gcc/ChangeLog

   1 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
   2
   3         PR target/63177
   4         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
   5         Don't handle -mcpu=power8 if -mpower9-vector is also used.
   6
   7 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
   8
   9         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
  10         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
  11         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
  12         several redundant entries.
  13
  14 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
  15
  16         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
  17         type from "rtx" to "rtx_insn *".
  18         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
  19         for local "call_insn", removing cast.
  20         (ix86_expand_call): Likewise, introducing a "call_insn" local.
  21
  22 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
  23
  24         PR tree-optimization/86066
  25         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
  26         for BIT_INSERT_EXPR stores.
  27
  28 2018-06-06  Richard Biener  <rguenther@suse.de>
  29
  30         PR tree-optimization/86062
  31         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
  32         component refs ontop
  33         of to be offsetted base.
  34
  35 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
  36
  37         * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
  38         to be static and remove check on interrupt attribute name.
  39
  40         gcc/testsuite/gcc.target/msp430/
  41         * function-attributes-4.c: New test.
  42         * static-interrupts.c: New test.
  43
  44 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
  45
  46         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
  47         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
  48
  49 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
  50
  51         PR target/79924
  52         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
  53         second argument.
  54         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
  55         Remove second argument, change how error is called.
  56         (aarch64_layout_arg): Remove second argument from
  57         aarch64_err_no_fpadvsimd call.
  58         (aarch64_init_cumulative_args): Ditto.
  59         (aarch64_gimplify_va_arg_expr): Ditto.
  60         * config/aarch64/aarch64.md (mov<mode>): Ditto.
  61
  62 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
  63
  64         * config/i386/i386.md (simple_return_indirect_internal): New expander.
  65         (*simple_return_indirect_internal<mode>): Rename from
  66         simple_return_indirect_internal.  Use W mode iterator.
  67         (rstorssp): New expander.
  68         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
  69         (clrssbsy): New expander.
  70         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
  71
  72 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
  73
  74         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
  75         __typeof__.
  76         (cmse_check_pointed_object): Likewise.
  77
  78 2018-06-05  Martin Liska  <mliska@suse.cz>
  79
  80         PR gcov-profile/47618
  81         * doc/invoke.texi: Document how -fprofile-dir format
  82         is extended.
  83
  84 2018-06-05  Richard Biener  <rguenther@suse.de>
  85
  86         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
  87         removal pretend DOM info isn't available so we do not update
  88         it and only remove edges, not dominated blocks.  Actually free
  89         DOM info in case we removed something.  Remove unreachable blocks.
  90         (mfb_keep_latches): Work with either DOM info or marked backedges.
  91         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
  92         first.  Mark backedges if DOM info isn't available.
  93         (Re-)compute DOM info after cleanup_control_flow_pre.
  94
  95 2018-06-05  Richard Biener  <rguenther@suse.de>
  96
  97         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
  98         (locus_discrim_hasher::hash): Adjust.
  99         (locus_discrim_hasher::equal): Likewise.
 100         (next_discriminator_for_locus): Work on line directly.
 101         (same_line_p): Pass in expanded locus1 as well.
 102         (assign_discriminators): Avoid redundant location expansions.
 103
 104 2018-06-05  Richard Biener  <rguenther@suse.de>
 105
 106         PR tree-optimization/86046
 107         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
 108         if required after clearing TREE_ADDRESSABLE.
 109
 110 2018-06-05  Richard Biener  <rguenther@suse.de>
 111
 112         PR tree-optimization/86047
 113         * tree-ssa-loop.c (for_each_index): Glob handling of all
 114         decls and constants and really handle all of them.
 115
 116 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 117
 118         PR target/81497
 119         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
 120         qualifier_void_pointer and qualifier_const_void_pointer.
 121         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
 122         (arm_init_builtins): Handle the above.
 123         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
 124         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
 125         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
 126         void intrinsics.
 127
 128 2018-06-05  Martin Liska  <mliska@suse.cz>
 129
 130         * auto-profile.c (read_autofdo_file): Do not use
 131         gcov_ctr_summary struct.
 132         (afdo_callsite_hot_enough_for_early_inline): Likewise.
 133         * coverage.c (struct counts_entry): Likewise.
 134         (read_counts_file): Read just single summary entry.
 135         (get_coverage_counts): Use gcov_summary struct.
 136         * coverage.h (get_coverage_counts): Likewise.
 137         * gcov-dump.c (dump_working_sets): Likewise.
 138         (tag_summary): Dump just single summary.
 139         * gcov-io.c (gcov_write_summary): Write just histogram
 140         summary.
 141         (gcov_read_summary): Read just single summary.
 142         (compute_working_sets): Use gcov_summary struct.
 143         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
 144         of GCOV_COUNTERS_SUMMABLE.
 145         (GCOV_COUNTERS_SUMMABLE): Remove.
 146         (GCOV_FIRST_VALUE_COUNTER): Replace with
 147         GCOV_COUNTER_V_INTERVAL.
 148         (struct gcov_ctr_summary): Remove.
 149         (struct gcov_summary): Directly use fields of former
 150         gcov_ctr_summary.
 151         (compute_working_sets): Use gcov_summary struct.
 152         * gcov.c (read_count_file): Do not use ctrs fields.
 153         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
 154         struct.
 155         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
 156         struct.
 157         * profile.c: Likewise.
 158         * profile.h: Likewise.
 159
 160 2018-06-05  Martin Liska  <mliska@suse.cz>
 161
 162         PR gcov-profile/84846
 163         * gcov.c (output_lines): Print working directory only
 164         in intermediate format.
 165
 166 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
 167
 168         * config/s390/s390-builtin-types.def: Add void function type.
 169         * config/s390/s390-builtins.def: Use the function type for the
 170         tbeginc builtin.
 171
 172 2018-06-04  Jim Wilson  <jimw@sifive.com>
 173
 174         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
 175         to int.
 176         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
 177         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
 178         handle EH_RETURN_DATA_REGNO registers properly.
 179         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
 180         (riscv_expand_epilogue): Update comment.  Change argument name and
 181         type.  Update code to use new name and type.  Pass new args to
 182         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
 183         EXCEPTION_RETURN.
 184         * config/riscv/riscv.md (NORMAL_RETURN): New.
 185         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
 186         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
 187         (eh_return): Call gen_eh_return_internal and emit barrier.
 188         (eh_return_internal): Call riscv_expand_epilogue.
 189
 190 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
 191
 192         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
 193         bit_insertion field and declare can_be_merged_into method.
 194         (merged_store_group::can_be_merged_into): New method.
 195         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
 196         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
 197         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
 198
 199 2018-06-04  Richard Biener  <rguenther@suse.de>
 200
 201         PR tree-optimization/85955
 202         * builtins.c (fold_builtin_sincos): Convert pointers to
 203         destination to appropriate type before dereferencing.
 204
 205 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
 206
 207         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
 208
 209 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
 210
 211         * expr.c (expand_expr_real_1): Force the operand into memory if
 212         its TYPE_MODE is BLKmode and if there is no integer mode for
 213         the number of bits being extracted.
 214
 215 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
 216
 217         PR target/85832
 218         PR target/86036
 219         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
 220         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
 221
 222 2018-06-04  Richard Biener  <rguenther@suse.de>
 223
 224         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
 225         (cleanup_tree_cfg_noloop): ... single caller.  Do
 226         start_recording_case_labels later.
 227
 228 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
 229
 230         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
 231         to _IMMINTRIN_H_INCLUDED.
 232         * config/i386/pconfigintrin.h: Ditto.
 233         * config/i386/waitpkgintrin.h: Ditto.
 234         * config/i386/immintrin.h: Add includes for sgxintrin.h,
 235         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
 236         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
 237         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
 238         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
 239         waitpkgintrin.h and cldemoteintrin.h.
 240
 241 2018-06-04  Richard Biener  <rguenther@suse.de>
 242
 243         PR tree-optimization/86038
 244         * tracer.c (find_best_successor): Check probability for
 245         being initialized, bail out if not.
 246
 247 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
 248
 249         PR target/86003
 250         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
 251         of bits to ignore when comparing architectures.
 252
 253 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
 254
 255         PR tree-optimization/69615
 256         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
 257         maximum or minimum of the type, try to merge it also as if
 258         range1 is + [-, x - 1] or + [x + 1, -].
 259
 260         PR c++/86025
 261         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
 262
 263 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
 264
 265         PR tree-optimization/86034
 266         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
 267         the unsigned bitfield type in a bit insertion sequence if it does not
 268         have a larger precision than the bitfield size.
 269         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
 270
 271 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
 272
 273         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
 274
 275 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
 276
 277         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
 278         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
 279         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
 280         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
 281
 282 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
 283
 284         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
 285         Disable -fdelete-null-pointer-checks for ELF toolchain.
 286
 287 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
 288             Kito Cheng  <kito.cheng@gmail.com>
 289
 290         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
 291         (nds32le-*-*, nds32be-*-*): Integrate checking process.
 292         (nds32*-*-*): Add glibc and uclibc conditions.
 293         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
 294         (TARGET_EXCEPT_UNWIND_INFO): Define.
 295         * config/nds32/elf.h: New file.
 296         * config/nds32/linux.h: New file.
 297         * config/nds32/nds32-elf.opt: New file.
 298         * config/nds32/nds32-linux.opt: New file.
 299         * config/nds32/nds32-fp-as-gp.c
 300         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
 301         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
 302         TARGET_LINUX_ABI.
 303         (nds32_asm_file_end): Ditto.
 304         (nds32_print_operand): Ditto.
 305         (nds32_insert_attributes): Ditto.
 306         (nds32_init_libfuncs): New function.
 307         (TARGET_HAVE_TLS): Define.
 308         (TARGET_INIT_LIBFUNCS): Define.
 309         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
 310         spec content.
 311         (TARGET_ELF): Apply different mcmodel setting.
 312         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
 313         been migrated into elf.h and linux.h files.
 314         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
 315         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
 316         (mcmodel): The content has been migrated into nds32-elf.opt and
 317         nds32-linux.opt files.
 318         * config/nds32/t-elf: New file.
 319         * config/nds32/t-linux: New file.
 320
 321 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
 322             Shiva Chen  <shiva0217@gmail.com>
 323
 324         * config/nds32/constants.md (unspec_volatile_element): Add
 325         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
 326         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
 327         optimization.
 328         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
 329         (make_pass_nds32_fp_as_gp): Declare.
 330         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
 331         optmization pass.
 332         (nds32_asm_function_end_prologue): Remove unused asm output.
 333         (nds32_asm_function_begin_epilogue): Remove unused asm output.
 334         (nds32_asm_file_start): Output necessary fp_as_gp information.
 335         (nds32_option_override): Adjust register usage.
 336         (nds32_expand_prologue): Consider fp_as_gp situation.
 337         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
 338         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
 339         (epilogue): Ditto.
 340         (return): Ditto.
 341         (simple_return): Ditto.
 342         (omit_fp_begin): Output special directive for fp_as_gp.
 343         (omit_fp_end): Output special directive for fp_as_gp.
 344         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
 345         mforbid-fp-as-gp): New options.
 346
 347 2018-06-01  Mark Wielaard  <mark@klomp.org>
 348
 349         * dwarf2out.c (dwarf2out_finish): Remove generation of
 350         DW_AT_loclists_base.
 351
 352 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
 353
 354         * gimple-ssa-store-merging.c: Include gimple-fold.h.
 355         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
 356         (struct merged_store_group): Add bit_insertion field.
 357         (dump_char_array): Use standard hexadecimal format.
 358         (merged_store_group::merged_store_group): Set bit_insertion to false.
 359         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
 360         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
 361         also print the mask in the dump file.
 362         (pass_store_merging::gate): Minor tweak.
 363         (imm_store_chain_info::coalesce_immediate): Fix wrong association
 364         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
 365         stores with INTEGER_CST stores.
 366         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
 367         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
 368         and use it throughout.  Generate bit insertion sequences if need be.
 369         (pass_store_merging::process_store): Remove redundant condition.
 370         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
 371
 372 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
 373
 374         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
 375         the 128-bit floating point types.  Fix function comment.
 376
 377 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 378
 379         * config/aarch64/aarch64-simd.md
 380         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
 381         mnemonics.
 382         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
 383         mnemonics.
 384
 385 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
 386
 387         PR tree-optimization/85989
 388         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
 389         variable.
 390         (backprop::intersect_uses): Check it when deciding whether this
 391         is a backedge reference.
 392         (backprop::process_block): Add each phi to m_visited_phis
 393         after visiting it, then clear it at the end.
 394
 395 2018-06-01  Richard Biener  <rguenther@suse.de>
 396
 397         * tree-vectorizer.h (vect_dr_stmt): New function.
 398         (vect_get_load_cost): Adjust.
 399         (vect_get_store_cost): Likewise.
 400         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
 401         Use vect_dr_stmt instead of DR_SMTT.
 402         (vect_record_base_alignments): Likewise.
 403         (vect_calculate_target_alignment): Likewise.
 404         (vect_compute_data_ref_alignment): Likewise and make static.
 405         (vect_update_misalignment_for_peel): Likewise.
 406         (vect_verify_datarefs_alignment): Likewise.
 407         (vector_alignment_reachable_p): Likewise.
 408         (vect_get_data_access_cost): Likewise.  Pass down
 409         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
 410         (vect_get_peeling_costs_all_drs): Likewise.
 411         (vect_peeling_hash_get_lowest_cost): Likewise.
 412         (vect_enhance_data_refs_alignment): Likewise.
 413         (vect_find_same_alignment_drs): Likewise.
 414         (vect_analyze_data_refs_alignment): Likewise.
 415         (vect_analyze_group_access_1): Likewise.
 416         (vect_analyze_group_access): Likewise.
 417         (vect_analyze_data_ref_access): Likewise.
 418         (vect_analyze_data_ref_accesses): Likewise.
 419         (vect_vfa_segment_size): Likewise.
 420         (vect_small_gap_p): Likewise.
 421         (vectorizable_with_step_bound_p): Likewise.
 422         (vect_prune_runtime_alias_test_list): Likewise.
 423         (vect_analyze_data_refs): Likewise.
 424         (vect_supportable_dr_alignment): Likewise.
 425         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
 426         (vect_gen_prolog_loop_niters): Likewise.
 427         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
 428         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
 429         modify DR_STMT.
 430         (vect_recog_mask_conversion_pattern): Likewise.
 431         (vect_try_gather_scatter_pattern): Likewise.
 432         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
 433         to vect_get_store_cost.
 434         (vect_get_store_cost): Get stmt_info instead of DR.
 435         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
 436         (vect_get_load_cost): Get stmt_info instead of DR.
 437
 438 2018-06-01  Richard Biener  <rguenther@suse.de>
 439
 440         PR middle-end/86017
 441         * gimple-fold.c (var_decl_component_p): Also allow offsetted
 442         vars wrapped in MEM_REFs.
 443
 444 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
 445
 446         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
 447         Fix subreg tests so that we only return a choice between
 448         GENERAL_REGS and FP_REGS if the original classes included both.
 449
 450 2018-06-01  Richard Biener  <rguenther@suse.de>
 451
 452         PR ipa/85960
 453         * tree-ssa-structalias.c (get_function_part_constraint):
 454         Handle NULL fi->decl.
 455         (find_func_aliases_for_call): Properly handle indirect
 456         fi from direct call.
 457         (find_func_clobbers): Likewise.
 458         (ipa_pta_execute): Likewise.
 459         (create_variable_info_for): For functions that are ifunc_resolver
 460         resolve to a varinfo that contains the result of the resolver call.
 461         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
 462         aliases.
 463
 464 2018-05-31  Michael Collison  <michael.collison@arm.com>
 465
 466         * config/aarch64/aarch64.md:
 467         (*fix_to_zero_extenddfdi2): New pattern.
 468         * gcc.target/aarch64/fix_extend1.c: New testcase.
 469
 470 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
 471
 472         PR middle-end/78809
 473         PR middle-end/83026
 474         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
 475         and BUILT_IN_STRNCMP_EQ.
 476         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
 477         BUILT_IN_STRNCMP_EQ.
 478         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
 479         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
 480         (gimple_fold_builtin): Likewise.
 481         * tree-ssa-strlen.c (compute_string_length): New function.
 482         (determine_min_obsize): New function.
 483         (handle_builtin_string_cmp): New function to handle calls to
 484         string compare functions.
 485         (strlen_optimize_stmt): Add handling to builtin string compare
 486         calls.
 487         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
 488         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
 489         * tree.c (build_common_builtin_nodes): Add new defines of
 490         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
 491
 492 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
 493
 494         PR target/85984
 495         * bb-reorder.c (pass_partition_blocks::gate): Return false for
 496         functions with naked attribute.
 497
 498 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
 499
 500         * config/i386/sse.md (avx_vec_concat<mode>):
 501         Substitute concat_tg_mode mode attribute with xtg_mode.
 502         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
 503         (concat_tg_mode): Remove mode attribute.
 504
 505 2018-05-31  Martin Sebor  <msebor@redhat.com>
 506
 507         PR c/82063
 508         * calls.c (alloc_max_size): Correct a logic error/typo.
 509         Treat excessive arguments as infinite.  Warn for invalid arguments.
 510         * doc/invoke.texi (-Walloc-size-larger-than): Update.
 511
 512 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
 513
 514         PR target/85829
 515         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
 516         and movx for Haswell.
 517
 518 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
 519             Cesar Philippidis  <cesar@codesourcery.com>
 520
 521         PR middle-end/85879
 522         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
 523         when emitting error on private/firstprivate reductions.
 524         * omp-low.c (lower_omp_target): Avoid reference-type processing
 525         on pointers for firstprivate clause.
 526
 527 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
 528
 529         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
 530         (st1x2): Likewise.
 531         (st1x3): Likewise.
 532         * config/aarch64/aarch64-simd.md
 533         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
 534         (aarch64_ld1_x3_<mode>): Likewise
 535         (aarch64_st1x2<VALLDIF:mode>): Likewise
 536         (aarch64_st1_x2_<mode>): Likewise
 537         (aarch64_st1x3<VALLDIF:mode>): Likewise
 538         (aarch64_st1_x3_<mode>): Likewise
 539         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
 540         (vld1_s8_x3): Likewise.
 541         (vld1_u16_x3): Likewise.
 542         (vld1_s16_x3): Likewise.
 543         (vld1_u32_x3): Likewise.
 544         (vld1_s32_x3): Likewise.
 545         (vld1_u64_x3): Likewise.
 546         (vld1_s64_x3): Likewise.
 547         (vld1_f16_x3): Likewise.
 548         (vld1_f32_x3): Likewise.
 549         (vld1_f64_x3): Likewise.
 550         (vld1_p8_x3): Likewise.
 551         (vld1_p16_x3): Likewise.
 552         (vld1_p64_x3): Likewise.
 553         (vld1q_u8_x3): Likewise.
 554         (vld1q_s8_x3): Likewise.
 555         (vld1q_u16_x3): Likewise.
 556         (vld1q_s16_x3): Likewise.
 557         (vld1q_u32_x3): Likewise.
 558         (vld1q_s32_x3): Likewise.
 559         (vld1q_u64_x3): Likewise.
 560         (vld1q_s64_x3): Likewise.
 561         (vld1q_f16_x3): Likewise.
 562         (vld1q_f32_x3): Likewise.
 563         (vld1q_f64_x3): Likewise.
 564         (vld1q_p8_x3): Likewise.
 565         (vld1q_p16_x3): Likewise.
 566         (vld1q_p64_x3): Likewise.
 567         (vst1_s64_x2): Likewise.
 568         (vst1_u64_x2): Likewise.
 569         (vst1_f64_x2): Likewise.
 570         (vst1_s8_x2): Likewise.
 571         (vst1_p8_x2): Likewise.
 572         (vst1_s16_x2): Likewise.
 573         (vst1_p16_x2): Likewise.
 574         (vst1_s32_x2): Likewise.
 575         (vst1_u8_x2): Likewise.
 576         (vst1_u16_x2): Likewise.
 577         (vst1_u32_x2): Likewise.
 578         (vst1_f16_x2): Likewise.
 579         (vst1_f32_x2): Likewise.
 580         (vst1_p64_x2): Likewise.
 581         (vst1q_s8_x2): Likewise.
 582         (vst1q_p8_x2): Likewise.
 583         (vst1q_s16_x2): Likewise.
 584         (vst1q_p16_x2): Likewise.
 585         (vst1q_s32_x2): Likewise.
 586         (vst1q_s64_x2): Likewise.
 587         (vst1q_u8_x2): Likewise.
 588         (vst1q_u16_x2): Likewise.
 589         (vst1q_u32_x2): Likewise.
 590         (vst1q_u64_x2): Likewise.
 591         (vst1q_f16_x2): Likewise.
 592         (vst1q_f32_x2): Likewise.
 593         (vst1q_f64_x2): Likewise.
 594         (vst1q_p64_x2): Likewise.
 595         (vst1_s64_x3): Likewise.
 596         (vst1_u64_x3): Likewise.
 597         (vst1_f64_x3): Likewise.
 598         (vst1_s8_x3): Likewise.
 599         (vst1_p8_x3): Likewise.
 600         (vst1_s16_x3): Likewise.
 601         (vst1_p16_x3): Likewise.
 602         (vst1_s32_x3): Likewise.
 603         (vst1_u8_x3): Likewise.
 604         (vst1_u16_x3): Likewise.
 605         (vst1_u32_x3): Likewise.
 606         (vst1_f16_x3): Likewise.
 607         (vst1_f32_x3): Likewise.
 608         (vst1_p64_x3): Likewise.
 609         (vst1q_s8_x3): Likewise.
 610         (vst1q_p8_x3): Likewise.
 611         (vst1q_s16_x3): Likewise.
 612         (vst1q_p16_x3): Likewise.
 613         (vst1q_s32_x3): Likewise.
 614         (vst1q_s64_x3): Likewise.
 615         (vst1q_u8_x3): Likewise.
 616         (vst1q_u16_x3): Likewise.
 617         (vst1q_u32_x3): Likewise.
 618         (vst1q_u64_x3): Likewise.
 619         (vst1q_f16_x3): Likewise.
 620         (vst1q_f32_x3): Likewise.
 621         (vst1q_f64_x3): Likewise.
 622         (vst1q_p64_x3): Likewise.
 623
 624 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
 625
 626         * config/msp430/msp430.c (msp430_output_labelref): Prepend
 627         user_label_prefix to name.
 628
 629         * tree-core.h: Update comment about the format of NAME string
 630         passed to handler in attribute_spec.
 631
 632         * config/msp430/msp430.md: Remove erroneous subreg expression from
 633         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
 634         zero_extend{q,h}isi2.
 635
 636 2018-05-30  Borislav Petkov  <bp@suse.de>
 637
 638         * doc/extend.texi: Document some architecture specific
 639         constraints and sort entries.
