
13550         (altivec_expand_vec_perm_const): Likewise.  Update call to
13551         altivec_expand_vec_perm_const_le.
13552         (rs6000_expand_vec_perm_const): Delete.
13553         (rs6000_vectorize_vec_perm_const_ok): Delete.
13554         (rs6000_vectorize_vec_perm_const): New function.
13555         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13556         an element count and rtx array.
13557         (rs6000_expand_extract_even): Update call accordingly.
13558         (rs6000_expand_interleave): Likewise.
13559         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13560         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13561         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13562         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13563         (rs6000_expand_vec_perm_const): Delete.
13564         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13565         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13566         (altivec_expand_vec_perm_const_le): Take each operand individually.
13567         Operate on constant selectors rather than rtxes.
13568         (altivec_expand_vec_perm_const): Likewise.  Update call to
13569         altivec_expand_vec_perm_const_le.
13570         (rs6000_expand_vec_perm_const): Delete.
13571         (rs6000_vectorize_vec_perm_const_ok): Delete.
13572         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
13573         reference to the SPE evmerge intructions.
13574         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13575         an element count and rtx array.
13576         (rs6000_expand_extract_even): Update call accordingly.
13577         (rs6000_expand_interleave): Likewise.
13578         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13579         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13580         new function.
13581         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13582
13583 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
13584
13585         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13586         vector mode and that that mode matches the mode of the data
13587         being permuted.
13588         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13589         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
13590         directly using expand_vec_perm_1 when forcing selectors into
13591         registers.
13592         (expand_vec_perm_var): New function, split out from expand_vec_perm.
13593
13594 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
13595
13596         * optabs-query.h (can_vec_perm_p): Delete.
13597         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13598         * optabs-query.c (can_vec_perm_p): Split into...
13599         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13600         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13601         particular selector is valid.
13602         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13603         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13604         (vect_grouped_load_supported): Likewise.
13605         (vect_shift_permute_load_chain): Likewise.
13606         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13607         (vect_transform_slp_perm_load): Likewise.
13608         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13609         (vectorizable_bswap): Likewise.
13610         (vect_gen_perm_mask_checked): Likewise.
13611         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
13612         implementations of variable permutation vectors into account
13613         when deciding which selector to use.
13614         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13615         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13616         with a false third argument.
13617         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13618         to test whether the constant selector is valid and can_vec_perm_var_p
13619         to test whether a variable selector is valid.
13620
13621 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
13622
13623         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13624         * optabs-query.c (can_vec_perm_p): Likewise.
13625         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13626         instead of vec_perm_indices.
13627         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13628         (vect_gen_perm_mask_checked): Likewise,
13629         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13630         (vect_gen_perm_mask_checked): Likewise,
13631
13632 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
13633
13634         * optabs-query.h (qimode_for_vec_perm): Declare.
13635         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13636         (qimode_for_vec_perm): ...this new function.
13637         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13638
13639 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
13640
13641         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13642         does not have a conditional at the top.
13643
13644 2018-01-02  Richard Biener  <rguenther@suse.de>
13645
13646         * ipa-inline.c (big_speedup_p): Fix expression.
13647
13648 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
13649
13650         PR target/81616
13651         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13652         for generic 4->6.
13653
13654 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
13655
13656         PR target/81616
13657         Generic tuning.
13658         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13659         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13660         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13661         cond_taken_branch_cost 3->4.
13662
13663 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
13664
13665         PR tree-optimization/83581
13666         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13667         TODO_cleanup_cfg if any changes have been made.
13668
13669         PR middle-end/83608
13670         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13671         convert_modes if target mode has the right side, but different mode
13672         class.
13673
13674         PR middle-end/83609
13675         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13676         last argument when extracting from CONCAT.  If either from_real or
13677         from_imag is NULL, use expansion through memory.  If result is not
13678         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13679         the parts directly to inner mode, if even that fails, use expansion
13680         through memory.
13681
13682         PR middle-end/83623
13683         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13684         check for bswap in mode rather than HImode and use that in expand_unop
13685         too.
13686 \f
13687 Copyright (C) 2018 Free Software Foundation, Inc.
13688
13689 Copying and distribution of this file, with or without modification,
13690 are permitted in any medium without royalty provided the copyright
13691 notice and this notice are preserved.