 640
 641 2018-05-30  Martin Sebor  <msebor@redhat.com>
 642
 643         PR middle-end/85369
 644         * builtins.c (expand_builtin_stpcpy_1): New function.
 645         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
 646         only if the former succeeds.
 647
 648 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
 649
 650         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
 651         in saphira.
 652
 653 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
 654
 655         * doc/invoke.texi (-flinker-output): Document
 656
 657 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
 658
 659         * passes.c (ipa_write_summaries): Only modify statements if body
 660         is in memory.
 661         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
 662         incrementally linking.
 663         (ipa_passes): Likewise.
 664         * lto-cgraph.c (lto_output_node): When incrementally linking do not
 665         pass down resolution info.
 666         * common.opt (flag_incremental_link): Update info.
 667         * gcc.c (plugin specs): Turn flinker-output=* to
 668         -plugin-opt=-linker-output-known
 669         * toplev.c (compile_file): Also cut compilation when doing incremental
 670         link.
 671         * flag-types. (enum lto_partition_model): Add
 672         LTO_LINKER_OUTPUT_NOLTOREL.
 673         (invoke.texi): Add -flinker-output docs.
 674         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
 675         link same way as WPA; do not stream in dead initializers.
 676
 677         * dwarf2out.c (dwarf2out_die_ref_for_decl,
 678         darf2out_register_external_decl): Support incremental link.
 679
 680 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
 681
 682         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
 683
 684 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
 685
 686         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
 687         it down to simple_object_copy_lto_debug_sections.
 688         (run_gcc): Determine incremental LTO link time and configure
 689         lto1 into non-wpa mode, disable renaming of debug sections.
 690
 691 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
 692
 693         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
 694         descriptions of various incorrectly documented functions.
 695
 696 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
 697
 698         Revert:
 699         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
 700         address check not strict.
 701
 702 2018-05-30  Richard Biener  <rguenther@suse.de>
 703
 704         PR tree-optimization/85964
 705         * tracer.c (better_p): Drop initialized count check, we only
 706         call the function with initialized counts now.
 707         (find_best_successor): Do find a best edge if one
 708         has uninitialized count.
 709         (find_best_predecessor): Likewise.  Do BB frequency check only
 710         if count is initialized.
 711
 712 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
 713
 714         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
 715         (aarch64_ldrstr_offset_compare): New.
 716         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
 717         load/store orderings.
 718         (aarch64_gen_adjusted_ldpstp): Likewise.
 719
 720 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
 721
 722         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
 723         Check for subset of GENERAL_REGS and FP_REGS.
 724         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
 725         r=w alternative.
 726
 727 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
 728
 729         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
 730         and wi::to_poly_offset.  Add the current offset and then check
 731         whether the sum fits, rather than using an unchecked addition of
 732         a checked term.  Check for a shwi rather than a uhwi.
 733         * expr.c (get_bit_range): Use tree_to_poly_uint64.
 734         (store_constructor): Use poly_int_tree_p.
 735         (expand_expr_real_1): Likewise.
 736         * function.c (assign_temp): Likewise.
 737         * fold-const.c (const_binop): Use poly_int_tree_p and
 738         wi::to_poly_offset.
 739         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
 740         division.
 741         * ipa-icf-gimple.c (func_checker::compare_operand): Use
 742         to_poly_offset for MEM offsets.
 743         * ipa-icf.c (sem_variable::equals): Likewise.
 744         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
 745         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
 746         wi::to_poly_offset for BIT_FIELD_REF offsets.
 747         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
 748         wi::to_poly_offset.
 749         * var-tracking.c (emit_note_insn_var_location): Use
 750         tree_to_poly_uint64.
 751
 752 2018-05-29  Jim Wilson  <jimw@sifive.com>
 753
 754         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
 755
 756 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
 757
 758         PR target/85950
 759         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
 760         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
 761         sequence.
 762         (sse4_1_round<mode>2): Use nonimmediate_operand
 763         for operand 1 predicate.
 764
 765 2018-05-29  Martin Sebor  <msebor@redhat.com>
 766             Richard Biener  <rguenther@suse.de>
 767
 768         PR testsuite/85888
 769         * calls.c (get_size_range): Call determine_value_range instead
 770         of get_value_range..
 771         * tree-vrp.h (determine_value_range): Declared new function.
 772         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
 773
 774 2018-05-29  Richard Biener  <rguenther@suse.de>
 775
 776         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
 777         sure to use non-pattern stmts for get_earlier_stmt arguments.
 778         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
 779         called on pattern stmts.
 780         (get_later_stmt): Likewise.
 781
 782 2018-05-29  Martin Liska  <mliska@suse.cz>
 783
 784         PR gcov-profile/85759
 785         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
 786         env variables.
 787
 788 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
 789
 790         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
 791         VEC_UNPACK_*_EXPR.
 792         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
 793         VEC_PACK_*_EXPR.
 794
 795         PR target/85918
 796         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
 797         VEC_PACK_FLOAT_EXPR): New tree codes.
 798         * tree-pretty-print.c (op_code_prio): Handle
 799         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
 800         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
 801         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
 802         * tree-inline.c (estimate_operator_cost): Likewise.
 803         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
 804         * fold-const.c (const_binop): Likewise.
 805         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
 806         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
 807         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
 808         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
 809         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
 810         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
 811         * expr.c (expand_expr_real_2): Likewise.
 812         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
 813         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
 814         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
 815         optabs.
 816         * optabs.c (expand_widen_pattern_expr): For
 817         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
 818         sign from result type rather than operand's type.
 819         (expand_binop_directly): For vec_packu_float_optab and
 820         vec_packs_float_optab allow result type to be different from operand's
 821         type.
 822         * optabs-tree.c (optab_for_tree_code): Handle
 823         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
 824         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
 825         * tree-vect-generic.c (expand_vector_operations_1):  Handle
 826         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
 827         VEC_PACK_FLOAT_EXPR.
 828         * tree-vect-stmts.c (supportable_widening_operation): Handle
 829         FIX_TRUNC_EXPR.
 830         (supportable_narrowing_operation): Handle FLOAT_EXPR.
 831         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
 832         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
 833         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
 834         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
 835         mode attributes.
 836         (vec_pack<floatprefix>_float_<mode>): New expander.
 837         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
 838         attributes.
 839         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
 840         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
 841         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
 842         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
 843         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
 844         Document.
 845         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
 846         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
 847         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
 848         VEC_PACK_FLOAT_EXPR): Document.
 849
 850 2018-05-29  Richard Biener  <rguenther@suse.de>
 851
 852         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
 853         member.
 854         (stmt_vec_info_vec): Make pointer.
 855         (init_stmt_vec_info_vec): Remove.
 856         (free_stmt_vec_info_vec): Likewise.
 857         (set_stmt_vec_info_vec): New function.
 858         (free_stmt_vec_infos): Likewise.
 859         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
 860         (set_vinfo_for_stmt): Likewise.
 861         (get_earlier_stmt): Likewise.
 862         (get_later_stmt): Likewise.
 863         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
 864         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
 865         (vec_info::~vec_info): Free stmt_vec_infos.
 866         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
 867         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
 868         (pass_slp_vectorize::execute): Likewise.
 869         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
 870         (free_stmt_vec_info_vec): Likewise.
 871         (set_stmt_vec_info_vec): New function.
 872         (free_stmt_vec_infos): Likewise.
 873         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
 874         the global stmt_vec_info_vec.
 875         * tree-parloops.c (gather_scalar_reductions): Use
 876         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
 877         vector.
 878
 879 2018-05-29  Richard Biener  <rguenther@suse.de>
 880
 881         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
 882
 883 2018-05-29  Martin Liska  <mliska@suse.cz>
 884             David Malcolm  <dmalcolm@redhat.com>
 885
 886         * vec.c (test_reverse): New.
 887         (vec_c_tests): Add new test.
 888         * vec.h (vl_ptr>::reverse): New function.
 889
 890 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
 891
 892         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
 893
 894         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
 895         and later.
 896
 897 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
 898
 899         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
 900
 901 2018-05-28  Richard Biener  <rguenther@suse.de>
 902
 903         PR tree-optimization/85933
 904         * tree-vect-data-refs.c (vect_record_base_alignments): Only
 905         look at stmts marked as vectorizable.
 906
 907 2018-05-28  Richard Biener  <rguenther@suse.de>
 908
 909         PR tree-optimization/85934
 910         * tree-vect-generic.c (expand_vector_operations_1): Hoist
 911         vector boolean check before scalar optimization.
 912
 913 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
 914
 915         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
 916         for armv5te.
 917
 918 2018-05-28  Mark Wielaard  <mark@klomp.org>
 919
 920         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
 921         if it is an expression containing a minus sign.
 922
 923 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
 924
 925         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
 926
 927 2018-05-27  Paul Koning  <ni1d@arrl.net>
 928
 929         * config/pdp11/pdp11.md (truncsihi2): Remove.
 930
 931 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
 932             Chung-Ju Wu  <jasonwucj@gmail.com>
 933
 934         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
 935         implementation.
 936         (unaligned_store_dw): Ditto.
 937         * config/nds32/nds32-memory-manipulation.c
 938         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
 939         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
 940         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
 941         (emit_setmem_word_loop): Rename to ...
 942         (emit_setmem_doubleword_loop): ... this.
 943         (nds32_gen_dup_4_byte_to_word_value): New function.
 944         (nds32_gen_dup_8_byte_to_double_word_value): New function.
 945         (nds32_expand_setmem_loop): Refine implementation.
 946         (nds32_expand_setmem_loop_v3m): Ditto.
 947         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
 948         pattern.
 949
 950 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
 951
 952         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
 953
 954 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
 955
 956         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
 957         (nds32_init_machine_status): Initialize machine->attr_naked_p and
 958         machine->attr_no_prologue_p.
 959         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
 960         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
 961         (nds32_expand_epilogue): Consider attr_naked_p.
 962         (nds32_expand_epilogue_v3pop): Likewise.
 963         (nds32_can_use_return_insn): Likewise.
 964         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
 965         attr_no_prologue_p fields.
 966         * config/nds32/nds32.opt (mret-in-naked-func): New option.
 967
 968 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
 969
 970         PR target/85918
 971         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
 972         attributes.
 973         * config/i386/sse.md
 974         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
 975         Rename to ...
 976         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
 977         ... this.
 978         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
 979         Rename to ...
 980         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
 981         ... this.
 982         (*<floatsuffix>floatv2div2sf2): Rename to ...
 983         (*float<floatunssuffix>v2div2sf2): ... this.
 984         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
 985         (float<floatunssuffix>v2div2sf2_mask): ... this.
 986         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
 987         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
 988         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
 989         to ...
 990         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
 991         ... this.
 992         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
 993         Rename to ...
 994         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
 995         ... this.
 996         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
 997         Rename to ...
 998         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
 999         ... this.
1000         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
1001         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
1002         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
1003         gen_ufix_truncv8dfv8si2.
1004         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
1005         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
1006         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
1007         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
1008         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
1009         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
1010         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
1011         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
1012
1013 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
1014
1015         PR target/85900
1016         PR target/85345
1017         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
1018
1019 2018-05-25  Jim Wilson  <jimw@sifive.com>
1020
1021         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
1022         * config/riscv/riscv.c (struct machine_function): Add
1023         interrupt_handler_p and attribute_checked_p fields.
1024         (riscv_attribute_table): Add interrupt.
1025         (riscv_interrupt_type_p): New.
1026         (riscv_save_reg_p): Save extra regs for interrupt handler.
1027         (riscv_use_save_libcall): Return false  for interrupt handler.
1028         (riscv_first_stack_step): Add forward declaration.
1029         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
1030         for interrupt handler with large frame.  Use it for saved reg list.
1031         (riscv_expand_prologue): Move flag_stack_usage_info support to
1032         eliminate duplication.
1033         (riscv_expand_epilogue): Generate mret for interrupt handler.
1034         (riscv_epilogue_uses): New.
1035         (riscv_can_use_return_insn): Return false for interrupt handler.
1036         (riscv_function_ok_for_sibcall): Likewise.
1037         (riscv_set_current_function): Add interrupt handler support.
1038         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
1039         * config/riscv/riscv.md (UNSPECV_MRET): New.
1040         (GP_REGNUM): New.
1041         (riscv_frflags, riscv_fsflags): Use tab after opcode.
1042         (riscv_mret): New.
1043         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
1044
1045 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
1046
1047         PR tree-optimization/85712
1048         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
1049         this candidate has already been replaced in-situ by a copy.
1050
1051 2018-05-25  Jason Merrill  <jason@redhat.com>
1052
1053         PR c++/80485 - inline function non-zero address.
1054         * symtab.c (nonzero_address): Check DECL_COMDAT.
1055
1056 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
1057
1058         PR target/83628
1059         * config/alpha/alpha.md (ashlsi3): New insn pattern.
1060         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
1061         extension of SImode operation.  Use const123_operand predicate.
1062         (*saddsi_1): Remove.
1063         (*saddl_se_1): Ditto.
1064         (*ssubsi_1): Ditto.
1065         (*ssubl_se_1): Ditto.
1066         * config/alpha/predicates.md (const123_operand): New predicate.
1067         * config/alpha/constraints.md (P): Use IN_RANGE.
1068
1069 2018-05-25  Richard Biener  <rguenther@suse.de>
1070
1071         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
1072         defaulted to true.
1073         (ref_maybe_used_by_stmt_p): Likewise.
1074         (stmt_may_clobber_ref_p): Likewise.
1075         (stmt_may_clobber_ref_p_1): Likewise.
1076         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
1077         and pass it along.
1078         (ref_maybe_used_by_stmt_p): Likewise.
1079         (stmt_may_clobber_ref_p): Likewise.
1080         (stmt_may_clobber_ref_p_1): Likewise.
1081         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
1082         the alias oracle to disambiguate DRs with stmts DR analysis
1083         couldn't handle.
1084         (vect_analyze_data_refs): Do not give up on not analyzable
1085         DRs for BB vectorization.  Remove code truncating the dataref
1086         vector.
1087
1088 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
1089
1090         PR target/85832
1091         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1092         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
1093         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
1094
1095 2018-05-25  Richard Biener  <rguenther@suse.de>
1096
1097         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
1098         function, combining stmt data ref gathering and fatal analysis
1099         parts.
1100         (vect_analyze_data_refs): Remove now redudnant code and simplify.
1101         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
1102         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
1103         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
1104         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
1105
1106 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
1107
1108         PR tree-optimization/85720
1109         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
1110         SCC if all partitions are builtins.
1111         (version_loop_by_alias_check): New parameter.  Generate cancelable
1112         runtime alias check if all partitions are builtins.
1113         (distribute_loop): Update call to above function.
1114
1115 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
1116
1117         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
1118         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
1119         (parm_default_def_partition_arg): Ditto.
1120         (set_parm_default_def_partition): Ditto.
1121         (get_parm_default_def_partitions): Ditto and make it static.
1122         (get_undefined_value_partitions): Ditto and make it static.
1123         (remove_ssa_form): Refactor call to init_var_map here.
1124         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
1125         computation for loop region.
1126         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
1127         (register_default_def): Delete.
1128         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
1129         (parm_default_def_partition_arg): Ditto.
1130         (set_parm_default_def_partition): Ditto.
1131         (get_parm_default_def_partitions): Ditto and make it static.
1132         (get_undefined_value_partitions): Ditto and make it static.
1133         (coalesce_with_default, coalesce_with_default): Update comment.
1134         (create_coalesce_list_for_region): New func factored out from
1135         create_outofssa_var_map.
1136         (populate_coalesce_list_for_outofssa): New func factored out from
1137         create_outofssa_var_map and coalesce_ssa_name.
1138         (create_outofssa_var_map): Delete.
1139         (coalesce_ssa_name): Refactor to support live range computation.
1140         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
1141         (get_parm_default_def_partitions): Delete.
1142         (get_undefined_value_partitions): Ditto.
1143         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
1144         computation for loop region.
1145         (new_tree_live_info, loe_visit_block): Ditto.
1146         (live_worklist, set_var_live_on_entry): Ditto.
1147         (calculate_live_on_exit, verify_live_on_entry): Ditto.
1148         * tree-ssa-live.h (struct _var_map): New fields.
1149         (init_var_map): Change decl.
1150         (region_contains_p): New.
1151
1152 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
1153
1154         * tree-ssa-live.h (live_merge_and_clear): Delete.
1155
1156 2018-05-25  Richard Biener  <rguenther@suse.de>
1157
1158         PR c++/85912
1159         * tree-dump.c (dequeue_and_dump): Remove access to removed
1160         operand 2 of a SWITCH_EXPR.
1161
1162 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
1163
1164         * doc/sourcebuild.texi (vect_double_cond_arith): Include
1165         multiplication and division.
1166         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
1167         (cond_udiv@var{m}, cond_umod@var{m}): Document.
1168         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
1169         (cond_udiv_optab, cond_umod_optab): New optabs.
1170         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
1171         (IFN_COND_RDIV): New internal functions.
1172         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
1173         TRUNC_MOD_EXPR and RDIV_EXPR.
1174         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
1175         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
1176         New unspecs.
1177         (SVE_INT_BINARY): Include mult.
1178         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
1179         (optab, sve_int_op): Handle mult.
1180         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
1181         UNSPEC_COND_DIV.
1182         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
1183         for SVE_INT_BINARY_SD.
1184
1185 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
1186
1187         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
1188         (optab, sve_int_op): Handle div and udiv.
1189         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
1190         for SVE_INT_BINARY_SD.
1191         (*<optab><mode>3): New insn for the same.
1192
1193 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
1194
1195         * tree-vect-patterns.c: Include predict.h.
1196         (vect_recog_divmod_pattern): Restrict check for division support
1197         to when optimizing for size.
1198
1199 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
1200
1201         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
1202         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
1203         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
1204         (gimple_match_op::set_op): Likewise.
1205         (gimple_resimplify4): Declare.
1206         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
1207         (expr::gen_transform): Likewise.
1208         (decision_tree::gen): Generate a simplification routine for 4 operands.
1209         * gimple-match-head.c (gimple_simplify): Add an overload for
1210         4 operands.  In the top-level function, handle up to 4 call
1211         arguments and call gimple_resimplify4.
1212         (gimple_resimplify4): New function.
1213         (build_call_internal): Pass a fourth operand.
1214         (maybe_push_to_seq): Likewise.
1215         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
1216         Fold VEC_COND_EXPRs of an operation and a default value into
1217         an IFN_COND_* function if possible.
1218         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
1219         New unspecs.
1220         (SVE_COND_FP_BINARY): Include them.
1221         (optab, sve_fp_op): Handle them.
1222         (SVE_INT_BINARY_REV): New code iterator.
1223         (SVE_COND_FP_BINARY_REV): New int iterator.
1224         (commutative): New int attribute.
1225         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
1226         Declare.
1227         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
1228         function.
1229         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
1230         (*cond_<optab><mode>): New patterns for reversed operands.
1231
1232 2018-05-25  Richard Biener  <rguenther@suse.de>
1233
1234         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
1235         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
1236         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
1237         (STMT_VINFO_GROUPED_ACCESS): Adjust.
1238         * tree-vect-data-refs.c (everywhere): Adjust users.
1239         * tree-vect-loop.c (everywhere): Likewise.
1240         * tree-vect-slp.c (everywhere): Likewise.
1241         * tree-vect-stmts.c (everywhere): Likewise.
1242         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1243
1244 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1245
1246         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
1247         Rename to...
1248         (gcc_cv_as_section_exclude): ... this.
1249         Try Solaris as #exclude syntax.
1250         * configure: Regenerate.
1251         * config.in: Regenerate.
1252         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
1253         SECTION_EXCLUDE.
1254         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
1255         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
1256
1257         * varasm.c (default_elf_asm_named_section): Don't check if
1258         HAVE_GAS_SECTION_EXCLUDE is defined.
1259
1260 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
1261
1262         * doc/md.texi: Update the documentation of the cond_* optabs
1263         to mention the new final operand.  Fix GET_MODE_NUNITS call.
1264         Describe the scalar case too.
1265         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
1266         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
1267         instead of 2.
1268         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
1269         (get_conditional_internal_fn): Update comment.
1270         * tree-vect-loop.c (vectorizable_reduction): Pass the original
1271         accumulator value as a final argument to conditional functions.
1272         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
1273         a define_expand and add an "else" operand.  Assert for now that
1274         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
1275         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
1276         (*cond_<optab><mode>): New patterns.
1277         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
1278         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
1279         (UNSPEC_COND_EOR): Delete.
1280         (optab): Remove associated mappings.
1281         (SVE_INT_BINARY): New code iterator.
1282         (sve_int_op): Remove int attribute and add "minus" to the code
1283         attribute.
1284         (SVE_COND_INT_OP): Delete.
1285         (SVE_COND_FP_OP): Rename to...
1286         (SVE_COND_FP_BINARY): ...this.
1287
1288 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
1289
1290         * optabs.c (can_reuse_operands_p): New function.
1291         (maybe_legitimize_operands): Try to reuse the results for
1292         earlier operands.
1293
1294 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
1295
1296         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1297         Add {q} suffix to insn mnemonic.
1298
1299 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1300
1301         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1302         (msp430_warn_func_return): New.
1303
1304 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
1305
1306         * fold-const.c (tree_nonzero_bits): New function.
1307         * fold-const.h (tree_nonzero_bits): Likewise.
1308         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1309         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1310
1311 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
1312
1313         PR target/85900
1314         PR target/85345
1315         * varasm.c (assemble_alias): Check ifunc_resolver only on
1316         FUNCTION_DECL.
1317
1318 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
1319
1320         PR target/85903
1321         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1322         when memory input operand is handled.
1323
1324 2018-05-24  Luis Machado  <luis.machado@linaro.org>
1325
1326         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1327         global.
1328         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1329
1330 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
1331
1332         * match.pd: Delay FMA folds until after vectorization.
1333
1334 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
1335
1336         PR target/83009
1337         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1338         address check not strict.
1339
1340 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
1341
1342         * gimple-match.h (gimple_match_op): New class.
1343         (mprts_hook): Replace parameters with a gimple_match_op *.
1344         (maybe_build_generic_op): Likewise.
1345         (gimple_simplified_result_is_gimple_val): Replace parameters with
1346         a const gimple_match_op *.
1347         (gimple_simplify): Replace code_helper * and tree * parameters with
1348         a gimple_match_op * parameter.
1349         (gimple_resimplify1): Replace code_helper *, tree and tree *
1350         parameters with a gimple_match_op * parameter.
1351         (gimple_resimplify2): Likewise.
1352         (gimple_resimplify3): Likewise.
1353         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1354         parameters with a gimple_match_op * parameter.
1355         * gimple-match-head.c (gimple_simplify): Change prototypes of
1356         auto-generated functions to take a gimple_match_op * instead of
1357         separate code_helper * and tree * parameters.  Make the same
1358         change in the top-level overload and update calls to the
1359         gimple_resimplify routines.  Update calls to the auto-generated
1360         functions and to maybe_push_res_to_seq in the publicly-facing
1361         operation-specific gimple_simplify overloads.
1362         (gimple_match_op::MAX_NUM_OPS): Define.
1363         (gimple_resimplify1): Replace rcode and ops with a single res_op
1364         parameter.  Update call to gimple_simplify.
1365         (gimple_resimplify2): Likewise.
1366         (gimple_resimplify3): Likewise.
1367         (mprts_hook): Replace parameters with a gimple_match_op *.
1368         (maybe_build_generic_op): Likewise.
1369         (build_call_internal): Replace type, nargs and ops with
1370         a gimple_match_op *.
1371         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1372         with a single gimple_match_op *.  Update calls to mprts_hook,
1373         build_call_internal and gimple_simplified_result_is_gimple_val.
1374         Factor out code that is common to the tree_code and combined_fn cases.
1375         * genmatch.c (expr::gen_transform): Replace tem_code and
1376         tem_ops with a gimple_match_op called tem_op.  Update calls
1377         to the gimple_resimplify functions and maybe_push_res_to_seq.
1378         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1379         res_ops.  Update call to the gimple_resimplify functions.
1380         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1381         (decision_tree::gen): Make the functions take a gimple_match_op *
1382         called res_op instead of separate res_code and res_ops parameters.
1383         Update call accordingly.
1384         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1385         and ops with a single res_op parameter.  Update calls to
1386         maybe_build_generic_op and maybe_push_res_to_seq.
1387         (fold_stmt_1): Update calls to gimple_simplify and
1388         replace_stmt_with_simplification.
1389         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1390         and gimple_simplified_result_is_gimple_val.
1391         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1392         gimple_simplify.
1393         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1394         with a gimple_match_op *.
1395         (vn_nary_build_or_lookup): Likewise.  Update call to
1396         vn_nary_build_or_lookup_1.
1397         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1398         gimple_match_op *.  Update calls to the gimple_resimplify routines
1399         and to gimple_simplified_result_is_gimple_val.
1400         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1401         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1402         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1403         (visit_nary_op): Likewise.
1404         (visit_reference_op_load): Likewise.
1405
1406 2018-05-23  Luis Machado  <luis.machado@linaro.org>
1407
1408         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1409         modifier for printing the step amount.
1410
1411 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
1412
1413         PR target/78849
1414         * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1415         types.
1416
1417 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
1418
1419         * doc/sourcebuild.texi (Endianness): New subsubsection.
1420
1421 2018-05-23  Luis Machado  <luis.machado@linaro.org>
1422
1423         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1424         <prefetch_dynamic_strides>: New const bool field.
1425         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1426         prefetch_dynamic_strides.
1427         (exynosm1_prefetch_tune): Likewise.
1428         (thunderxt88_prefetch_tune): Likewise.
1429         (thunderx_prefetch_tune): Likewise.
1430         (thunderx2t99_prefetch_tune): Likewise.
1431         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1432         false.
1433         (aarch64_override_options_internal): Update to set
1434         PARAM_PREFETCH_DYNAMIC_STRIDES.
1435         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1436         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1437         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1438         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1439         prefetch-dynamic-strides setting.
1440
1441 2018-05-23  Luis Machado  <luis.machado@linaro.org>
1442
1443         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1444         <minimum_stride>: New const int field.
1445         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1446         minimum_stride field defaulting to -1.
1447         (exynosm1_prefetch_tune): Likewise.
1448         (thunderxt88_prefetch_tune): Likewise.
1449         (thunderx_prefetch_tune): Likewise.
1450         (thunderx2t99_prefetch_tune): Likewise.
1451         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1452         <default_opt_level>: Set to 3.
1453         (aarch64_override_options_internal): Update to set
1454         PARAM_PREFETCH_MINIMUM_STRIDE.
1455         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1456         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1457         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1458         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1459         stride is constant and is below the minimum stride threshold.
1460
1461 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1462
1463         * config/arm/arm-cpus.in (mode26): Delete.
1464         (armv4): Delete mode26 reference.
1465         * config/arm/arm.c (arm_configure_build_target): Delete use of
1466         isa_bit_mode26.
1467
1468 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
1469
1470         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1471         New insn pattern.
1472         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1473         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
1474         for non-SSE modes.
1475         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
1476         (floatunsdidf2): Ditto.
1477
1478 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
1479
1480         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1481         (fixuns_trunc<mode>si2_avx512f): Ditto.
1482         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1483         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1484         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1485
1486 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
1487
1488         PR rtl-optimization/79985
1489         * df-scan.c (df_insn_refs_collect): Remove special case for
1490         global registers and asm statements.
1491
1492 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
1493
1494         * extend.texi (Global Register Variables): Rewrite the bullet list.
1495         Note that the register is available for allocation. Note that access
1496         via inline asm must use constraints. Add note about async-signal
1497         handlers. Remove paragraph about automagic register selection.
1498
1499 2018-05-23  Richard Biener  <rguenther@suse.de>
1500
1501         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1502         of fixed offset from memset VN.
1503
1504 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
1505
1506         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1507         first_interp field.
1508         (alloc_cand_and_find_basis): Initialize first_interp field.
1509         (slsr_process_mul): Modify first_interp field.
1510         (slsr_process_add): Likewise.
1511         (slsr_process_cast): Modify first_interp field for each new
1512         interpretation.
1513         (slsr_process_copy): Likewise.
1514         (dump_candidate): Dump first_interp field.
1515         (replace_mult_candidate): Process all interpretations, not just
1516         subsequent ones.
1517         (replace_rhs_if_not_dup): Likewise.
1518         (replace_one_candidate): Likewise.
1519
1520 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
1521
1522         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1523         Add new boolean.
1524         (aarch64_needs_frame_chain): New function.
1525         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1526
1527 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
1528
1529         PR target/84882
1530         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1531         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1532         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1533         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1534         as true for strict-align.
1535         (aarch64_can_inline_p): Perform checks even when callee has no
1536         attributes to check for strict alignment.
1537         * doc/extend.texi (AArch64 Function Attributes): Document
1538         no-strict-align.
1539         * doc/invoke.texi: (AArch64 Options): Likewise.
1540
1541 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
1542
1543         PR tree-optimization/85853
1544         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1545         the handling of the root of the node to...
1546         (vect_slp_analyze_node_operations_1): ...this new function,
1547         and run the whole thing with the child nodes' def types
1548         set according to their SLP node's def type.
1549
1550 2018-05-23  Richard Biener  <rguenther@suse.de>
1551
1552         PR middle-end/85874
1553         * tree-data-ref.c (create_runtime_alias_checks): Defer
1554         and ignore overflow warnings.
1555
1556 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
1557
1558         PR tree-optimization/85822
1559         * tree-vrp.c (is_masked_range_test): Fix handling of negative
1560         constants.
1561
1562 2018-05-23  Richard Biener  <rguenther@suse.de>
1563
1564         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1565         memset constants via native_interpret_expr.
1566
1567 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
1568
1569         PR target/85345
1570         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1571         attribute.
1572         (cgraph_node::create_alias): Likewise.
1573         (cgraph_node::get_availability): Check ifunc_resolver instead
1574         of looking up ifunc attribute.
1575         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1576         * varasm.c (do_assemble_alias): Likewise.
1577         (assemble_alias): Likewise.
1578         (default_binds_local_p_3): Likewise.
1579         * cgraph.h (cgraph_node): Add ifunc_resolver.
1580         (cgraph_node::only_called_directly_or_aliased_p): Return false
1581         for IFUNC resolver.
1582         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1583         attribute.
1584         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1585         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1586         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1587         instead of looking up ifunc attribute.
1588
1589 2018-05-22  Luis Machado  <luis.machado@linaro.org>
1590
1591         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1592
1593 2018-05-22  Martin Sebor  <msebor@redhat.com>
1594
1595         PR middle-end/85359
1596         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1597         only when expasion succeeds.
1598         (expand_builtin_strcmp): Same.
1599         (expand_builtin_strncmp): Same.
1600
1601 2018-05-22  Martin Sebor  <msebor@redhat.com>
1602
1603         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1604
1605 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
1606             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1607
1608         * config/aarch64/aarch64-ldpstp.md: Replace uses of
1609         aarch64_mem_pair_operand with memory_operand and delete operand swapping
1610         code.
1611         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1612         Add check for legitimate_address.
1613         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1614         (aarch64_swap_ldrstr_operands): New.
1615         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1616         Define prototype.
1617
1618 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
1619             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1620
1621         * config/aarch64/aarch64.md: New patterns to generate stp
1622         and ldp.
1623         (store_pair_sw, store_pair_dw): New patterns to generate stp for
1624         single words and double words.
1625         (load_pair_sw, load_pair_dw): Likewise.
1626         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1627         Delete.
1628         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1629         Delete.
1630         * config/aarch64/aarch64-ldpstp.md: Modify peephole
1631         for different mode ldpstp and add peephole for merged zero stores.
1632         Likewise for loads.
1633         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1634         Add size check.
1635         (aarch64_gen_store_pair): Rename calls to match new patterns.
1636         (aarch64_gen_load_pair): Rename calls to match new patterns.
1637         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1638         (load_pair<DREG:mode><DREG2:mode>): ... This.
1639         (store_pair<mode>): Rename to...
1640         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1641         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1642         New mode iterators.
1643         (V_INT_EQUIV): Handle SImode.
1644         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1645         New predicate.
1646
1647 2018-05-22  Martin Sebor  <msebor@redhat.com>
1648
1649         PR c/85623
1650         * calls.c (maybe_warn_nonstring_arg): Use string length to set
1651         or ajust the presumed bound on an operation to avoid unnecessary
1652         warnings.
1653
1654 2018-05-22  Martin Sebor  <msebor@redhat.com>
1655
1656         PR tree-optimization/85826
1657         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1658         assuming that a DECL necesarily has a constant size.
1659
1660 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
1661
1662         PR middle-end/85862
1663         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1664
1665 2018-05-22  Richard Biener  <rguenther@suse.de>
1666
1667         PR tree-optimization/85834
1668         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1669         non-constant and non-zero memset arguments.
1670
1671 2018-05-22  Martin Liska  <mliska@suse.cz>
1672
1673         PR ipa/85607
1674         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1675
1676 2018-05-22  Richard Biener  <rguenther@suse.de>
1677
1678         PR tree-optimization/85863
1679         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1680         comparisons when vectype is specified.
1681         (vectorizable_condition): Do not specify vectype for
1682         vect_is_simple_cond when SLP vectorizing.
1683
1684 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
1685
1686         PR target/85657
1687         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1688         define __ibm128 as long double.
1689         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1690         as a distinct type when IEEE 128-bit support is enabled.
1691         (init_float128_ieee): Fix up conversions between IFmode and IEEE
1692         128-bit types to use the correct functions.
1693         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1694         convert between 128-bit floating point types that have different
1695         modes but the same representation, instead of using gen_lowpart to
1696         makean alias.
1697         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1698         KFmode.
1699         (IFKF_reg): New attributes to give the register constraints for
1700         IFmode and KFmode.
1701         (extend<mode>tf2_internal): New insns to mark an explicit
1702         conversion between 128-bit floating point types that have a
1703         different mode but share the same representation.
1704
1705 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
1706
1707         PR tree-optimization/85814
1708         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1709         a null return from get_strinfo when unsharing the next
1710         strinfo in the chain.
1711
1712 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
1713
1714         PR gcc/84923
1715         * varasm.c (weak_finish): Clean up weak_decls.
1716
1717 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1718
1719         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1720         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1721         UNSPEC_UADALP values.
1722         * config/aarch64/iterators.md (ABAL): New int iterator.
1723         (ABDL2): Likewise.
1724         (ADALP): Likewise.
1725         (sur): Add mappings for the above.
1726         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1727         New define_insn.
1728         (aarch64_<sur>abal<mode>_4): Likewise.
1729         (aarch64_<sur>adalp<mode>_3): Likewise.
1730         (<sur>sadv16qi): New define_expand.
1731
1732 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
1733
1734         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1735         (*movdf_internal): Ditto.
1736         (*rcpsf2_sse): Ditto.
1737         (*rsqrtsf2_sse): Ditto.
1738         (*sqrt<mode>2_sse): Ditto.
1739
1740 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
1741
1742         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1743         eor3q<mode>4.
1744         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1745         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1746         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1747         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1748         vbcaxq_s64): New.
1749         * config/aarch64/arm_neon.h: Likewise.
1750         * config/aarch64/iterators.md (VQ_I): New.
1751
1752 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
1753
1754         * config.gcc: Add arc/t-multilib-linux to tmake_file for
1755         arc*-*-linux*.
1756         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1757         MULTILIB_DIRNAMES
1758
1759 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
1760
1761         * config/nds32/constraints.md (S): New constraint.
1762         * config/nds32/nds32.md (call_internal): Use constraint S.
1763         (call_value_internal): Likewise.
1764         (sibcall_internal): Likewise.
1765         (sibcall_value_internal): Likewise.
1766
1767 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
1768             Chung-Ju Wu  <jasonwucj@gmail.com>
1769
1770         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1771         into consideration.
1772
1773 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
1774             Chung-Ju Wu  <jasonwucj@gmail.com>
1775
1776         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1777         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1778         (nds32_rtx_costs_impl): Simplify.
1779         (nds32_address_cost_impl): Simplify.
1780         (nds32_init_rtx_costs): New function.
1781         (nds32_rtx_costs_speed_prefer): Likewise.
1782         (nds32_rtx_costs_size_prefer): Likewise.
1783         (nds32_address_cost_speed_prefer): Likewise.
1784         (nds32_address_cost_speed_fwprop): Likewise.
1785         (nds32_address_cost_size_prefer): Likewise.
1786         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1787         * config/nds32/nds32.c (nds32_option_override): Use
1788         nds32_init_rtx_costs function.
1789
1790 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
1791
1792         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1793         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1794         (TARGET_PIPELINE_N8): Likewise.
1795         (TARGET_PIPELINE_N10): Likewise.
1796         (TARGET_PIPELINE_N13): Likewise.
1797         (TARGET_PIPELINE_GRAYWOLF): Likewise.
1798
1799 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
1800
1801         * config/nds32/nds32-fpu.md: Update copyright year.
1802
1803 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1804
1805         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1806
1807 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1808
1809         * config/nds32/nds32.c
1810         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1811         * config/nds32/nds32.opt (minline-asm-r15): New option.
1812
1813 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1814
1815         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1816         MASK_HW_ABS.
1817         * config/nds32/nds32.md (abssi2): New pattern.
1818
1819 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
1820
1821         * config/i386/i386.md (rex64namesuffix): New mode attribute.
1822         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1823         Merge insn pattern from sse_cvtsi2ss<round_name> and
1824         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1825         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1826         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1827         using SWI48 mode iterator.
1828         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1829         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1830         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1831         pattern from sse_cvttss2si<round_saeonly_name>
1832         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1833         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1834         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1835         using SWI48 mode iterator.
1836         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1837         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1838         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1839         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1840         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1841         using SWI48 mode iterator.
1842         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1843         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1844         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1845         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1846         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1847         SWI48 mode iterator.
1848         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1849         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1850         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1851         pattern from sse_cvttsd2si<round_saeonly_name>
1852         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1853
1854 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1855
1856         * config/nds32/nds32-md-auxiliary.c
1857         (nds32_valid_smw_lwm_base_p): Refine.
1858         (nds32_output_smw_single_word): Refine.
1859         (nds32_output_smw_double_word): New.
1860         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1861
1862 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1863
1864         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1865         (nds32_output_stack_pop): Refine.
1866         (nds32_expand_unaligned_load): Refine.
1867         (nds32_expand_unaligned_store): Refine.
1868
1869 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
1870             Chung-Ju Wu  <jasonwucj@gmail.com>
1871
1872         * config/nds32/constants.md: Add TP_REGNUM constant.
1873         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1874         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1875         UNSPEC_ADD32.
1876         * config/nds32/nds32-doubleword.md: Consider flag_pic.
1877         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1878         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1879         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1880         and PIC code generation.
1881         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1882         code generation.
1883         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1884         optimization.
1885         * config/nds32/nds32.md: Support TLS and PIC.
1886         * config/nds32/nds32.c: Support TLS and PIC.
1887         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1888         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1889         predicate.
1890
1891 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
1892
1893         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1894         mode with E_ prefix.
1895
1896 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
1897             Chung-Ju Wu  <jasonwucj@gmail.com>
1898
1899         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1900         * config/nds32/nds32-md-auxiliary.c
1901         (symbolic_reference_mentioned_p): New.
1902         (nds32_legitimize_ict_address): New.
1903         (nds32_expand_ict_move): New.
1904         (nds32_indirect_call_referenced_p): New.
1905         (nds32_symbol_binds_local_p): Delete.
1906         (nds32_long_call_p): Modify.
1907         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1908         * config/nds32/nds32-protos.h
1909         (symbolic_reference_mentioned_p): Declare.
1910         (nds32_legitimize_ict_address): Declare.
1911         (nds32_expand_ict_move): Declare.
1912         (nds32_indirect_call_referenced_p): Declare.
1913         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1914         (nds32_relax_group): Use nds32_ict_const_p as condition.
1915         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1916         (nds32_asm_file_start): Output ict_model directive in asm code.
1917         (nds32_legitimate_address_p): Consider indirect call.
1918         (nds32_print_operand): Consider indirect call.
1919         (nds32_print_operand_address): Consider indirect call.
1920         (nds32_insert_attributes): Handle "indirect_call" attribute.
1921         (TARGET_LEGITIMATE_ADDRESS_P): Define.
1922         (TARGET_LEGITIMATE_CONSTANT_P): Define.
1923         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1924         (TARGET_DELEGITIMIZE_ADDRESS): Define.
1925         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1926         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1927         (TARGET_ICT_MODEL_SMALL): Define.
1928         (TARGET_ICT_MODEL_LARGE): Define.
1929         * config/nds32/nds32.md (movsi): Consider ict model.
1930         (call, call_value): Consider ict model.
1931         (sibcall, sibcall_value): Consider ict model.
1932         * config/nds32/nds32.opt (mict-model): New option.
1933         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1934         model.
1935
1936 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
1937             Monk Chiang  <sh.chiang04@gmail.com>
1938             Jim Wilson <jimw@sifive.com>
1939
1940         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1941         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
1942         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1943         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1944         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
1945         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1946         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1947         compute save_libcall_adjustment properly.
1948         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1949         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1950         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1951         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1952         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1953         (ABI_SPEC): Handle mabi=ilp32e.
1954         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1955         (RVE): Add RVE mask.
1956         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1957         <-march>: Add rv32e as an example.
1958
1959 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
1960
1961         PR c++/82899
1962         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1963         (intra_create_variable_infos): Handle C++ constructors.
1964
1965 2018-05-18  Martin Liska  <mliska@suse.cz>
1966
1967         * passes.def: Remove a redundant pass.
1968
1969 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
1970
1971         PR bootstrap/85838
1972         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1973
1974 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1975
1976         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1977         (ARMv4): Update.
1978         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1979         (ARMv6m): Update.
1980         (armv2, armv2a, armv3, armv3m): Delete architectures.
1981         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1982         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1983         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1984         Delete cpus.
1985         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1986         (*mulsidi3adddi): Likewise.
1987         (mulsidi3): Likewise.
1988         (*mulsidi3_nov6): Likewise.
1989         (umulsidi3): Likewise.
1990         (umulsidi3_nov6): Likewise.
1991         (umaddsidi4): Likewise.
1992         (*umulsidi3adddi): Likewise.
1993         (smulsi3_highpart): Likewise.
1994         (*smulsi3_highpart_nov6): Likewise.
1995         (umulsi3_highpart): Likewise.
1996         (*umulsi3_highpart_nov6): Likewise.
1997         * config/arm/arm.h (arm_arch3m): Delete.
1998         * config/arm/arm.c (arm_arch3m): Delete.
1999         (arm_option_override_internal): Update armv3-related comment.
2000         (arm_configure_build_target): Delete use of isa_bit_mode32.
2001         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
2002         (arm_rtx_costs_internal): Delete check of arm_arch3m.
2003         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
2004         (mulsa3): Likewise.
2005         (mulusa3): Likewise.
2006         * config/arm/arm-protos.h (arm_arch3m): Delete.
2007         * config/arm/arm-tables.opt: Regenerate.
2008         * config/arm/arm-tune.md: Likewise.
2009         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
2010         deleted architectures.
2011
2012 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2013
2014         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
2015         (armv5t, armv5te): New features.
2016         (ARMv5, ARMv5e): Delete fgroups.
2017         (ARMv5t, ARMv5te): Adjust for above changes.
2018         (ARMv6m): Likewise.
2019         (armv5, armv5e): Delete arches.
2020         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
2021         arm_arch5.
2022         (*call_reg_arm): Likewise.
2023         (*call_value_reg_armv5): Likewise.
2024         (*call_value_reg_arm): Likewise.
2025         (*call_symbol): Likewise.
2026         (*call_value_symbol): Likewise.
2027         (*sibcall_insn): Likewise.
2028         (*sibcall_value_insn): Likewise.
2029         (clzsi2): Likewise.
2030         (prefetch): Likewise.
2031         (define_split and define_peephole2 dependent on arm_arch5):
2032         Likewise.
2033         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
2034         arm_arch5e.
2035         (TARGET_ARM_QBIT): Likewise.
2036         (TARGET_DSP_MULTIPLY): Likewise.
2037         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
2038         (arm_arch5, arm_arch5e): Delete.
2039         (arm_arch5t, arm_arch5te): Declare.
2040         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
2041         (arm_arch5t): Declare.
2042         (arm_option_reconfigure_globals): Update for the above.
2043         (arm_options_perform_arch_sanity_checks): Update comment, replace
2044         use of arm_arch5 with arm_arch5t.
2045         (use_return_insn): Likewise.
2046         (arm_emit_call_insn): Likewise.
2047         (output_return_instruction): Likewise.
2048         (arm_final_prescan_insn): Likewise.
2049         (arm_coproc_builtin_available): Likewise.
2050         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
2051         arm_arch5e with arm_arch5t and arm_arch5te.
2052         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
2053         (arm_arch5t, arm_arch5te): Declare.
2054         * config/arm/arm-tables.opt: Regenerate.
2055         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
2056         * config/arm/t-multilib: Likewise.
2057         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
2058         instead of arm_arch5.
2059         (*call_reg_thumb1): Likewise.
2060         (*call_value_reg_thumb1_v5): Likewise.
2061         (*call_value_reg_thumb1): Likewise.
2062         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
2063         unreachable path.
2064         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
2065
2066 2018-05-18  Martin Liska  <mliska@suse.cz>
2067
2068         PR gcov-profile/84846
2069         * doc/gcov.texi: Document -t option of gcov tool.
2070
2071 2018-05-18  Martin Liska  <mliska@suse.cz>
2072
2073         PR gcov-profile/84846
2074         * gcov.c (print_usage): Add new -t option.
2075         (process_args): Handle the option.
2076         (generate_results): Use stdout as output when requested by
2077         the option.
2078
2079 2018-05-18  Martin Liska  <mliska@suse.cz>
2080
2081         PR gcov-profile/84846
2082         * coverage.c (coverage_init): Write PWD to .gcno file.
2083         * doc/gcov.texi: Document how working directory is printed.
2084         * gcov-dump.c (dump_gcov_file): Print PWD.
2085         * gcov.c (output_intermediate_file): Likewise.
2086         (read_graph_file): Read PWD string.
2087         (output_lines): Print PWD.
2088
2089 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
2090
2091         PR middle-end/85817
2092         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
2093         for retval and return false if all args to phi are zero.
2094
2095 2018-05-18  Richard Biener  <rguenther@suse.de>
2096
2097         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
2098         method.
2099         (evrp_dom_walker::before_dom_children): Call it.
2100
2101 2018-05-18  Richard Biener  <rguenther@suse.de>
2102
2103         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
2104         results when processing array refs with variable index.
2105
2106 2018-05-18  Toon Moene  <toon@moene.org>
2107
2108         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
2109         directly after that of -floop-interchange. Indicate that both
2110         options are enabled by default when specifying -O3.
2111
2112 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2113
2114         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
2115         iterator.  Delete separate integer-mode vec_set<mode> expander.
2116         (aarch64_simd_vec_setv2di): Delete.
2117         (vec_setv2di): Delete.
2118         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
2119         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
2120         the "w, r" alternative.
2121
2122 2018-05-18  Martin Liska  <mliska@suse.cz>
2123
2124         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
2125         * tree-pass.h (make_pass_lower_switch_O0): New function.
2126         * tree-switch-conversion.c (node_has_low_bound): Remove.
2127         (node_has_high_bound): Likewise.
2128         (node_is_bounded): Likewise.
2129         (class pass_lower_switch): Make it a template type and create
2130         two instances.
2131         (pass_lower_switch::execute): Add template argument.
2132         (make_pass_lower_switch): New function.
2133         (make_pass_lower_switch_O0): New function.
2134         (do_jump_if_equal): Remove.
2135         (emit_case_nodes): Simplify to just handle all 3 cases and leave
2136         all the hard work to tree optimization passes.
2137
2138 2018-05-18  Martin Liska  <mliska@suse.cz>
2139
2140         * dbgcnt.c (limit_low): Renamed from limit.
2141         (limit_high): New variable.
2142         (dbg_cnt_is_enabled): Check for upper limit.
2143         (dbg_cnt): Adjust dumping.
2144         (dbg_cnt_set_limit_by_index): Add new argument for high
2145         value.
2146         (dbg_cnt_set_limit_by_name): Likewise.
2147         (dbg_cnt_process_single_pair): Parse new format.
2148         (dbg_cnt_process_opt): Use strtok.
2149         (dbg_cnt_list_all_counters): Remove 'value' and add
2150         'limit_high'.
2151         * doc/invoke.texi: Document changes.
2152
2153 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
2154
2155         * doc/sourcebuild.texi (scalar_all_fma): Document.
2156         * tree.def (FMA_EXPR): Delete.
2157         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
2158         * internal-fn.c (ternary_direct): New macro.
2159         (expand_ternary_optab_fn): Likewise.
2160         (direct_ternary_optab_supported_p): Likewise.
2161         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
2162         * builtins.c (fold_builtin_fma): Delete.
2163         (fold_builtin_3): Don't call it.
2164         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
2165         * expr.c (expand_expr_real_2): Likewise.
2166         * fold-const.c (operand_equal_p): Likewise.
2167         (fold_ternary_loc): Likewise.
2168         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
2169         * gimple.c (DEFTREECODE): Likewise.
2170         * gimplify.c (gimplify_expr): Likewise.
2171         * optabs-tree.c (optab_for_tree_code): Likewise.
2172         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
2173         * tree-eh.c (operation_could_trap_p): Likewise.
2174         (stmt_could_throw_1_p): Likewise.
2175         * tree-inline.c (estimate_operator_cost): Likewise.
2176         * tree-pretty-print.c (dump_generic_node): Likewise.
2177         (op_code_prio): Likewise.
2178         * tree-ssa-loop-im.c (stmt_cost): Likewise.
2179         * tree-ssa-operands.c (get_expr_operands): Likewise.
2180         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
2181         * fold-const-call.h (fold_fma): Delete.
2182         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
2183         CFN_FNMA and CFN_FNMS.
2184         (fold_fma): Delete.
2185         * genmatch.c (combined_fn): New enum.
2186         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
2187         (commutative_op): New function.
2188         (commutate): Use it.  Handle more than 2 operands.
2189         (dt_operand::gen_gimple_expr): Use commutative_op.
2190         (parser::parse_expr): Allow :c to be used with non-binary
2191         operators if the commutative operand is known.
2192         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
2193         CFN_FMS, CFN_FNMA and CFN_FNMS.
2194         (backprop::process_assign_use): Remove FMA_EXPR handling.
2195         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
2196         (gen_hsa_fma): New function.
2197         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
2198         IFN_FNMA and IFN_FNMS.
2199         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
2200         * gimple-fold.h (follow_all_ssa_edges): Declare.
2201         * gimple-fold.c (follow_all_ssa_edges): New function.
2202         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
2203         gimple_build interface and use follow_all_ssa_edges to fold the result.
2204         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
2205         instead of checking for optabs directly.
2206         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
2207         rather than FMA_EXPRs.
2208         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
2209         call to IFN_FMA instead of an FMA_EXPR.
2210
2211 2018-05-17  Jim Wilson  <jimw@sifive.com>
2212
2213         * expr.c (do_tablejump): When converting index to Pmode, if we have a
2214         sign extended promoted subreg, and the range does not have the sign bit
2215         set, then do a sign extend.
2216
2217         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
2218         test, check for sign extended subreg and/or constant operands, and
2219         do a sign extend in that case.
2220
2221 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
2222
2223         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
2224         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
2225         Add untyped.
2226         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
2227         Change logics_shift_reg to logics_shift_imm.
2228         (thunderx2t99_fp_loadpair_basic): Delete.
2229         (thunderx2t99_fp_storepair_basic): Delete.
2230         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
2231         (thunderx2t99_asimd_polynomial): Delete.
2232         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
2233         and neon_fp_mul_d_scalar_q.
2234         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
2235         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
2236         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
2237         (thunderx2t99_asimd_lut): Add missing tbl types.
2238         (thunderx2t99_asimd_ext): Delete.
2239         (thunderx2t99_asimd_load1_1_mult): Delete.
2240         (thunderx2t99_asimd_load1_2_mult): Delete.
2241         (thunderx2t99_asimd_load1_ldp): New.
2242         (thunderx2t99_asimd_load1): New.
2243         (thunderx2t99_asimd_load2): Add missing *load2* types.
2244         (thunderx2t99_asimd_load3): New.
2245         (thunderx2t99_asimd_load4): New.
2246         (thunderx2t99_asimd_store1_1_mult): Delete.
2247         (thunderx2t99_asimd_store1_2_mult): Delete.
2248         (thunderx2t99_asimd_store2_mult): Delete.
2249         (thunderx2t99_asimd_store2_onelane): Delete.
2250         (thunderx2t99_asimd_store_stp): New.
2251         (thunderx2t99_asimd_store1): New.
2252         (thunderx2t99_asimd_store2): New.
2253         (thunderx2t99_asimd_store3): New.
2254         (thunderx2t99_asimd_store4): New.
2255
2256 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
2257
2258         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
2259         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
2260
2261 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
2262             Segher Boessenkool  <segher@kernel.crashing.org>
2263
2264         PR target/85698
2265         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
2266         operand.
2267
2268 2018-05-17  Richard Biener  <rguenther@suse.de>
2269
2270         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
2271         for pruning loop and prune defs feeding only already visited PHIs.
2272
2273 2018-05-17  Richard Biener  <rguenther@suse.de>
2274
2275         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
2276
2277 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
2278             Richard Biener  <rguenther@suse.de>
2279
2280         PR tree-optimization/85793
2281         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
2282         for VMAT_ELEMENTWISE.
2283
2284 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
2285
2286         * internal-fn.h (lookup_internal_fn): Declare
2287         * internal-fn.c (lookup_internal_fn): New function.
2288         * gimple.c (gimple_build_call_from_tree): Handle calls to
2289         internal functions.
2290         * gimple-pretty-print.c (dump_gimple_call): Print "." before
2291         internal function names.
2292         * tree-pretty-print.c (dump_generic_node): Likewise.
2293         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2294
2295 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
2296
2297         * gimple-fold.h (gimple_build): Make the function forms take
2298         combined_fn rather than built_in_function.
2299         (gimple_simplify): Likewise.
2300         * gimple-match-head.c (gimple_simplify): Likewise.
2301         * gimple-fold.c (gimple_build): Likewise.
2302         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2303         rather than gimple_build_call_internal.
2304         (get_initial_defs_for_reduction): Likewise.
2305         (vect_create_epilog_for_reduction): Likewise.
2306         (vectorizable_live_operation): Likewise.
2307
2308 2018-05-17  Martin Liska  <mliska@suse.cz>
2309
2310         * gimple-ssa-sprintf.c (format_directive): Do not use
2311         space in between 'G_' and '('.
2312
2313 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
2314
2315         PR target/85323
2316         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2317         even if the mask is not all ones.
2318
2319         PR target/85323
2320         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2321         vector.
2322         (ix86_gimple_fold_builtin): Likewise.
2323
2324         PR target/85323
2325         * config/i386/i386.c: Include tree-vector-builder.h.
2326         (ix86_vector_shift_count): New function.
2327         (ix86_fold_builtin): Fold shift builtins by scalar count.
2328         (ix86_gimple_fold_builtin): Likewise.
2329
2330         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2331         _mm512_setzero): New intrinsics.
2332
2333 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
2334             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2335
2336         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2337         code generation for cases where splatting a value is not useful.
2338         * simplify-rtx.c (simplify_ternary_operation): Simplify
2339         vec_merge across a vec_duplicate and a paradoxical subreg forming
2340         a vector mode to a vec_concat.
2341
2342 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
2343
2344         * config.gcc: Support "goldmont-plus".
2345         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2346         "goldmont-plus".
2347         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2348         PROCESSOR_GOLDMONT_PLUS.
2349         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2350         (processor_target_table): Add "goldmont-plus".
2351         (PTA_GOLDMONT_PLUS): Define.
2352         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2353         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2354         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2355         (fold_builtin_cpu): Add "goldmont-plus".
2356         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2357         (ix86_option_override_internal): Add "goldmont-plus".
2358         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2359         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2360         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2361         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2362
2363 2018-05-17  Richard Biener  <rguenther@suse.de>
2364
2365         PR tree-optimization/85757
2366         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2367         remove defs that only feed that PHI from further processing.
2368
2369 2018-05-16  Jim Wilson  <jimw@sifive.com>
2370
2371         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2372         asterisk to name.
2373         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2374
2375 2018-05-16  Mark Wielaard  <mark@klomp.org>
2376
2377         * dwarf2out.c (count_index_strings): New function.
2378         (output_indirect_strings): Call count_index_strings and generate
2379         header for dwarf_version >= 5.
2380
2381 2018-05-16  Mark Wielaard  <mark@klomp.org>
2382
2383         * dwarf2out.c (dwarf_FORM): New function.
2384         (set_indirect_string): Use dwarf_FORM.
2385         (reset_indirect_string): Likewise.
2386         (size_of_die): Likewise.
2387         (value_format): Likewise.
2388         (output_die): Likewise.
2389         (add_skeleton_AT_string): Likewise.
2390         (output_macinfo_op): Likewise.
2391         (index_string): Likewise.
2392         (output_index_string_offset): Likewise.
2393         (output_index_string): Likewise.
2394         (count_index_strings): Likewise.
2395
2396 2018-05-16  Carl Love  <cel@us.ibm.com>
2397
2398         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2399         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2400
2401 2018-05-16  Martin Jambor  <mjambor@suse.cz>
2402
2403         * ipa-prop.c (ipa_free_all_edge_args): Remove.
2404         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2405
2406 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
2407
2408         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2409         (fnma<mode>4): Likewise.
2410         (fms<mode>4): Likewise.
2411         (fnms<mode>4): Likewise.
2412         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2413         (aarch64_fnma<mode>4): Likewise.
2414         (aarch64_fms<mode>4): Likewise.
2415         (aarch64_fnms<mode>4): Likewise.
2416         (aarch64_fnmadd<mode>4): Likewise.
2417
2418 2018-05-16  Jason Merrill  <jason@redhat.com>
2419
2420         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
2421
2422 2018-05-16  Richard Biener  <rguenther@suse.de>
2423
2424         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2425         (dump_stmt_cost): Declare.
2426         (add_stmt_cost): Dump cost we add.
2427         (add_stmt_costs): New function.
2428         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2429         No longer exported.
2430         (vect_analyze_stmt): Adjust prototype.
2431         (vectorizable_condition): Likewise.
2432         (vectorizable_live_operation): Likewise.
2433         (vectorizable_reduction): Likewise.
2434         (vectorizable_induction): Likewise.
2435         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2436         cost vector to pass to vectorizable_ and record afterwards.
2437         (vect_model_reduction_cost): Take cost vector argument and adjust.
2438         (vect_model_induction_cost): Likewise.
2439         (vectorizable_reduction): Likewise.
2440         (vectorizable_induction): Likewise.
2441         (vectorizable_live_operation): Likewise.
2442         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2443         SLP_TREE_NUMBER_OF_VEC_STMTS.
2444         (vect_analyze_slp_cost_1): Remove.
2445         (vect_analyze_slp_cost): Likewise.
2446         (vect_slp_analyze_node_operations): Take visited args and
2447         a target cost vector.  Avoid processing already visited stmt sets.
2448         (vect_slp_analyze_operations): Use a local cost vector to gather
2449         costs and register those of non-discarded instances.
2450         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2451         (vect_schedule_slp_instance): Remove copying of
2452         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
2453         zero.
2454         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2455         adding cost.  Record cost entry location.
2456         (vect_prologue_cost_for_slp_op): Function to compute cost of
2457         a constant or invariant generated for SLP vect in the prologue,
2458         split out from vect_analyze_slp_cost_1.
2459         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
2460         (vect_model_promotion_demotion_cost): Likewise.
2461         (vect_model_store_cost): Likewise, make static.
2462         (vect_model_load_cost): Likewise.
2463         (vectorizable_bswap): Add cost vector arg and adjust.
2464         (vectorizable_call): Likewise.
2465         (vectorizable_simd_clone_call): Likewise.
2466         (vectorizable_conversion): Likewise.
2467         (vectorizable_assignment): Likewise.
2468         (vectorizable_shift): Likewise.
2469         (vectorizable_operation): Likewise.
2470         (vectorizable_store): Likewise.
2471         (vectorizable_load): Likewise.
2472         (vectorizable_condition): Likewise.
2473         (vectorizable_comparison): Likewise.
2474         (can_vectorize_live_stmts): Likewise.
2475         (vect_analyze_stmt): Likewise.
2476         (vect_transform_stmt): Adjust calls to vectorizable_*.
2477         * tree-vectorizer.c: Include gimple-pretty-print.h.
2478         (dump_stmt_cost): New function.
2479
2480 2018-05-16  Richard Biener  <rguenther@suse.de>
2481
2482         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2483         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2484         * tree-ssa-dse.c: Include tree-ssa-loop.h.
2485         (check_name): New callback.
2486         (dse_classify_store): Track cycles via a visited bitmap of PHI
2487         defs and simplify handling of in-loop and across loop dead stores
2488         and properly fail for loop-variant refs.  Handle byte-tracking with
2489         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2490         limiting the walk.
2491
2492 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
2493
2494         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2495         (vect_get_mask_type_for_stmt): Likewise.
2496         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2497         split out from...
2498         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
2499         to determine the statement's vector type and the vector type that
2500         should be used for calculating nunits.  Deal with cases in which
2501         the type has to be deferred.
2502         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2503         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2504         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2505         (vect_determine_vf_for_stmt): New functions, split out from...
2506         (vect_determine_vectorization_factor): ...here.
2507         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2508         (vect_get_mask_type_for_stmt): New functions, split out from
2509         vect_determine_vectorization_factor.
2510
2511 2018-05-16  Richard Biener  <rguenther@suse.de>
2512
2513         * tree-cfg.c (verify_gimple_assign_ternary): Properly
2514         verify the [VEC_]COND_EXPR embedded comparison.
2515
2516 2018-05-15  Martin Sebor  <msebor@redhat.com>
2517
2518         PR tree-optimization/85753
2519         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2520         RECORD_TYPE in addition to ARRAY_TYPE.
2521
2522 2018-05-15  Martin Sebor  <msebor@redhat.com>
2523
2524         PR middle-end/85643
2525         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2526
2527 2018-05-15  Richard Biener  <rguenther@suse.de>
2528
2529         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2530         add by_clobber_p one.  Change algorithm to collect all defs
2531         representing uses we need to walk and try reducing them to
2532         a single one before failing.
2533         (dse_dom_walker::dse_optimize_stmt): Adjust.
2534
2535 2018-05-13  Mark Wielaard  <mark@klomp.org>
2536
2537         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2538         (size_of_loc_descr): Likewise.
2539         (output_loc_operands): Likewise.
2540         (output_loc_operands_raw): Likewise.
2541         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2542         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2543         (hash_loc_operands): Likewise.
2544         (compare_loc_operands): Likewise.
2545
2546 2018-05-14  Mark Wielaard  <mark@klomp.org>
2547
2548         * dwarf2out.c (count_index_addrs): New function.
2549         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2550
2551 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
2552
2553         PR tree-optimization/83648
2554         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2555         return value as malloc candidate.
2556
2557 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
2558
2559         PR ipa/85734
2560         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
2561         param as true in call to suggest_attribute.
2562
2563 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
2564
2565         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2566         -mreadonly-in-sdata.
2567
2568 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2569
2570         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2571         New pattern.
2572         (aarch64_crypto_aesd_fused): Likewise.
2573
2574 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
2575
2576         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2577         (movsi_aarch64): Likewise.
2578         (load_pairsi): Likewise.
2579         (load_pairdi): Likewise.
2580         (store_pairsi): Likewise.
2581         (store_pairdi): Likewise.
2582         (load_pairsf): Likewise.
2583         (load_pairdf): Likewise.
2584         (store_pairsf): Likewise.
2585         (store_pairdf): Likewise.
2586         (zero_extend): Likewise.
2587         (trunc): Swap alternatives.
2588         (fcvt_target): Add '?' to prefer w over r.
2589
2590 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
2591
2592         PR target/85756
2593         * config/i386/i386.md: Disallow non-commutative arithmetics in
2594         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2595         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
2596         in the peephole2 before it.
2597
2598 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
2599
2600         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2601         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2602         (ix86_handle_option): Handle -mcldemote.
2603         * config.gcc: New header.
2604         * config/i386/cldemoteintrin.h: New file.
2605         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2606         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2607         -mcldemote.
2608         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2609         OPTION_MASK_ISA_CLDEMOTE.
2610         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2611         (ix86_valid_target_attribute_inner_p): Ditto.
2612         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2613         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2614         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2615         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2616         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2617         (cldemote): New.
2618         * config/i386/i386.opt: Add -mcldemote.
2619         * config/i386/x86intrin.h: New header.
2620         * doc/invoke.texi: Add -mcldemote.
2621
2622 2018-05-14  Richard Biener  <rguenther@suse.de>
2623
2624         * doc/match-and-simplify.texi: Adjust :s documentation.
2625
2626 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
2627
2628         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2629         intended memcpy size.
2630         (REORDER_45): Likewise.
2631
2632 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
2633
2634         * sort.cc: New file.
2635         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2636         * vec.c (qsort_chk): Use gcc_qsort.
2637         * Makefile.in (OBJS-libcommon): Add sort.o.
2638         (build/sort.o): New target.  Use it...
2639         (BUILD_RTL): ... here, and...
2640         (build/gencfn-macros): ... here, and...
2641         (build/genmatch): ... here.
2642
2643 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
2644             Chung-Ju Wu  <jasonwucj@gmail.com>
2645
2646         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2647         * config/nds32/nds32-graywolf.md: New file.
2648         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2649         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2650         pipeline.
2651         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2652         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2653         * config/nds32/nds32.md (pipeline_model): Add graywolf.
2654         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2655         * config/nds32/pipelines.md: Include n15 settings.
2656
2657 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
2658             Chung-Ju Wu  <jasonwucj@gmail.com>
2659
2660         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2661         * config/nds32/nds32-n13.md: New file.
2662         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2663         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2664         pipeline.
2665         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2666         * config/nds32/nds32.md (pipeline_model): Add n13.
2667         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2668         * config/nds32/pipelines.md: Include n13 settings.
2669
2670 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
2671             Chung-Ju Wu  <jasonwucj@gmail.com>
2672
2673         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2674         * config/nds32/nds32-n10.md: New file.
2675         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2676         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2677         pipeline.
2678         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2679         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2680         * config/nds32/nds32.md (pipeline_model): Add n10.
2681         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2682         * config/nds32/pipelines.md: Include n10 settings.
2683
2684 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
2685             Kito Cheng  <kito.cheng@gmail.com>
2686             Chung-Ju Wu  <jasonwucj@gmail.com>
2687
2688         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2689         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2690         Add enum values for DSP extension instructions.
2691         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2692         New constraints.
2693         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2694         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2695         New code iterators.
2696         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2697         * config/nds32/nds32-dspext.md: New file for DSP implementation.
2698         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2699         * config/nds32/nds32-intrinsic.md: Likewise.
2700         * config/nds32/nds32_intrinsic.h: Likewise.
2701         * config/nds32/nds32-md-auxiliary.c: Likewise.
2702         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2703         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2704         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2705         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2706         * config/nds32/nds32-protos.h: New declarations for DSP extension.
2707         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2708         TYPE_DMAC in switch statement.
2709         * config/nds32/nds32.c: New checking and implementation for DSP
2710         extension instructions.
2711         * config/nds32/nds32.h: Likewise.
2712         * config/nds32/nds32.md: Likewise.
2713         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2714         * config/nds32/predicates.md: Implement new predicates for DSP
2715         extension.
2716
2717 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
2718
2719         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2720         Reformat alternatives and attributes so it is easier to identify
2721         which constraints/attributes go with which instruction.
2722         (mov<mode>_hardfloat32, FMOVE64): Likewise.
2723         (mov<mode>_softfloat32, FMOVE64): Likewise.
2724         (mov<mode>_hardfloat64, FMOVE64): Likewise.
2725         (mov<mode>_softfloat64, FMOVE64): Likewise.
2726
2727 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
2728
2729         * doc/extend.texi (PowerPC Built-in Functions): Rename this
2730         subsection.
2731         (Basic PowerPC Built-in Functions): The new name of the
2732         subsection previously known as "PowerPC Built-in Functions".
2733         (Basic PowerPC Built-in Functions Available on all Configurations):
2734         New subsubsection.
2735         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2736         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2737         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2738         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2739
2740 2018-05-11  Martin Jambor  <mjambor@suse.cz>
2741
2742         PR ipa/85655
2743         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2744         single const.
2745
2746 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
2747
2748         PR target/85733
2749         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2750
2751 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
2752
2753         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2754         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2755         (ix86_handle_option): Handle -mwaitpkg.
2756         * config.gcc: New header.
2757         * config/i386/cpuid.h (bit_WAITPKG): New bit.
2758         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2759         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2760         function type.
2761         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2762         OPTION_MASK_ISA_WAITPKG.
2763         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2764         (ix86_option_override_internal): Add PTA_WAITPKG.
2765         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2766         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2767         IX86_BUILTIN_TPAUSE.
2768         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2769         __builtin_ia32_umwait and __builtin_ia32_tpause.
2770         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2771         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2772         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2773         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2774         UNSPECV_TPAUSE): New.
2775         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2776         * config/i386/i386.opt: Add -mwaitpkg.
2777         * config/i386/waitpkgintrin.h: New file.
2778         * config/i386/x86intrin.h: New header.
2779         * doc/invoke.texi: Add -mwaitpkg.
2780
2781 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
2782
2783         PR target/85606
2784         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2785         equivalent.
2786         (cortex-m0): Use armv6s-m isa.
2787         (cortex-m0plus): Likewise.
2788         (cortex-m1): Likewise.
2789         (cortex-m0.small-multiply): Likewise.
2790         (cortex-m0plus.small-multiply): Likewise.
2791         (cortex-m1.small-multiply): Likewise.
2792
2793 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
2794             Jakub Jelinek  <jakub@redhat.com>
2795
2796         PR tree-optimization/85692
2797         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2798         source permute as well.
2799
2800 2018-05-11  Martin Liska  <mliska@suse.cz>
2801
2802         PR sanitizer/85556
2803         * doc/extend.texi: Document LLVM style format for no_sanitize
2804         attribute.
2805
2806 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
2807
2808         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2809         mode_supports_vsx_dform_quad to mode_supports_dq_form.
2810         (mode_supports_vsx_dform_quad): Likewise.
2811         (mode_supports_vmx_dform): Move these functions to be next to the
2812         other mode_supports functions.
2813         (mode_supports_dq_form): Likewise.
2814         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2815         mode_supports_dq_form.
2816         (reg_offset_addressing_ok_p): Likewise.
2817         (offsettable_ok_by_alignment): Likewise.
2818         (rs6000_legitimate_offset_address_p): Likewise.
2819         (legitimate_lo_sum_address_p): Likewise.
2820         (rs6000_legitimize_address): Likewise.
2821         (rs6000_legitimize_reload_address): Likewise.
2822         (rs6000_secondary_reload_inner): Likewise.
2823         (rs6000_preferred_reload_class): Likewise.
2824         (rs6000_output_move_128bit): Likewise.
2825
2826 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
2827
2828         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2829         Generate SImode target register for null target.
2830         <case IX86_BUILTIN_XGETBV>: Ditto.
2831         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2832         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2833
2834 2018-05-10  Carl Love  <cel@us.ibm.com>
2835
2836         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2837         dcbtt and dcbtstt if operands[2] is 0.
2838
2839 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
2840
2841         PR target/85693
2842         * config/i386/sse.md (usadv64qi): New expander.
2843
2844 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
2845
2846         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2847         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2848         -maltivec=be support.
2849         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2850         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2851         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2852         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2853         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2854         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2855         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2856         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2857         altivec_vsumsws): Adjust.
2858         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2859         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2860         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2861         support.
2862         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2863         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2864         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2865         (altivec_lve<VI_char>x): Delete expand.
2866         (*altivec_lve<VI_char>x_internal): Rename to...
2867         (altivec_lve<VI_char>x): ... this.
2868         (altivec_lvxl_<mode>): Delete expand.
2869         (*altivec_lvxl_<mode>_internal): Rename to ...
2870         (altivec_lvxl_<mode>): ... this.
2871         (altivec_stvxl_<mode>): Delete expand.
2872         (*altivec_stvxl_<mode>_internal): Rename to ...
2873         (altivec_stvxl_<mode>): ... this.
2874         (altivec_stve<VI_char>x): Delete expand.
2875         (*altivec_stve<VI_char>x_internal): Rename to ...
2876         (altivec_stve<VI_char>x): ... this.
2877         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2878         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2879         reduc_plus_scal_<mode>): Adjust.
2880         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2881         comment.
2882         (rs6000_cpu_cpp_builtins): Adjust.
2883         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2884         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2885         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2886         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2887         -maltivec=be support.
2888         (rs6000_split_vec_extract_var): Adjust.
2889         (rs6000_split_v4si_init): Adjust.
2890         (swap_selector_for_mode): Delete.
2891         (altivec_expand_lvx_be, altivec_expand_stvx_be,
2892         altivec_expand_stvex_be): Delete.
2893         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2894         -maltivec=be support.
2895         (rs6000_gimple_fold_builtin): Ditto.
2896         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2897         Adjust.
2898         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2899         (TARGET_DIRECT_MOVE_64BIT): Adjust.
2900         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2901         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2902         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2903         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2904         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2905         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2906         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2907         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2908         anonymous split): Adjust.
2909         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2910         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2911
2912 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
2913
2914         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2915         when --with-gxx-include-dir is also specified.
2916         * configure: Regenerate.
2917
2918 2018-05-09  Jim Wilson  <jimw@sifive.com>
2919
2920         PR target/84797
2921         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2922         * config/riscv/t-withmultilib: New.
2923         * config/riscv/withmultilib.h: New.
2924         * doc/install.texi: Document RISC-V --with-multilib-list support.
2925
2926 2018-05-09  Richard Biener  <rguenther@suse.de>
2927
2928         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2929         vector.
2930         (vect_bb_vectorization_profitable_p): Adjust.  Compute
2931         actual scalar cost using the cost vector and the add_stmt_cost
2932         machinery.
2933
2934 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
2935
2936         PR rtl-optimization/85645
2937         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2938         in the REG_CFA_REGISTER note for LR, don't leave it empty.
2939
2940 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
2941
2942         PR rtl-optimization/85645
2943         * shrink-wrap.c (spread_components): Return a boolean saying if
2944         anything was changed.
2945         (try_shrink_wrapping_separate): Iterate spread_components until
2946         nothing changes anymore.
2947
2948 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
2949
2950         PR rtl-optimization/85645
2951         * regrename.c (build_def_use): Also kill the chains that include the
2952         destination of a REG_CFA_REGISTER note.
2953
2954 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
2955
2956         PR rtl-optimization/85645
2957         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2958         insn that has a REG_CFA_REGISTER note.
2959
2960 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
2961
2962         * cfgexpand.c (expand_clobber): New function.
2963         (expand_gimple_stmt_1): Use it.
2964         * tree-vect-stmts.c (vect_clobber_variable): New function,
2965         split out from...
2966         (vectorizable_simd_clone_call): ...here.
2967         (vectorizable_store): Emit a clobber either side of an
2968         IFN_STORE_LANES sequence.
2969         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2970
2971 2018-05-09  Tom de Vries  <tom@codesourcery.com>
2972
2973         PR target/85626
2974         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2975         (define_insn "trap_if_false"): Add exit after trap.
2976
2977 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
2978
2979         PR rtl-optimization/85638
2980         * bb-reorder.c: Include common/common-target.h.
2981         (create_forwarder_block): New function extracted from...
2982         (fix_up_crossing_landing_pad): ...here.  Rename into...
2983         (dw2_fix_up_crossing_landing_pad): ...this.
2984         (sjlj_fix_up_crossing_landing_pad): New function.
2985         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2986         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2987         from both partitions and exit the loop after one iteration.
2988
2989 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
2990
2991         Revert:
2992         * doc/extend.texi (PowerPC Built-in Functions): Rename this
2993         subsection.
2994         (Basic PowerPC Built-in Functions): The new name of the
2995         subsection previously known as "PowerPC Built-in Functions".
2996         (Basic PowerPC Built-in Functions Available on all Configurations):
2997         New subsubsection.
2998         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2999         subsubsection.
3000         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
3001         subsubsection.
3002         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
3003         subsubsection.
3004         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
3005         subsubsection.
3006
3007 2018-05-08  Jim Wilson  <jimw@sifive.com>
3008
3009         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
3010         (LD_EMUL_SUFFIX): New.
3011         (LINK_SPEC): Use it.
3012
3013 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
3014
3015         * doc/extend.texi (PowerPC Built-in Functions): Rename this
3016         subsection.
3017         (Basic PowerPC Built-in Functions): The new name of the
3018         subsection previously known as "PowerPC Built-in Functions".
3019         (Basic PowerPC Built-in Functions Available on all Configurations):
3020         New subsubsection.
3021         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
3022         subsubsection.
3023         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
3024         subsubsection.
3025         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
3026         subsubsection.
3027         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
3028         subsubsection.
3029
3030 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
3031
3032         PR target/85683
3033         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
3034         after cmpelim optimization.
3035
3036 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
3037
3038         * config.gcc: Support "goldmont".
3039         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
3040         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3041         PROCESSOR_GOLDMONT.
3042         * config/i386/i386.c (m_GOLDMONT): Define.
3043         (processor_target_table): Add "goldmont".
3044         (PTA_GOLDMONT): Define.
3045         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
3046         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
3047         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
3048         (fold_builtin_cpu): Add "goldmont".
3049         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
3050         (ix86_option_override_internal): Add "goldmont".
3051         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
3052         (processor_type): Add PROCESSOR_GOLDMONT.
3053         * config/i386/i386.md: Add CPU "glm".
3054         * config/i386/glm.md: New file.
3055         * config/i386/x86-tune.def: Add m_GOLDMONT.
3056         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
3057
3058 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
3059
3060         PR target/85572
3061         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
3062         E_V4DImode.
3063         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
3064         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
3065         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
3066
3067         PR target/85317
3068         * config/i386/i386.c (ix86_fold_builtin): Handle
3069         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
3070
3071         PR target/85480
3072         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
3073         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
3074
3075 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
3076
3077         PR target/85658
3078         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
3079         (check_arch): Likewise.
3080         (check_fpu): Return the result rather than printing it.
3081         (end arch): Fix operator precedence.
3082         (end cpu): Likewise.
3083         (END): Print the result from check_fpu.
3084
3085 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
3086             Alan Hayward  <alan.hayward@arm.com>
3087             David Sherwood  <david.sherwood@arm.com>
3088
3089         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
3090         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
3091         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
3092         (*fcmuo<mode>_and): New patterns.
3093
3094 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
3095
3096         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
3097         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
3098         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
3099         (cmp_op, sve_imm_con): New code attributes.
3100         (SVE_COND_INT_CMP, imm_con): Delete.
3101         (cmp_op): Remove above unspecs from int attribute.
3102         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
3103         to...
3104         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
3105         comparison-specific unspecs.
3106         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
3107         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
3108         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
3109         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
3110         (*vec_fcm<cmp_op><mode>): Rename to...
3111         (*fcm<cmp_op><mode>): ...this and adjust likewise.
3112         (*vec_fcmuo<mode>): Rename to...
3113         (*fcmuo<mode>): ...this and adjust likewise.
3114         (*pred_fcm<cmp_op><mode>): New pattern.
3115         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
3116         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
3117         functions.
3118         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
3119         and UNORDERED.
3120         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
3121         (aarch64_emit_sve_predicated_cond): New function.
3122         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
3123         (aarch64_emit_unspec_cond_or): Replace with...
3124         (aarch64_emit_sve_or_conds): ...this new function.  Use
3125         aarch64_emit_sve_ptrue_op for the individual comparisons and
3126         aarch64_emit_binop to OR them together.
3127         (aarch64_emit_inverted_unspec_cond): Replace with...
3128         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
3129         aarch64_emit_sve_ptrue_op for the comparison and
3130         aarch64_emit_unop to invert the result.
3131         (aarch64_expand_sve_vec_cmp_float): Update after the above
3132         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
3133
3134 2018-05-07  Nathan Sidwell  <nathan@acm.org>
3135
3136         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
3137         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
3138         (Backwards Compatibility): Likewise.
3139
3140 2018-05-07  Luis Machado  <luis.machado@linaro.org>
3141
3142         PR bootstrap/85681
3143         Revert:
3144         2018-05-07  Luis Machado  <luis.machado@linaro.org>
3145
3146         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3147         <prefetch_dynamic_strides>: New const bool field.
3148         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3149         prefetch_dynamic_strides.
3150         (exynosm1_prefetch_tune): Likewise.
3151         (thunderxt88_prefetch_tune): Likewise.
3152         (thunderx_prefetch_tune): Likewise.
3153         (thunderx2t99_prefetch_tune): Likewise.
3154         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3155         to false.
3156         (aarch64_override_options_internal): Update to set
3157         PARAM_PREFETCH_DYNAMIC_STRIDES.
3158         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3159         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3160         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3161         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3162         prefetch-dynamic-strides setting.
3163
3164         2018-05-07  Luis Machado  <luis.machado@linaro.org>
3165
3166         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3167         <minimum_stride>: New const int field.
3168         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3169         minimum_stride field.
3170         (exynosm1_prefetch_tune): Likewise.
3171         (thunderxt88_prefetch_tune): Likewise.
3172         (thunderx_prefetch_tune): Likewise.
3173         (thunderx2t99_prefetch_tune): Likewise.
3174         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3175         (aarch64_override_options_internal): Update to set
3176         PARAM_PREFETCH_MINIMUM_STRIDE.
3177         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3178         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3179         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3180         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3181         stride is constant and is below the minimum stride threshold.
3182
3183 2018-05-07  Luis Machado  <luis.machado@linaro.org>
3184
3185         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
3186         to 512.
3187
3188 2018-05-07  Luis Machado  <luis.machado@linaro.org>
3189
3190         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3191         <prefetch_dynamic_strides>: New const bool field.
3192         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3193         prefetch_dynamic_strides.
3194         (exynosm1_prefetch_tune): Likewise.
3195         (thunderxt88_prefetch_tune): Likewise.
3196         (thunderx_prefetch_tune): Likewise.
3197         (thunderx2t99_prefetch_tune): Likewise.
3198         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3199         to false.
3200         (aarch64_override_options_internal): Update to set
3201         PARAM_PREFETCH_DYNAMIC_STRIDES.
3202         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3203         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3204         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3205         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3206         prefetch-dynamic-strides setting.
3207
3208 2018-05-07  Luis Machado  <luis.machado@linaro.org>
3209
3210         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3211         <minimum_stride>: New const int field.
3212         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3213         minimum_stride field.
3214         (exynosm1_prefetch_tune): Likewise.
3215         (thunderxt88_prefetch_tune): Likewise.
3216         (thunderx_prefetch_tune): Likewise.
3217         (thunderx2t99_prefetch_tune): Likewise.
3218         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3219         (aarch64_override_options_internal): Update to set
3220         PARAM_PREFETCH_MINIMUM_STRIDE.
3221         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3222         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3223         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3224         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3225         stride is constant and is below the minimum stride threshold.
3226
3227 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
3228
3229         PR c++/85659
3230         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
3231         the type is addressable.  Don't force op into register if it has
3232         BLKmode.
3233
3234 2018-05-05  Roland McGrath  <mcgrathr@google.com>
3235
3236         PR other/77609
3237         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
3238         any section for which we don't know a specific type it should have,
3239         regardless of name.  Previously this was done only for the exact
3240         names ".init_array", ".fini_array", and ".preinit_array".
3241         (default_elf_asm_named_section): Add comment about
3242         relationship with default_section_type_flags and SECTION_NOTYPE.
3243         (get_section): Don't consider it a type conflict if one side has
3244         SECTION_NOTYPE and the other doesn't, as long as neither has the
3245         SECTION_BSS et al used in the default_section_type_flags logic.
3246
3247 2018-05-05  Tom de Vries  <tom@codesourcery.com>
3248
3249         PR target/85653
3250         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
3251         (workaround_barsyncs): New function.
3252         (nvptx_reorg): Use workaround_barsyncs.
3253         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3254         (define_expand "nvptx_membar_cta"): New define_expand.
3255         (define_insn "*nvptx_membar_cta"): New insn.
3256
3257 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
3258
3259         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
3260         To improve optimization opportunities.
3261         * builtin-types.def: The new needed builtin types for the above.
3262
3263 2018-05-04  Richard Biener  <rguenther@suse.de>
3264
3265         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
3266         * gimple-ssa-store-merging.c
3267         (imm_store_chain_info::output_merged_store): Remove redundant create,
3268         release split_store vector contents on failure.
3269         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
3270         scalar stmt vector on cache hit.
3271
3272 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
3273
3274         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
3275         Xilinx FP support.
3276         * config.gcc (powerpc-xilinx-eabi*): Remove.
3277         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
3278         support.
3279         (fusion_addis_mem_combo_load): Ditto.
3280         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
3281         FP support.
3282         (rs6000_cpu_cpp_builtins): Ditto.
3283         * config/rs6000/rs6000-linux.c
3284         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
3285         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
3286         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3287         support.
3288         (rs6000_setup_reg_addr_masks): Ditto.
3289         (rs6000_init_hard_regno_mode_ok): Ditto.
3290         (rs6000_option_override_internal): Ditto.
3291         (legitimate_lo_sum_address_p): Ditto.
3292         (rs6000_legitimize_address): Ditto.
3293         (rs6000_legitimize_reload_address): Ditto.
3294         (rs6000_legitimate_address_p): Ditto.
3295         (abi_v4_pass_in_fpr): Ditto.
3296         (setup_incoming_varargs): Ditto.
3297         (rs6000_gimplify_va_arg): Ditto.
3298         (rs6000_split_multireg_move): Ditto.
3299         (rs6000_savres_strategy): Ditto.
3300         (rs6000_emit_prologue_components): Ditto.
3301         (rs6000_emit_epilogue_components): Ditto.
3302         (rs6000_emit_prologue): Ditto.
3303         (rs6000_emit_epilogue): Ditto.
3304         (rs6000_elf_file_end): Ditto.
3305         (rs6000_function_value): Ditto.
3306         (rs6000_libcall_value): Ditto.
3307         * config/rs6000/rs6000.h: Ditto.
3308         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3309         (TARGET_MINMAX): ... this.  New.
3310         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3311         * config/rs6000/rs6000.md: Remove Xilinx FP support.
3312         (*movsi_internal1_single): Delete.
3313         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3314         mfpu=, mxilinx-fpu): Delete.
3315         * config/rs6000/singlefp.h: Delete.
3316         * config/rs6000/sysv4.h: Remove Xilinx FP support.
3317         * config/rs6000/t-rs6000: Ditto.
3318         * config/rs6000/t-xilinx: Delete.
3319         * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3320         * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3321         (VStype_simple): Delete.
3322         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3323         * config/rs6000/xfpu.h: Delete.
3324         * config/rs6000/xfpu.md: Delete.
3325         * config/rs6000/xilinx.h: Delete.
3326         * config/rs6000/xilinx.opt: Delete.
3327         * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3328         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3329
3330 2018-05-04  Tom de Vries  <tom@codesourcery.com>
3331
3332         PR libgomp/85639
3333         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3334         if ignore == 0.
3335
3336 2018-05-04  Richard Biener  <rguenther@suse.de>
3337
3338         PR middle-end/85627
3339         * tree-complex.c (update_complex_assignment): We are always in SSA form.
3340         (expand_complex_div_wide): Likewise.
3341         (expand_complex_operations_1): Likewise.
3342         (expand_complex_libcall): Preserve EH info of the original stmt.
3343         (tree_lower_complex): Handle removed blocks.
3344         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3345         on complex multiplication and division libcall builtins.
3346
3347 2018-05-04  Richard Biener  <rguenther@suse.de>
3348
3349         PR middle-end/85574
3350         * fold-const.c (negate_expr_p): Restrict negation of operand
3351         zero of a division to when we know that can happen without
3352         overflow.
3353         (fold_negate_expr_1): Likewise.
3354
3355 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
3356
3357         PR libstdc++/85466
3358         * real.h (real_nextafter): Declare.
3359         * real.c (real_nextafter): New function.
3360         * fold-const-call.c (fold_const_nextafter): New function.
3361         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3362         CASE_CFN_NEXTTOWARD.
3363         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3364         even when arg1_mode is different from arg0_mode.
3365
3366 2018-05-03  Nathan Sidwell  <nathan@acm.org>
3367
3368         * doc/extend.texi (Deprecated Features): Remove
3369         -ffriend-injection.
3370         (Backwards Compatibility): Likewise.
3371         * doc/invoke.texi (C++ Language Options): Likewise.
3372         (C++ Dialect Options): Likewise.
3373
3374 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
3375
3376         PR target/85530
3377         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3378         _mm512_mask_mullox_epi64): New intrinsics.
3379
3380 2018-05-03  Tom de Vries  <tom@codesourcery.com>
3381
3382         PR testsuite/85106
3383         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3384         dump files): Add offload-tree.
3385
3386 2018-05-03  Richard Biener  <rguenther@suse.de>
3387
3388         PR tree-optimization/85615
3389         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3390         to loops not nested in BBs loop father to avoid creating multi-entry
3391         loops.
3392
3393 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3394
3395         PR tree-optimization/70291
3396         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3397         arguments.  Change return type to tree.  Emit libcall as a new
3398         statement rather than replacing existing one when inplace_p is true.
3399         (expand_complex_multiplication_components): New function.
3400         (expand_complex_multiplication): Expand floating-point complex
3401         multiplication using the above.
3402         (expand_complex_division): Rename inner_type parameter to type.
3403         Update expand_complex_libcall call-site.
3404         (expand_complex_operations_1): Update expand_complex_multiplication
3405         and expand_complex_division call-sites.
3406
3407 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
3408
3409         PR target/85582
3410         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3411         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3412         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3413         the highest significant bit of the shift count mask is clear.  In
3414         check whether and[sq]i3 is needed verify that all significant bits
3415         of the shift count other than the highest are set.
3416
3417 2018-05-02  Tom de Vries  <tom@codesourcery.com>
3418
3419         PR libgomp/82428
3420         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3421         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3422         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3423         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3424         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3425         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3426         __builtin_goacc_parlevel_size.
3427
3428 2018-05-02  Richard Biener  <rguenther@suse.de>
3429
3430         PR tree-optimization/85597
3431         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3432         do not use split vect_get_vec_defs call but call vect_get_slp_defs
3433         directly.
3434
3435 2018-05-02  Tom de Vries  <tom@codesourcery.com>
3436
3437         PR testsuite/85106
3438         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3439         dump files): Add ltrans-tree.
3440
3441 2018-05-02  Tom de Vries  <tom@codesourcery.com>
3442
3443         PR testsuite/85106
3444         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3445         dump files): Add wpa-ipa.
3446
3447 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
3448
3449         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
3450         powerpc*-*-linux*paired* target.
3451         * config/rs6000/750cl.h: Delete.
3452         * config/rs6000/paired.h: Delete.
3453         * config/rs6000/paired.md: Delete.
3454         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3455         float support.
3456         * config/rs6000/rs6000-builtin.def: Remove paired float support.
3457         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3458         comment.  Remove paired float support.
3459         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3460         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3461         VECTOR_PAIRED.
3462         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3463         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3464         declarations.
3465         * config/rs6000/rs6000.c: Remove paired float support.
3466         (paired_expand_vector_init, paired_expand_vector_move,
3467         paired_emit_vector_compare, paired_emit_vector_cond_expr,
3468         (paired_expand_lv_builtin, paired_expand_stv_builtin,
3469         paired_expand_builtin, paired_expand_predicate_builtin,
3470         paired_init_builtins): Delete.
3471         * config/rs6000/rs6000.h: Remove paired float support.
3472         * config/rs6000/rs6000.md: Remove paired float support.
3473         (move_from_CR_ov_bit): Delete.
3474         * config/rs6000/rs6000.opt (mpaired): Delete.
3475         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3476         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3477
3478 2018-05-02  Richard Biener  <rguenther@suse.de>
3479
3480         PR middle-end/85567
3481         * gimplify.c (gimplify_save_expr): When in SSA form allow
3482         SAVE_EXPRs to compute to SSA vars.
3483
3484 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
3485
3486         PR target/85582
3487         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3488         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3489         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3490         clobber operands[2], instead use a new pseudo.  Formatting fixes.
3491
3492 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
3493
3494         PR tree-optimization/85586
3495         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3496         exit early for statements in the same group if the accesses are
3497         not strided.
3498
3499 2018-05-02  Tom de Vries  <tom@codesourcery.com>
3500
3501         PR lto/85451
3502         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3503         error message.
3504
3505 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
3506
3507         PR tree-optimization/85143
3508         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3509
3510 2018-05-01  Tom de Vries  <tom@codesourcery.com>
3511
3512         PR lto/85451
3513         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3514         not found" error message.
3515
3516 2018-05-01  Tom de Vries  <tom@codesourcery.com>
3517
3518         PR other/83786
3519         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3520         * vec.c (test_ordered_remove_if): New function.
3521         (vec_c_tests): Call test_ordered_remove_if.
3522         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3523         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3524         * tree-vect-patterns.c (vect_pattern_recog_1): Use
3525         VEC_ORDERED_REMOVE_IF.
3526
3527 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
3528
3529         PR tree-optimization/82665
3530         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3531         pointer subtraction where arguments come from a memchr call.
3532
3533 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
3534
3535         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3536         --push-state --as-needed and --pop-state instead of --as-needed and
3537         --no-as-needed if ld supports it.
3538         * configure: Regenerated.
3539
3540         PR web/85578
3541         * doc/install.texi2html: Replace _002d with - and _002a with * in
3542         generated html files using sed.
3543
3544 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
3545
3546         PR c++/85523
3547         * gcc-rich-location.c (blank_line_before_p): New function.
3548         (use_new_line): New function.
3549         (gcc_rich_location::add_fixit_insert_formatted): New function.
3550         * gcc-rich-location.h
3551         (gcc_rich_location::add_fixit_insert_formatted): New function.
3552
3553 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
3554
3555         * selftest.c (assert_streq): Rename "expected" and "actual" to
3556         "val1" and "val2".  Extend NULL-handling to cover both inputs
3557         symmetrically, while still requiring both to be non-NULL for a pass.
3558         * selftest.h (assert_streq): Rename "expected" and "actual" to
3559         "val1" and "val2".
3560         (ASSERT_EQ): Likewise.
3561         (ASSERT_EQ_AT): Likewise.
3562         (ASSERT_KNOWN_EQ): Likewise.
3563         (ASSERT_KNOWN_EQ_AT): Likewise.
3564         (ASSERT_NE): Likewise.
3565         (ASSERT_MAYBE_NE): Likewise.
3566         (ASSERT_MAYBE_NE_AT): Likewise.
3567         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
3568         the assertion to pass.
3569         (ASSERT_STREQ_AT): Likewise.
3570
3571 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
3572
3573         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3574         interaction with -pie.
3575
3576 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
3577
3578         * selftest.h: Fix alphabetization of per-source-file selftest
3579         declarations.
3580
3581 2018-04-30  Jason Merrill  <jason@redhat.com>
3582
3583         PR c++/61982 - dead stores to destroyed objects.
3584         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3585         of clobber.
3586
3587 2018-04-30  Jason Merrill  <jason@redhat.com>
3588
3589         * tree.c (build_clobber): New.
3590         * tree.h: Declare it.
3591         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3592
3593 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
3594
3595         * diagnostic-show-locus.c (layout::layout): Update for
3596         location_get_source_line returning a char_span.
3597         (struct char_span): Move to input.h.
3598         (struct correction): Update for fields in char_span becoming
3599         private.
3600         (struct source_line): Update for location_get_source_line
3601         returning a char_span.
3602         (layout::print_line): Likewise.
3603         * edit-context.c (edited_file::print_content): Likewise.
3604         (edited_file::print_diff_hunk): Likewise.
3605         (edited_file::print_run_of_changed_lines): Likewise.
3606         (edited_file::get_num_lines): Likewise.
3607         (edited_line::edited_line): Likewise.
3608         * final.c (asm_show_source): Likewise.
3609         * input.c (location_get_source_line): Convert return type
3610         from const char * to char_span, losing the final "line_len"
3611         param.
3612         (dump_location_info): Update for the above.
3613         (get_substring_ranges_for_loc): Likewise.  Use a char_span
3614         when handling the literal within the line.
3615         (test_reading_source_line): Update for location_get_source_line
3616         returning a char_span.
3617         * input.h (class char_span): Move here from
3618         diagnostic-show-locus.c, converting from a struct to a class.
3619         Make data members private.
3620         (char_span::operator bool): New.
3621         (char_span::length): New.
3622         (char_span::get_buffer): New.
3623         (char_span::operator[]): New.
3624         (char_span::subspan): Make const.
3625         (char_span::xstrdup): New.
3626         (location_get_source_line): Convert return type from const char *
3627         to char_span, losing the final "line_size" param.
3628
3629 2018-04-30  Jan Hubicka  <jh@suse.cz>
3630
3631         * lto-wrapper.c (ltrans_priorities): New static var.
3632         (cmp_priority): New.
3633         (run_gcc): Read priorities and if doing parallel build order
3634         the Makefile by them.
3635
3636 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
3637
3638         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3639
3640 2018-04-30  Richard Biener  <rguenther@suse.de>
3641
3642         * tree-cfg.c (verify_address): Remove base argument, add
3643         flag whether to check TREE_ADDRESSABLE and do that.
3644         (verify_expr): Remove.
3645         (verify_types_in_gimple_reference): Add pieces from verify_expr.
3646         (verify_gimple_assign_single): Likewise.
3647         (verify_gimple_switch): Likewise.
3648         (verify_expr_location_1): Dereference tp once.  Add (disabled)
3649         piece from verify_expr.
3650         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3651
3652 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
3653
3654         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3655
3656 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
3657
3658         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3659         (small_data_pattern): Likewise.
3660         (arc_rewrite_small_data): Likewise.
3661         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3662         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3663         (get_symbol_alignment): New function.
3664         (legitimate_small_data_address_p): Likewise.
3665         (legitimate_scaled_address): Update, call
3666         legitimate_small_data_address_p.
3667         (output_sdata): New static variable.
3668         (arc_print_operand): Update how we handle small data operands.
3669         (arc_print_operand_address): Likewise.
3670         (arc_legitimate_address_p): Update, use
3671         legitimate_small_data_address_p.
3672         (arc_rewrite_small_data_p): Remove.
3673         (arc_rewrite_small_data_1): Likewise.
3674         (arc_rewrite_small_data): Likewise.
3675         (small_data_pattern): Likewise.
3676         (compact_sda_memory_operand): Update to use
3677         legitimate_small_data_address_p and get_symbol_alignment.
3678         (prepare_move_operands): Don't rewite sdata pattern.
3679         (prepare_extend_operands): Remove.
3680         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3681         pattern.
3682         (zero_extendqisi2): Likewise.
3683         (zero_extendhisi2): Likewise.
3684         (extendqihi2): Likewise.
3685         (extendqisi2): Likewise.
3686         (extendhisi2): Likewise.
3687         (addsi3): Likewise.
3688         (subsi3): Likewise.
3689         (andsi3): Likewise.
3690         * config/arc/constraints.md (Usd): Change it to memory constraint.
3691
3692 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
3693
3694         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3695         as source of std instructions.
3696         * config/arc/arc.md (movsi_insn): Update pattern predicate to
3697         allow 6-bit constants as source for store instructions.
3698         (movdi_insn): Update instruction pattern to allow 6-bit constants
3699         as source for store instructions.
3700
3701 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
3702
3703         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3704
3705 2018-04-30  Nathan Sidwell  <nathan@acm.org>
3706             Sandra Loosemore <sandra@codesourcery.com>
3707
3708         * dumpfile.c (dump_open): Allow '-' for stdout.
3709         * doc/invoke.texi (Developer Options): Document dump filename
3710         determination early.  Document stdin/stdout selection.
3711
3712 2018-04-30 Andrew Sadek  <andrew.sadek.se@gmail.com>
3713
3714         Microblaze Target: PIC data text relative
3715
3716         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3717         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3718         Add declaration.
3719         * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3720         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3721         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3722         New addressing mode for data-text relative position indepenedent code.
3723         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3724         'ADDRESS_SYMBOLIC_TXT_REL'.
3725         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3726         (microblaze_legitimate_pic_operand): Exclude function calls from
3727         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3728         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3729         addresses cases.
3730         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3731         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3732         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3733         for 'address + offset'.
3734         (microblaze_expand_prologue): Add new function prologue call for
3735         'r20' assignation.
3736         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3737         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3738         table in case of TARGET_PIC_DATA_TEXT_REL.
3739         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3740         * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3741         Add new macros 'UNSPEC_TEXT',
3742         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3743         + exclude function calls from 'UNSPEC_PLT' in case of data text
3744         relative mode.
3745         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3746         new target hook for generating address diff vector tables in case of
3747         flag_pic.
3748         * doc/tm.texi : Regenerate.
3749         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3750         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3751         of addr diff vector generation.
3752         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3753         target hook definition.
3754         * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3755         Add default function for generate_pic_addr_diff_vec -> flag_pic.
3756         * doc/invoke.texi (Add new pic option): Add new microblaze pic
3757         option for data text relative.
3758
3759 2018-04-30  Richard Biener  <rguenther@suse.de>
3760
3761         * tree-chrec.h (evolution_function_is_constant_p): Remove
3762         redundant check.
3763         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3764
3765 2018-04-30  Richard Biener  <rguenther@suse.de>
3766
3767         PR bootstrap/85571
3768         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3769
3770 2018-04-30  Richard Biener  <rguenther@suse.de>
3771
3772         PR tree-optimization/28364
3773         PR tree-optimization/85275
3774         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3775         copying first exit test.
3776
3777 2018-04-28  Mark Wielaard  <mark@klomp.org>
3778
3779         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3780         dwarf_version >= 5.
3781         (dwarf_AT): Handle DW_AT_addr_base.
3782         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3783
3784 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
3785
3786         PR target/84431
3787         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3788         (*ashl<dwi>3_doubleword_mask_1): Ditto.
3789         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3790         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3791
3792 2018-04-28  Richard Biener  <rguenther@suse.de>
3793
3794         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3795         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3796         to reflect use.  Only add interesting stmts.
3797
3798 2018-04-27  Martin Jambor  <mjambor@suse.cz>
3799
3800         PR ipa/85549
3801         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3802         the jump function allows for passing through aggregate values.
3803
3804 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
3805
3806         * input.h (in_system_header_at): Convert from macro to inline
3807         function.
3808         (from_macro_expansion_at): Likewise.
3809         (from_macro_definition_at): Likewise.
3810
3811 2018-04-27  Jeff Law  <law@redhat.com>
3812
3813         * config.gcc: Mark tile* targets as deprecated/obsolete.
3814
3815 2018-04-27  Richard Biener  <rguenther@suse.de>
3816
3817         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3818         fix for ILP32.
3819
3820 2018-04-27  Richard Biener  <rguenther@suse.de>
3821
3822         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3823
3824 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
3825
3826         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3827         with Yd constraint. Set "preferred_for_speed" attribute from
3828         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3829         with Yd constraint.
3830         (*movdi_internal): Ditto.
3831         (movti_interunit splitters): Remove
3832         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3833         (movdi_interunit splitters): Ditto.
3834         * config/i386/constraints.md (Ye): Remove.
3835         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3836
3837 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3838
3839         PR target/85512
3840         * config/aarch64/constraints.md (Usg): Limit to 31.
3841         (Usj): Limit to 63.
3842
3843 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
3844
3845         PR tree-optimization/85529
3846         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3847         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3848         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3849         zero extension or masking of the MSB bit.
3850         (optimize_range_tests): Add FIRST_BB argument, pass it through
3851         to optimize_range_tests_var_bound.
3852         (maybe_optimize_range_tests, reassociate_bb): Adjust
3853         optimize_range_tests callers.
3854
3855 2018-04-26  Richard Biener  <rguenther@suse.de>
3856             Jakub Jelinek  <jakub@redhat.com>
3857
3858         * cgraph.h (symbol_table): Just declare debug method here.
3859         * symtab.c (symbol_table::debug): Define.
3860
3861 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
3862
3863         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3864
3865 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
3866
3867         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3868         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3869         (*movdi_internal): Substitute Yi and Yj constraint with x
3870         and Ym and Yn constraint with y constraint.  Update "isa"
3871         attribute and set "preferred_for_speed" attribute from
3872         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3873         (*movsi_internal): Ditto.
3874         (*movdf_internal): Ditto.
3875         (*movsf_internal): Ditto.
3876         (*zero_extendsidi2): Ditto.
3877         * config/i386/sse.md (vec_set<mode>_0): Ditto.
3878         (sse2_loadld): Ditto.
3879         (*vec_extract<ssevecmodelower>_0): Ditto.
3880         (*vec_extractv4si_0_zext_sse4): Ditto.
3881         (vec_concatv2di): Ditto.
3882         (*vec_dup<mode>): Ditto.
3883         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3884         * config/i386/constraints.md (Yi): Remove.
3885         (Yj): Remove.
3886         (Ym): Remove.
3887         (Yn): Remove.
3888
3889 2018-04-26  Nathan Sidwell  <nathan@acm.org>
3890
3891         * dumpfile.c (dump_open): New.
3892         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3893         (dump_finish): Detect stdio/stderr by value not name.
3894
3895 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
3896
3897         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3898
3899 2018-04-26  Tom de Vries  <tom@codesourcery.com>
3900
3901         PR target/84952
3902         * config/nvptx/nvptx.c (verify_neutering_jumps)
3903         (verify_neutering_labels): New function
3904         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3905
3906 2018-04-26  Tom de Vries  <tom@codesourcery.com>
3907
3908         PR target/84025
3909         * config/nvptx/nvptx.c (needs_neutering_p): New function.
3910         (nvptx_single): Use needs_neutering_p to skip over insns that do not
3911         need neutering.
3912
3913 2018-04-26  Richard Biener <rguenther@suse.de>
3914             Tom de Vries  <tom@codesourcery.com>
3915
3916         PR lto/85422
3917         * lto-streamer-out.c (output_function): Fixup loops if required to match
3918         discovery done in the reader.
3919
3920 2018-04-26  Richard Biener  <rguenther@suse.de>
3921
3922         PR tree-optimization/85116
3923         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3924         have a loop exit from the single latch predecessor.  Remove
3925         case of header with just condition.
3926         (ch_base::copy_headers): Exclude infinite loops from any
3927         processing.
3928         (pass_ch::execute): Record exits.
3929
3930 2018-04-26  Richard Biener  <rguenther@suse.de>
3931
3932         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3933         prologue cost vector and pass it to vect_get_load_cost.
3934         (vect_get_peeling_costs_all_drs): Likewise.
3935         (vect_peeling_hash_get_lowest_cost): Likewise.
3936         (vect_enhance_data_refs_alignment): Likewise.
3937
3938 2018-04-26  Richard Biener  <rguenther@suse.de>
3939
3940         PR middle-end/85450
3941         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3942         checking of integer<->pointer conversions.
3943         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3944         sign-/zero-extending pointer types.
3945         (expand_omp_for_static_chunk): Likewise.
3946
3947 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
3948             Jean Lee  <xiaoyur347@gmail.com>
3949
3950         * config/mips/mips.c (mips_asan_shadow_offset): New function.
3951         (TARGET_ASAN_SHADOW_OFFSET): Define.
3952         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3953         true for -fsanitize=address.
3954
3955 2018-04-25  Mark Wielaard  <mark@klomp.org>
3956
3957         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3958         shorter ones.
3959
3960 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
3961
3962         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3963         than "alu", remove explicit "memory" and "imm_disp" attributes.
3964         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3965
3966         PR middle-end/85414
3967         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3968         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3969         gen_lowpart_no_emit.
3970
3971 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
3972
3973         PR target/85473
3974         * config/i386/i386.c (ix86_expand_builtin): Change memory
3975         operand to XI, extend p0 to Pmode.
3976         * config/i386/i386.md: Change unspec volatile and operand
3977         1 mode to XI, change operand 0 mode to P.
3978
3979 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
3980
3981         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3982         GET_MODE_MASK before any checking.
3983         (nds32_can_use_bset_p): Likewise.
3984         (nds32_can_use_btgl_p): Likewise.
3985
3986 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
3987
3988         * config/nds32/nds32-doubleword.md: New define_split pattern for
3989         illegal register number.
3990
3991 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
3992
3993         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3994
3995 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
3996
3997         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3998
3999 2018-04-25  Richard Biener  <rguenther@suse.de>
4000
4001         * lto-streamer.h (LTO_major_version): Bump to 8.
4002
4003 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
4004
4005         * BASE-VER: Set to 9.0.0.
4006
4007 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
4008
4009         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
4010         in __abskf2 and __powikf2.
4011
4012 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4013
4014         PR target/85512
4015         * config/aarch64/constraints.md (Usg, Usj): New constraints.
4016         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
4017         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
4018         Use the above on operand 2.  Reindent.
4019         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
4020
4021 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
4022
4023         PR target/85485
4024         * common/config/i386/i386-common.c (ix86_handle_option): Don't
4025         handle OPT_mcet.
4026         * config/i386/i386.opt (mcet): Removed.
4027         * doc/install.texi: Remove -mcet documentation.
4028         * doc/invoke.texi: Likewise.
4029
4030 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
4031
4032         PR target/85485
4033         * doc/install.texi: Remove -mcet from bootstrap-cet.
4034
4035 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
4036
4037         PR target/85511
4038         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
4039         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
4040         if TARGET_64BIT.
4041
4042         PR target/85503
4043         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
4044         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
4045         containing a CONST_VECTOR.
4046
4047 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
4048
4049         * doc/install.texi: Update newlib dependency for nvptx.
4050
4051 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
4052
4053         PR target/85508
4054         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
4055         instead of INTVAL when shifting x left.
4056
4057 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
4058
4059         PR tree-optimization/85478
4060         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
4061         vect_grouped_store_supported for single element vectors.
4062
4063 2018-04-24  Richard Biener  <rguenther@suse.de>
4064
4065         PR target/85491
4066         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
4067         load cost increase to the case of non-constant step.
4068
4069 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
4070
4071         PR target/84828
4072         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
4073         destination if any_malformed_asm.
4074
4075 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
4076
4077         PR middle-end/85496
4078         * expr.c (store_field): In the bitfield case, if the value comes from
4079         a function call and is returned in registers by means of a PARALLEL,
4080         do not change the mode of the temporary unless BLKmode and VOIDmode.
4081
4082 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
4083
4084         PR rtl-optimization/85423
4085         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
4086         dependencies to debug insns when the previous insn is non-debug.
4087
4088 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
4089
4090         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
4091         enums into a single definition.
4092         (fls): Fix predicates and printing.
4093         (seti): Likewise.
4094
4095 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
4096
4097         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
4098         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
4099         and short u6 immediate.
4100         (check_if_valid_sleep_operand): Remove.
4101         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
4102
4103 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
4104
4105         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
4106         flag_always_save_lp condition.
4107         * config/nds32/nds32.opt (malways-save-lp): New option.
4108
4109 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
4110
4111         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
4112         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
4113         * config/nds32/nds32.h
4114         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
4115         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
4116
4117 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
4118
4119         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
4120         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
4121
4122 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
4123             Chung-Ju Wu  <jasonwucj@gmail.com>
4124
4125         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
4126         Declare.
4127         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
4128         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
4129
4130 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
4131
4132         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
4133
4134 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
4135
4136         * config/nds32/nds32-protos.h (nds32_data_alignment,
4137         nds32_local_alignment): Declare.
4138         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
4139         nds32_local_alignment): New functions.
4140         (TARGET_CONSTANT_ALIGNMENT): Define.
4141         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
4142
4143 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
4144
4145         * config/nds32/nds32.c
4146         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
4147         (TARGET_MODES_TIEABLE_P): Likewise.
4148
4149 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
4150
4151         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
4152         level Ofast and Og.
4153
4154 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
4155             Chung-Ju Wu  <jasonwucj@gmail.com>
4156
4157         * config/nds32/constants.md (unspec_volatile_element): Add enum values
4158         for unaligned access.
4159         * config/nds32/nds32-intrinsic.c: Implementation of expanding
4160         unaligned access.
4161         * config/nds32/nds32-intrinsic.md: Likewise.
4162         * config/nds32/nds32_intrinsic.h: Likewise.
4163         * config/nds32/nds32.h (nds32_builtins): Likewise.
4164         * config/nds32/nds32.opt (munaligned-access): New option.
4165         * config/nds32/nds32.c (nds32_asm_file_start): Display
4166         flag_unaligned_access status.
4167
4168 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
4169
4170         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
4171         -mno-relax is present.
4172         * config/riscv/linux.h (LINK_SPEC): Ditto.
4173
4174 2018-04-20  Martin Sebor  <msebor@redhat.com>
4175
4176         PR c/85365
4177         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
4178         for null pointers.
4179         (gimple_fold_builtin_stxcpy_chk): Same.
4180         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
4181
4182 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
4183
4184         PR target/85456
4185         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
4186         __powikf2 when long double is IEEE 128-bit.
4187
4188 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
4189
4190         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4191         step to make sure stack always aligned.
4192
4193 2018-04-20  Carl Love  <cel@us.ibm.com>
4194
4195         PR target/83402
4196         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
4197         size check for arg0.
4198
4199 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
4200             Tom de Vries  <tom@codesourcery.com>
4201
4202         PR target/85445
4203         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
4204         Emit insns for calls too.
4205         (nvptx_find_par): Always look for worker-level predecessor insn.
4206         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
4207         calls.
4208         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
4209         (nvptx_process_pars): Propagate frames for calls.
4210
4211 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
4212
4213         PR target/85469
4214         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
4215         Removed.
4216         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
4217         (ix86_handle_option): Don't handle OPT_mibt.
4218         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
4219         __SHSTK__.
4220         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4221         has_ibt and ibt.
4222         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4223         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4224         (ix86_target_macros): Define __CET__ with flag_cf_protection
4225         for -fcf-protection.
4226         * config/i386/i386.c (isa2_opts): Remove -mibt.
4227         * config/i386/i386.h (TARGET_IBT): Removed.
4228         (TARGET_IBT_P): Likewise.
4229         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4230         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4231         * config/i386/i386.opt (mcet): Update help message.
4232         (mshstk): Likewise.
4233         (mibt): Removed.
4234         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
4235         -mcet as an alias for -mshstk.
4236
4237 2018-04-20  Richard Biener <rguenther@suse.de>
4238
4239         PR middle-end/85475
4240         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4241         complexity by forcing a single use of the multiply operand.
4242
4243 2018-04-20  Martin Jambor  <mjambor@suse.cz>
4244
4245         ipa/85449
4246         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4247         recursion dependency to only apply to non-clones.
4248
4249 2018-04-20  Martin Jambor  <mjambor@suse.cz>
4250
4251         ipa/85447
4252         * ipa-cp.c (create_specialized_node): Check that clones of
4253         self-recursive edges exist during IPA-CP.
4254
4255 2018-04-19  Toon Moene  <toon@moene.org>
4256
4257         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4258         by -O3.
4259
4260 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
4261
4262         PR tree-optimization/85467
4263         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4264         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
4265         VECTOR_CST element to type.
4266
4267 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
4268
4269         PR target/85397
4270         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4271         * config/i386/i386.md (builtin_setjmp_setup): Removed.
4272         (builtin_longjmp): Likewise.
4273         (save_stack_nonlocal): New pattern.
4274         (restore_stack_nonlocal): Likewise.
4275
4276 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
4277
4278         PR target/85404
4279         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4280         Replace ASM_OUTPUT_LABEL with fprintf.
4281
4282 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
4283
4284         PR target/85417
4285         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4286         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4287         * config/i386/i386-c.c (ix86_target_macros_internal): Also
4288         define __IBT__ and __SHSTK__ for -fcf-protection.
4289         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4290         TARGET_IBT.
4291         (ix86_trampoline_init): Likewise.
4292         (x86_output_mi_thunk): Likewise.
4293         (ix86_notrack_prefixed_insn_p): Likewise.
4294         (ix86_option_override_internal): Don't disallow -fcf-protection.
4295         * config/i386/i386.md (rdssp<mode>): Also enable for
4296         -fcf-protection.
4297         (incssp<mode>): Likewise.
4298         (nop_endbr): Likewise.
4299         * config/i386/i386.opt (mcet): Change help message to built-in
4300         functions only.
4301         (mibt): Likewise.
4302         (mshstk): Likewise.
4303         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4304         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
4305         enable CET built-in functions.
4306
4307 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
4308
4309         * common/config/i386/i386-common.c
4310         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4311         OPTION_MASK_ISA_MOVDIRI_UNSET,
4312         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4313         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4314         * config.gcc (movdirintrin.h): New header.
4315         * config/i386/cpuid.h (bit_MOVDIRI,
4316         bit_MOVDIR64B): New bits.
4317         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4318         and -mmvodir64b.
4319         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4320         (VOID, PVOID, PCVOID)): New function types.
4321         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4322         __builtin_ia32_directstoreu_u64,
4323         __builtin_ia32_movdir64b): New builtins.
4324         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4325         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4326         and -mmovdiri.
4327         (ix86_valid_target_attribute_inner_p): Ditto.
4328         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4329         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4330         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4331         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4332         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4333         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4334         (movdiri<mode>, movdir64b_<mode>): New.
4335         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4336         * config/i386/immintrin.h: Include movdirintrin.h.
4337         * config/i386/movdirintrin.h: New file.
4338         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4339
4340 2018-04-19  Richard Biener  <rguenther@suse.de>
4341
4342         PR middle-end/85455
4343         * cfg.c (clear_bb_flags): When loop state says we have
4344         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4345
4346 2018-04-19  Richard Biener  <rguenther@suse.de>
4347
4348         PR tree-optimization/84737
4349         * tree-vect-data-refs.c (vect_copy_ref_info): New function
4350         copying restrict info.
4351         (vect_setup_realignment): Use it.
4352         * tree-vectorizer.h (vect_copy_ref_info): Declare.
4353         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4354         the first DR to all generated stores.
4355         (vectorizable_load): Likewise for loads.
4356
4357 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
4358
4359         PR tree-optimization/85446
4360         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4361         the integral and pointer types to have the same precision.
4362
4363         * doc/install.texi: Document --disable-cet being the default and
4364         --enable-cet=auto.
4365
4366 2018-04-18  Martin Liska  <mliska@suse.cz>
4367
4368         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4369         style.
4370
4371 2018-04-18  Martin Liska  <mliska@suse.cz>
4372
4373         Revert
4374         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
4375
4376         PR ipa/83983
4377         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4378         arguments if they are comparable.
4379
4380 2018-04-18  Martin Liska  <mliska@suse.cz>
4381
4382         Revert
4383         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
4384
4385         PR lto/84805
4386         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4387         incomplete types.
4388
4389 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
4390
4391         PR target/85388
4392         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4393         ENDBR after calling __morestack.
4394
4395 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
4396
4397         PR jit/85384
4398         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4399         by using gcc_base_ver to generate a gcc_driver_version, and use
4400         it when generating GCC_DRIVER_NAME.
4401         * configure: Regenerate.
4402
4403 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
4404
4405         PR target/81084
4406         * config.gcc: Obsolete powerpc*-*-*spe*.
4407
4408 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
4409
4410         PR debug/84637
4411         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4412         (stabstr_D): Change type of unum from unsigned int to
4413         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
4414         type.
4415
4416 2018-04-17  Jim Wilson  <jimw@sifive.com>
4417
4418         PR 84856
4419         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4420         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4421         Set arg_pointer_offset after using pretend_args_size.
4422
4423 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
4424
4425         PR rtl-optimization/85431
4426         * dse.c (record_store): Ignore zero width stores.
4427
4428         PR sanitizer/85230
4429         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
4430         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4431         __builtin_stack_restore rather than after it.
4432         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4433         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4434         argument instead of virtual_dynamic_stack_rtx.
4435
4436 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
4437
4438         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4439         New prototype.
4440         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4441         Add note to error message to explain internal mapping of overloaded
4442         built-in function name to non-overloaded built-in function name.
4443         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4444         function.
4445
4446 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
4447
4448         PR target/85424
4449         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4450         where the inputs overlap with the output.
4451
4452 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
4453
4454         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4455         (=v, v) alternative and explicit "memory" attribute.
4456         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
4457         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4458         attributes.
4459         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4460         "sselog1" type instead of "sselog".
4461         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4462         "sselog".  Remove explicit "memory" attribute.
4463         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4464         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4465         attributes.
4466         (vec_extract_hi_v32hi): Merge all alternatives into one, use
4467         "sselog1" type instead of "sselog".  Remove explicit "memory"
4468         attribute.
4469         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4470         use "sselog1" type instead of "sselog".  Remove explicit "memory"
4471         attribute.
4472         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4473         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4474         attributes.
4475         (vec_extract_hi_v64qi): Merge all alternatives into one, use
4476         "sselog1" type instead of "sselog".  Remove explicit "memory"
4477         attribute.
4478         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4479         use "sselog1" type instead of "sselog".  Remove explicit "memory"
4480         attribute.
4481
4482         PR target/85430
4483         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4484
4485         PR middle-end/85414
4486         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4487         on a SUBREG.
4488
4489 2018-04-17  Martin Jambor  <mjambor@suse.cz>
4490
4491         PR ipa/85421
4492         * ipa-cp.c (create_specialized_node): Call
4493         expand_all_artificial_thunks if necessary.
4494
4495 2018-04-17  Martin Liska  <mliska@suse.cz>
4496
4497         PR lto/85405
4498         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4499         in message, remote space in between '_G' and '('.
4500
4501 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
4502
4503         PR target/85281
4504         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4505         avx512f_vmcmp<mode>3<round_saeonly_name>,
4506         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4507         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4508         avx512f_rndscale<mode><round_saeonly_name>,
4509         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4510         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4511         Use %<iptr>2 instead of %2 for -masm=intel.
4512         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4513         avx512f_vcvttss2usi<round_saeonly_name>,
4514         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4515         -masm=intel.
4516         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4517         avx512f_vcvttsd2usi<round_saeonly_name>,
4518         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4519         Use %q1 instead of %1 for -masm=intel.
4520         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4521         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4522         of %3 for -masm=intel.
4523         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4524         -masm=intel.
4525         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4526         -masm=intel.
4527         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4528         -masm=intel.
4529         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4530         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4531         %g1.
4532         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4533         -masm=intel.
4534         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4535         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4536         %g1 and one with %0 and %1.
4537         (avx512er_vmrcp28<mode><round_saeonly_name>,
4538         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4539         %1 for -masm=intel.
4540         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4541         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4542         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4543         of %0 and %{%4%} for -masm=intel.
4544         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4545         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4546         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4547         order of %0 and %{%5%}%{z%} for -masm=intel.
4548
4549 2018-04-17  Jan Hubicka  <jh@suse.cz>
4550
4551         PR lto/85405
4552         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4553
4554 2018-04-17  Martin Liska  <mliska@suse.cz>
4555
4556         PR ipa/85329
4557         * multiple_target.c (create_dispatcher_calls): Set apostrophes
4558         for target_clone error message.  Make default implementation
4559         clone to be a local declaration.
4560         (separate_attrs): Add new argument and check for an empty
4561         string.
4562         (expand_target_clones): Handle it.
4563         (ipa_target_clone): Make redirection just for target_clones
4564         functions.
4565
4566 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
4567             Tom de Vries  <tom@codesourcery.com>
4568
4569         PR middle-end/84955
4570         * omp-expand.c (expand_oacc_for): Add dummy false branch for
4571         tiled basic blocks without omp continue statements.
4572
4573 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
4574
4575         PR target/83660
4576         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4577         vec_extract expression as having side effects to make sure it gets
4578         a cleanup point.
4579
4580 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
4581
4582         PR target/85403
4583         * config/i386/i386.c (get_builtin_code_for_version): Check
4584         error_mark_node.
4585
4586 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
4587
4588         PR target/84331
4589         * gcc/config.gcc: Support "skylake".
4590         * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4591         PROCESSOR_SKYLAKE.
4592         * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4593         (processor_target_table): Add "skylake".
4594         (ix86_option_override_internal): Add "skylake".
4595         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4596         PROCESSOR_CANNONLAKE.
4597         (get_builtin_code_for_version): Fix priority for
4598         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4599         PROCESSOR_SKYLAKE-AVX512.
4600         * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4601         (processor_type): Add PROCESSOR_SKYLAKE.
4602
4603 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
4604             Jason Merrill  <jason@redhat.com>
4605
4606         PR c++/85112
4607         * convert.c (convert_to_integer_1): Use direct recursion for
4608         enumeral types and types with a precision less than the number
4609         of bits in their mode.
4610
4611 2018-04-16  Julia Koval  <julia.koval@intel.com>
4612
4613         PR target/84413
4614         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4615         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4616
4617 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
4618
4619         PR target/85293
4620         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4621         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4622         and -mno-direct-move.
4623
4624 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
4625
4626         PR target/83402
4627         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4628         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4629         Ensure negative shifts result in {0}.
4630
4631 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
4632
4633         PR rtl-optimization/79916
4634         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4635         regs (if any) to define how to gnerate SD moves when LRA is in
4636         progress.
4637
4638 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
4639
4640         PR rtl-optimization/85393
4641         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4642         * except.c (expand_dw2_landing_pad_for_region): Make static.
4643         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4644         a label and unconditional jump to old_bb, rather than
4645         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4646         basic block.
4647
4648         PR rtl-optimization/85376
4649         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4650         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4651         instead of a specific value.
4652
4653 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
4654             Bin Cheng  <bin.cheng@arm.com>
4655
4656         PR tree-optimization/82965
4657         PR tree-optimization/83991
4658         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4659         by_profile_only parameter.
4660         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4661         information if the loop was predicted to iterate too many times.
4662         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4663
4664 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
4665
4666         PR lto/71991
4667         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4668         always inline.
4669
4670 2018-04-13  Martin Liska  <mliska@suse.cz>
4671             Jakub Jelinek  <jakub@redhat.com>
4672
4673         PR middle-end/81657
4674         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4675         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4676         * builtins.c (expand_builtin_memory_copy_args): Use
4677         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4678         handle dest_addr == pc_rtx.
4679
4680 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
4681
4682         PR target/85291
4683         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4684         asked to not generate direct moves.
4685         (fix_trunc<mode>si2_stfiwx): Similar.
4686         (fix_trunc<mode>si2_internal): Similar.
4687
4688 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
4689
4690         PR debug/83157
4691         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4692         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4693         lookup if dest in some wider mode is known to be const0_rtx and
4694         if so, record permanent equivalence for it to be ZERO_EXTEND of
4695         the narrower mode destination.
4696
4697 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
4698
4699         * lto-streamer-out.c (output_function): Revert 259346.
4700         * omp-expand.c (expand_oacc_for): Likewise.
4701
4702 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
4703
4704         PR rtl-optimization/85354
4705         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4706         * sel-sched.c (sel_global_init): ... here.
4707
4708 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
4709
4710         PR target/85238
4711         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4712         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4713         mode for PE-COFF targets.
4714         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4715         (i386_pe_asm_lto_end): Likewise.
4716         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4717         (TARGET_ASM_LTO_END): Likewise.
4718         * config/i386/winnt.c (saved_debug_info_level): New static variable.
4719         (i386_pe_asm_lto_start): New function.
4720         (i386_pe_asm_lto_end): Likewise.
4721
4722 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
4723             Richard Biener  <rguenther@suse.de>
4724
4725         PR middle-end/84955
4726         * lto-streamer-out.c (output_function): Fix CFG loop state before
4727         streaming out.
4728         * omp-expand.c (expand_oacc_for): Handle calls to internal
4729         functions like regular functions.
4730
4731 2018-04-12  Richard Biener  <rguenther@suse.de>
4732
4733         PR lto/85371
4734         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4735         for the early LTO debug to properly generate references to it
4736         during DIE emission.  Do not re-use that for the skeleton for
4737         split-dwarf.
4738         (dwarf2out_early_finish): Likewise.
4739
4740 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
4741
4742         PR target/85328
4743         * config/i386/sse.md
4744         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4745         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4746         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4747         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4748         and output is a reg, avoid creating invalid lowpart subreg, but
4749         instead split into a 512-bit move.  Don't split if not AVX512VL,
4750         input is xmm16+ reg and output is a mem.
4751         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4752         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4753         xmm16+ reg and output is a mem.
4754
4755 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
4756
4757         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4758         also for flag_dwarf2_cfi_asm.
4759
4760 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
4761
4762         PR rtl-optimization/85342
4763         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4764         a bool scalar var inside of the loop instead.  Don't try to update
4765         recog_data.operand after failed apply_change_group.
4766
4767 2018-04-12  Tom de Vries  <tom@codesourcery.com>
4768
4769         PR target/85296
4770         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4771         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
4772         array with flexible array member as array without given dimension.
4773         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4774         argument for undefined param to true.
4775
4776 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
4777
4778         PR target/85321
4779         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4780         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4781         from PowerPC section.
4782         * config/rs6000/sysv4.opt (mcall-): Improve help text.
4783         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4784         help text that is too long.
4785         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4786         help text that is too long.
4787         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4788         help text that is too long.
4789
4790 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
4791
4792         * config/alpha/alpha.md (stack_probe_internal): Rename
4793         from "probe_stack".  Update all callers.
4794
4795 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
4796
4797         PR rtl-optimization/84566
4798         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4799         sched_macro_fuse_insns.
4800
4801 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
4802
4803         PR target/84301
4804         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4805         (compute_block_dependences): ... from here.
4806
4807 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
4808
4809         PR tree-optimization/85331
4810         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4811         from int to HOST_WIDE_INT.
4812
4813 2018-04-11  Martin Jambor  <mjambor@suse.cz>
4814
4815         PR ipa/84149
4816         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4817         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4818         not the same as the source val.
4819         (cgraph_edge_brings_value_p): New parameter.
4820         (gather_edges_for_value): Pass destination value to
4821         cgraph_edge_brings_value_p.
4822         (perhaps_add_new_callers): Likewise.
4823         (get_info_about_necessary_edges): Likewise and exclude values brought
4824         only by self-recursive edges.
4825         (create_specialized_node): Redirect only clones of self-calling edges.
4826         (+self_recursive_pass_through_p): New function.
4827         (find_more_scalar_values_for_callers_subset): Use it.
4828         (find_aggregate_values_for_callers_subset): Likewise.
4829         (known_aggs_to_agg_replacement_list): Removed.
4830         (decide_whether_version_node): Re-calculate known constants for all
4831         remaining context clones.
4832
4833 2018-04-11  Richard Biener  <rguenther@suse.de>
4834
4835         PR lto/85339
4836         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4837         from early DWARF output.
4838         (dwarf2out_early_finish): Output line info unconditionally into
4839         early DWARF and add reference to it.
4840
4841 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
4842
4843         PR target/85281
4844         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4845         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4846         other than V2DFmode using iptr mode attribute.
4847         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4848
4849 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
4850
4851         PR rtl-optimization/84659
4852         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4853
4854 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
4855
4856         PR debug/85302
4857         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4858         SIZEP is NULL.
4859         (output_loc_list): Pass address of a dummy size variable even in the
4860         locview handling loop.
4861         (index_location_lists): Add comment on why skip_loc_list_entry can't
4862         call size_of_locs.
4863
4864 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
4865
4866         PR target/85261
4867         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4868         into register.
4869
4870 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
4871
4872         PR target/85321
4873         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4874         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4875         and -mstring-compare-inline-limit.
4876
4877 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
4878
4879         PR target/85287
4880         * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4881         for stack clash protection in a register whenever we need it to be in
4882         a register.
4883
4884 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
4885
4886         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4887         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4888
4889 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
4890
4891         PR target/85321
4892         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4893         the help text.
4894         (mlong-double-): Ditto.
4895         * config/rs6000/sysv4.opt (msdata=): Ditto.
4896         (mtls-size=): Ditto.
4897
4898 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
4899
4900         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4901         erroneous entries for
4902         "vector int vec_ldl (int, long int *)", and
4903         "vector unsigned int vec_ldl (int, unsigned long int *)".
4904         Add comments and entries for
4905         "vector bool char vec_ldl (int, bool char *)",
4906         "vector bool short vec_ldl (int, bool short *)",
4907         "vector bool int vec_ldl (int, bool int *)",
4908         "vector bool long long vec_ldl (int, bool long long *)",
4909         "vector pixel vec_ldl (int, pixel *)",
4910         "vector long long vec_ldl (int, long long *)",
4911         "vector unsigned long long vec_ldl (int, unsigned long long *)".
4912         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4913         type tree bool_long_long_type_node and correct definition of
4914         bool_V2DI_type_node to make reference to this new type tree.
4915         (rs6000_mangle_type): Replace erroneous reference to
4916         bool_long_type_node with bool_long_long_type_node.
4917         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4918         comments to emphasize sign distinctions for char and int types and
4919         replace RS6000_BTI_bool_long constant with
4920         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
4921         use of RS6000_BTI_pixel.
4922         (bool_long_type_node): Remove this macro definition.
4923         (bool_long_long_type_node): New macro definition
4924
4925 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
4926
4927         PR rtl-optimization/85300
4928         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4929         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4930         simplify_unary_operation fails.
4931
4932 2018-04-10  Martin Liska  <mliska@suse.cz>
4933
4934         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4935         cgraph_edge and ipa_ref.
4936
4937 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
4938
4939         PR target/85177
4940         PR target/85255
4941         * config/i386/sse.md
4942         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4943         computation of the VEC_MERGE selector from mask.
4944         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4945         Fix decoding of the VEC_MERGE selector into mask.
4946
4947 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
4948
4949         PR tree-optimization/85286
4950         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4951
4952 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
4953
4954         * final.c (final_1): Set insn_last_address as well as
4955         insn_current_address.
4956
4957 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4958
4959         PR target/85173
4960         * explow.c (emit_stack_probe): Call validize_mem on memory location
4961         before passing it to gen_probe_stack.  Create address operand and
4962         legitimize it for the probe_stack_address case.
4963
4964 2018-04-09  Jan Hubicka  <jh@suse.cz>
4965
4966         PR lto/85078
4967         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4968         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4969         * tree.c (free_lang_data_in_type): Fix handling of binfos;
4970         walk basetypes.
4971         (free_lang_data): Rebuild type inheritance graph.
4972
4973 2018-04-09  Martin Sebor  <msebor@redhat.com>
4974
4975         * invoke.texi (-finline-small-functions): Mention other optimization
4976         options.
4977         (-findirect-inlining, -fpartial-inlining): Same.
4978         (-finline-functions-called-once): Same.
4979         (-freorder-blocks-and-partition): Same.
4980
4981 2018-04-09  Jan Hubicka  <jh@suse.cz>
4982
4983         PR rtl/84058
4984         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4985         jumps; choose last target that matches the criteria (i.e.
4986         no partition changes for non-crossing jumps).
4987         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4988         support for redirecting crossing jumps to non-crossing.
4989
4990 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
4991
4992         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4993         also for naked functions.
4994
4995 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
4996
4997         * config/arc/arc.md (add_shift): New pattern.
4998         (add_shift2): Likewise.
4999         (sub_shift): Likewise.
5000         (sub_shift_cmp0_noout): Likewise.
5001         (compare_si_ashiftsi): Likewise.
5002         (xbfu_cmp0_noout): New combine pattern.
5003         (xbfu_cmp0"): Likewise.
5004         (movsi_set_cc_insn): Place the predicable variant first.
5005         (commutative_binary_cmp0_noout): Remove clobber.
5006         (commutative_binary_cmp0): New pattern.
5007         (noncommutative_binary_cmp0): Likewise.
5008         (noncommutative_binary_cmp0_noout): Likewise.
5009         (noncommutative_binary_comparison_result_used): Removed.
5010         (rsub_cmp0): New pattern.
5011         (rsub_cmp0_noout): Likewise.
5012         (extzvsi): Changed, keep only meaningful variants.
5013         (SQH, SEZ): New iterators.
5014         (SQH_postfix): New mode attribute.
5015         (SEZ_prefix): New code attribute.
5016         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
5017         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
5018         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
5019         of numerical value.
5020         (noncommutative_operator): Check the availability of barrel
5021         shifter option.
5022
5023 2018-04-09  Richard Biener  <rguenther@suse.de>
5024
5025         PR tree-optimization/85284
5026         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
5027         Only use the niter constraining form of simple_iv when the exit
5028         is always executed.
5029
5030 2018-04-09  Tom de Vries  <tom@codesourcery.com>
5031
5032         PR target/84041
5033         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5034         (define_expand "*memory_barrier"): New define_expand.
5035         (define_insn "memory_barrier"): New insn.
5036
5037 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
5038
5039         PR rtl-optimization/80463
5040         PR rtl-optimization/83972
5041         PR rtl-optimization/83480
5042
5043         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
5044         correct producer for the insn.
5045         (tidy_control_flow): Fixup seqnos in case of debug insns.
5046
5047 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
5048
5049         PR rtl-optimization/83913
5050
5051         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
5052         different sched-times when merging exprs.
5053
5054 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
5055
5056         PR rtl-optimization/83962
5057
5058         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
5059         tidy_fallthru_edge and tidy_control_flow.
5060
5061 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
5062
5063         PR rtl-optimization/83530
5064
5065         * sel-sched.c (force_next_insn): New global variable.
5066         (remove_insn_for_debug): When force_next_insn is true, also leave only
5067         next insn in the ready list.
5068         (sel_sched_region): When the region wasn't scheduled, make another pass
5069         over it with force_next_insn set to 1.
5070
5071 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
5072
5073         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
5074         into tm_file.
5075         * config/nds32/constants.md (unspec_volatile_element): Add enum values
5076         for interrupt control.
5077         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5078         functions for interrupt control.
5079         * config/nds32/nds32-intrinsic.md: Likewise.
5080         * config/nds32/nds32_intrinsic.h: Likewise.
5081         * config/nds32/nds32.h (nds32_builtins): Likewise.
5082
5083 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
5084
5085         * config/nds32/nds32.c (nds32_init_machine_status,
5086         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
5087         strict_aligned_p field.
5088         (nds32_expand_to_rtl_hook): New function.
5089         (TARGET_EXPAND_TO_RTL_HOOK): Define.
5090         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
5091
5092 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
5093             Chung-Ju Wu  <jasonwucj@gmail.com>
5094
5095         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
5096         * config/nds32/nds32-n7.md: New file.
5097         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
5098         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
5099         pipeline.
5100         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
5101         * config/nds32/nds32.md (pipeline_model): Add n7.
5102         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
5103         * config/nds32/pipelines.md: Include n7 settings.
5104
5105 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
5106             Chung-Ju Wu  <jasonwucj@gmail.com>
5107
5108         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
5109         * config/nds32/nds32-e8.md: New file.
5110         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
5111         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
5112         pipeline.
5113         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
5114         * config/nds32/nds32.md (pipeline_model): Add e8.
5115         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
5116         * config/nds32/pipelines.md: Include e8 settings.
5117
5118 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
5119             Chung-Ju Wu  <jasonwucj@gmail.com>
5120
5121         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
5122         * config/nds32/nds32-n8.md: New file.
5123         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
5124         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
5125         pipeline.
5126         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
5127         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
5128         * config/nds32/nds32.md (pipeline_model): Add n8.
5129         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
5130         * config/nds32/pipelines.md: Include n8 settings.
5131
5132 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
5133             Chung-Ju Wu  <jasonwucj@gmail.com>
5134
5135         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
5136         * config/nds32/nds32-n9-2r1w.md: New file.
5137         * config/nds32/nds32-n9-3r2w.md: New file.
5138         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
5139         nds32_register_ports): New or modify for cpu n9.
5140         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
5141         pipeline.
5142         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
5143         * config/nds32/nds32-utils.c: New file.
5144         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
5145         TARGET_MUL_SLOW): Define.
5146         * config/nds32/nds32.md (pipeline_model): New attribute.
5147         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
5148         New options that support cpu n9.
5149         * config/nds32/pipelines.md: Include n9 settings.
5150         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
5151
5152 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
5153
5154         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
5155         information if necessary.
5156         (output_cond_branch_compare_zero): Likewise.
5157         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
5158         (nds32_target_alignment): Refine for alignment.
5159         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
5160         (FUNCTION_BOUNDARY): Modify.
5161         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
5162         align case.
5163         * config/nds32/nds32.opt (malways-align, malign-functions): New.
5164
5165 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
5166
5167         * config/nds32/constants.md (unspec_volatile_element): Add values for
5168         TLB operation and data prefetch.
5169         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5170         functions for TLB operation and data prefetch.
5171         * config/nds32/nds32-intrinsic.md: Likewise.
5172         * config/nds32/nds32_intrinsic.h: Likewise.
5173         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
5174         (nds32_print_operand): Likewise.
5175         * config/nds32/nds32.h (nds32_builtins): Likewise.
5176
5177 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
5178         Andrew Pinski <pinsika@gcc.gnu.org>
5179
5180         PR middle-end/82976
5181         * match.pd: Use constant_boolean_node of correct type instead of
5182         boolean_true_node or boolean_false_node for simplifying
5183         pointer comparisons to zero.
5184
5185 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
5186
5187         PR tree-optimization/80021
5188         * tree.c (verify_type_variant): Make error call in verify_variant_match
5189         translatable and remove final full stop.
5190
5191 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
5192
5193         * config/nds32/constants.md (unspec_volatile_element): Add
5194         UNSPEC_VOLATILE_EH_RETURN.
5195         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
5196         nds32_output_stack_pop): Support dwarf exception handling process.
5197         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
5198         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
5199         exception handling process.
5200         (nds32_compute_stack_frame): Likewise.
5201         (nds32_return_addr_rtx): Likewise.
5202         (nds32_initial_elimination_offset): Likewise.
5203         (nds32_expand_prologue): Likewise.
5204         (nds32_expand_epilogue): Likewise.
5205         (nds32_dynamic_chain_address): New function.
5206         * config/nds32/nds32.h (machine_function): Add fields for dwarf
5207         exception handling.
5208         (DYNAMIC_CHAIN_ADDRESS): Define.
5209         (EH_RETURN_DATA_REGNO): Define.
5210         (EH_RETURN_STACKADJ_RTX): Define.
5211         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
5212         patterns for dwarf exception handling.
5213
5214 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
5215
5216         * config/nds32/nds32.h: Clean up obsolete macros.
5217
5218 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
5219
5220         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5221         Add enum values for particular instructions.
5222         * config/nds32/nds32-intrinsic.c: Implementation of expanding
5223         particular intrinsic functions.
5224         * config/nds32/nds32-intrinsic.md: Likewise.
5225         * config/nds32/nds32_intrinsic.h: Likewise.
5226         * config/nds32/nds32.h (nds32_builtins): Likewise.
5227         * config/nds32/nds32.md (type): Add pbsad and pbsada.
5228         (btst, ave): New patterns for particular instructions.
5229
5230 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
5231
5232         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5233         Add enum values for atomic load/store and memory sync.
5234         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5235         and memory sync.
5236         * config/nds32/nds32-intrinsic.md: Likewise.
5237         * config/nds32/nds32_intrinsic.h: Likewise.
5238         * config/nds32/nds32.h (nds32_builtins): Likewise.
5239
5240 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
5241
5242         PR tree-optimization/85257
5243         * fold-const.c (native_encode_vector): If not all elts could fit
5244         and off is -1, return 0 rather than offset.
5245         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5246         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5247         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
5248         adjust buffer in native_interpret_expr call.
5249
5250 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
5251
5252         * config/nds32/constants.md (unspec_volatile_element): Add cache
5253         control enum values.
5254         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5255         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5256         * config/nds32/nds32.c (nds32_cctl_names): New.
5257         (nds32_print_operand): Handle cache control register names.
5258         * config/nds32/nds32.h (nds32_builtins): New enum values.
5259         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5260         macros.
5261         * config/nds32/nds32.md (type): Add mmu.
5262         * config/nds32/pipelines.md (simple_insn): Add mmu.
5263
5264 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
5265
5266         * config/nds32/nds32.md (type): Remove call.
5267         * config/nds32/pipelines.md (simple_insn): Likewise.
5268
5269 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
5270
5271         * config/nds32/constants.md (unspec_volatile_element): Add
5272         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5273         UNSPEC_VOLATILE_FMFCFG.
5274         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5275         description for fmfcfg and fmfcsr.
5276         (bdesc_1arg): Add fmtcsr.
5277         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5278         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5279         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5280         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5281         unspec_fmfcfg): New patterns.
5282         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5283         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5284         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5285         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5286         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5287         __nds32__fmfcfg): Define.
5288
5289 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
5290
5291         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5292         intrinsic register names.
5293         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5294         intrinsic register enum values and macros.
5295
5296 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
5297
5298         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5299         for load/store addressing form.
5300         (nds32_print_operand_address): Likewise.
5301
5302 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
5303
5304         PR target/85196
5305         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5306         based on LABEL_REF.  Remove useless assertion.
5307         (pic_address_needs_scratch): Fix formatting.
5308         (sparc_legitimize_pic_address): Minor tweaks.
5309         (sparc_delegitimize_address): Adjust assertion accordingly.
5310         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5311         into symbolic_operand.
5312         (movsi_high_pic_label_ref): Likewise.
5313         (movsi_lo_sum_pic_label_ref): Likewise.
5314         (movdi_pic_label_ref): Likewise.
5315         (movdi_high_pic_label_ref): Likewise.
5316         (movdi_lo_sum_pic_label_ref): Likewise.
5317
5318 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
5319
5320         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5321         custom LIB_SPEC setup.
5322
5323 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
5324             Kito Cheng  <kito.cheng@gmail.com>
5325
5326         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5327         * config/riscv/freebsd.h: New.
5328
5329 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
5330
5331         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5332         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5333         file.
5334
5335 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
5336             Kito Cheng  <kito.cheng@gmail.com>
5337
5338         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5339         nds32_output_call, nds32_symbol_binds_local_p): New functions.
5340         * config/nds32/nds32-protos.h (nds32_output_call,
5341         nds32_output_return): Declare.
5342         * config/nds32/nds32.md: Refine all the call and return patterns.
5343
5344 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
5345
5346         PR debug/85252
5347         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5348         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5349
5350         PR rtl-optimization/84872
5351         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5352         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5353         EDGE_CROSSING edge.
5354
5355 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
5356
5357         * expr.c (copy_blkmode_to_reg): Revert 254862.
5358         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5359
5360 2018-04-06  Richard Biener  <rguenther@suse.de>
5361
5362         PR middle-end/85244
5363         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5364         after seeing a component reference with an adjacent field.  Treat
5365         refs to arrays at struct end of external decls similar to
5366         refs to unconstrained commons.
5367
5368 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
5369
5370         PR sanitizer/85213
5371         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5372         look through SAVE_EXPRs with non-side-effects argument.  Adjust
5373         recursive calls.
5374         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5375         save_p here.
5376
5377 2018-04-06  Richard Biener  <rguenther@suse.de>
5378
5379         PR middle-end/85180
5380         * alias.c (find_base_term): New wrapper around find_base_term
5381         unwinding CSELIB_VAL_PTR changes.
5382         (find_base_term): Do not restore CSELIB_VAL_PTR during the
5383         recursion.
5384
5385 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
5386
5387         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5388         instructions.
5389         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5390         constant definitions.
5391         ("nop"): lr 0,0 -> nopr r0
5392         ("nop_lr0", "nop_lr1"): New insn definitions.
5393
5394 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
5395
5396         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5397         NDS32_V3PUSH_AVAILABLE_P macro.
5398
5399 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
5400             Chung-Ju Wu  <jasonwucj@gmail.com>
5401
5402         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5403         (nds32*-*-*): Add float and fpu_config into supported_defaults.
5404         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5405         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5406         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5407         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5408         * config/nds32/constraints.md: New constraints and checking for hard
5409         float configuration.
5410         * config/nds32/iterators.md: New mode iterator and attribute for hard
5411         float configuration.
5412         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5413         patterns.
5414         * config/nds32/nds32-fpu.md: New file.
5415         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5416         deal with hard float code generation.
5417         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5418         ARCH_V3S.
5419         (abi_type, float_reg_number): New enum type.
5420         * config/nds32/nds32-predicates.c: New predicates for hard float.
5421         * config/nds32/nds32-protos.h: Declare functions for hard float.
5422         * config/nds32/nds32.c: Implementation for hard float configuration.
5423         * config/nds32/nds32.h: Definitions for hard float configuration.
5424         * config/nds32/nds32.md: Include hard float machine description and
5425         modify patterns for hard float configuration.
5426         * config/nds32/nds32.opt: New options for hard float configuration.
5427         * config/nds32/predicates.md: New predicates for hard float
5428         configuration.
5429
5430 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
5431
5432         * common/config/nds32/nds32-common.c
5433         (nds32_option_optimization_table): Enable -mreleax-hint by default.
5434
5435 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
5436
5437         PR middle-end/85195
5438         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5439         CONSTRUCTOR_ELT (ctor, ...)->value.
5440
5441 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
5442
5443         PR target/85193
5444         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5445
5446 2018-04-05  Tom de Vries  <tom@codesourcery.com>
5447
5448         PR target/85204
5449         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5450         cond jump.
5451
5452 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
5453             Kito Cheng  <kito.cheng@gmail.com>
5454
5455         * config/nds32/constraints.md (U33): Fine-tune checking condition.
5456         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5457         * config/nds32/nds32.h (nds32_16bit_address_type): Add
5458         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5459
5460 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
5461             Kito Cheng  <kito.cheng@gmail.com>
5462
5463         * config/nds32/constraints.md (Ufe): New memory constraint.
5464         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5465         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5466         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5467         operands.
5468         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5469         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5470
5471 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
5472
5473         * config/nds32/nds32.md: Use optimize_size in the condition for
5474         alu-shift instructions.
5475
5476 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
5477
5478         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5479
5480 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
5481
5482         * config/nds32/nds32.md (negsi2): Refine pattern.
5483
5484 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
5485             Chung-Ju Wu  <jasonwucj@gmail.com>
5486
5487         * config/nds32/iterators.md (shift_rotate): New code iterator.
5488         (shift): New code attribute.
5489         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5490         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5491         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5492         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5493         bit-wise operations.
5494         (andsi3, *andsi3): Ditto.
5495         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5496         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5497         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5498         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5499         nds32_ior_operand, nds32_xor_operand): New predicates.
5500
5501 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
5502
5503         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5504         (addsi3, subsi3): ... this.
5505
5506 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
5507
5508         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5509
5510 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
5511
5512         * config/nds32/nds32.md: Adjust indention.
5513
5514 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
5515
5516         * config/nds32/nds32.md (feature): New attribute.
5517
5518 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
5519
5520         * config/nds32/nds32.md (subtype): New attribute.
5521
5522 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
5523
5524         PR target/85203
5525         * config/arm/arm-builtins.c (arm_expand_builtin): Change
5526         expansion to perform a bitwise AND of the argument followed by a
5527         boolean negation of the result.
5528
5529 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
5530
5531         PR rtl-optimization/84878
5532         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5533         the basic block.  Assert the use reference is not artificial and that
5534         it has an associated insn.
5535
5536 2018-04-04  Michael Matz  <matz@suse.de>
5537
5538         * builtins.c (compute_objsize): Pass correct operand
5539         to array_at_struct_end_p.
5540
5541 2018-04-04  Richard Biener  <rguenther@suse.de>
5542
5543         PR lto/85176
5544         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5545         from contexts for DINFO_LEVEL_TERSE and below.
5546
5547 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
5548
5549         * config/nds32/nds32-doubleword.md (move_<mode>): Require
5550         resiter_operand condition.
5551         * config/nds32/nds32.md (*move<mode>): Ditto.
5552
5553 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
5554             Monk Chiang  <sh.chiang04@gmail.com>
5555
5556         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5557
5558 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
5559
5560         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5561
5562 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
5563             Kito Cheng  <kito.cheng@gmail.com>
5564
5565         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5566         nds32_cond_code_str, output_cond_branch,
5567         output_cond_branch_compare_zero, nds32_expand_cbranch,
5568         nds32_expand_cstore, nds32_expand_movcc,
5569         nds32_output_cbranchsi4_equality_zero,
5570         nds32_output_cbranchsi4_equality_reg,
5571         nds32_output_cbranchsi4_equality_reg_or_const_int,
5572         nds32_output_cbranchsi4_greater_less_zero: New functions.
5573         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5574         nds32_expand_cstore, nds32_expand_movcc,
5575         nds32_output_cbranchsi4_equality_zero,
5576         nds32_output_cbranchsi4_equality_reg,
5577         nds32_output_cbranchsi4_equality_reg_or_const_int,
5578         nds32_output_cbranchsi4_greater_less_zero): Declare.
5579         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5580         nds32_rimm11s_operand): New predicates.
5581         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5582         * config/nds32/nds32.md: Rewrite all the branch and conditional move
5583         patterns.
5584
5585 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
5586
5587         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5588         * config/nds32/nds32.md: Ditto.
5589         * config/nds32/pipelines.md: Ditto.
5590
5591 2018-04-04  Richard Biener  <rguenther@suse.de>
5592
5593         PR tree-optimization/85168
5594         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5595         propagating abnormals.
5596
5597 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
5598
5599         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5600
5601 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
5602             Kito Cheng  <kito.cheng@gmail.com>
5603
5604         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5605         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5606         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5607         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5608         * config/nds32/nds32.md (sibcall_internal): New.
5609         (sibcall_register): Remove.
5610         (sibcall_immediate): Remove.
5611         (sibcall_value_internal): New.
5612         (sibcall_value_register): Remove.
5613         (sibcall_value_immediate): Remove.
5614         * config/nds32/predicates.md (nds32_general_register_operand): New.
5615         (nds32_call_address_operand): New.
5616
5617 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
5618
5619         PR rtl-optimization/85167
5620         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5621         bb_defs if *split_p, instead preinitialize it to NULL.
5622
5623         PR tree-optimization/85156
5624         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5625         evaluating the argument multiple times.
5626
5627 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
5628
5629         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5630         than vector.
5631         (_mm_cvtpd_ps): Likewise.
5632         (_mm_cvttpd_epi32): Likewise.
5633         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5634         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5635         vector, pixel, and bool following altivec.h include.
5636
5637 2018-04-03  Martin Sebor  <msebor@redhat.com>
5638
5639         * doc/extend.texi (Common Function Attributes): Clarify.
5640         (const attribute): Likewise.
5641         (pure attribute): Likewise.
5642
5643 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
5644
5645         PR target/85169
5646         * config/i386/i386.c (ix86_expand_vector_set): Use
5647         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
5648
5649 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
5650
5651         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5652         instructions when changing rounding bits to preserve precision bits
5653         in the x87 control word.
5654
5655 2018-04-03  Martin Liska  <mliska@suse.cz>
5656
5657         PR tree-optimization/82491
5658         * rtl.h (strip_offset_and_add): Replace += suboffset with
5659         poly_uint64 () + suboffset.
5660
5661 2018-03-29  Martin Liska  <mliska@suse.cz>
5662             Martin Jambor  <mjambor@suse.cz>
5663
5664         PR ipa/84947
5665         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5666         param_type is not an integral or pointer type.
5667
5668 2018-04-03  Richard Biener  <rguenther@suse.de>
5669
5670         * sese.h (recompute_all_dominators): Remove.
5671
5672 2018-04-02  Martin Sebor  <msebor@redhat.com>
5673
5674         * doc/invoke.texi (-Wrestrict): Fix typos.
5675
5676 2018-04-02  Jim Wilson  <jimw@sifive.com>
5677
5678         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5679         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5680         (<optab>di3, <optab>si3_extend): Likewise.
5681         (<optab>si3_mask, <optab>si3_mask_1): New.
5682         (<optab>di3_mask, <optab>di3_mask_1): New.
5683         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5684         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5685         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5686
5687 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
5688
5689         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5690         example.
5691
5692 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
5693
5694         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5695         (nds32_canonicalize_comparison): New function.
5696
5697 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
5698             Kito Cheng  <kito.cheng@gmail.com>
5699             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
5700
5701         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5702         * config/nds32/constants.md (unspec_volatile_element): Add
5703         UNSPEC_VOLATILE_RELAX_GROUP.
5704         * config/nds32/nds32-relax-opt.c: New file.
5705         * config/nds32/nds32-predicates.c
5706         (nds32_symbol_load_store_p): New function.
5707         * config/nds32/nds32-protos.h
5708         (nds32_symbol_load_store_p): Declare function.
5709         (make_pass_nds32_relax_opt): Declare new rtl pass function.
5710         * config/nds32/nds32.c
5711         (nds32_register_pass): New function to register pass.
5712         (nds32_register_passes): New function to register passes.
5713         * config/nds32/nds32.md (relax_group): New pattern.
5714         * config/nds32/nds32.opt (mrelax-hint): New option.
5715         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5716
5717 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
5718
5719         * config/nds32/t-nds32: Modify files dependency.
5720
5721 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
5722
5723         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5724         (PROFILE_HOOK): Define its implementation.
5725
5726 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
5727
5728         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5729         type and 32-bit size.
5730
5731 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
5732
5733         PR middle-end/85090
5734         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5735         (V_128_256): New mode iterator.
5736         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5737         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5738         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5739         of V.
5740         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5741         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5742
5743 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
5744
5745         PR target/83315
5746         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5747         NaN inputs correctly.
5748
5749 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
5750
5751         PR target/80546
5752         * config/rs6000/vsx.md (??r): New mode attribute.
5753         (*vsx_mov<mode>_64bit): Use it.
5754         (*vsx_mov<mode>_32bit): Likewise.
5755
5756 2018-03-30  Martin Sebor  <msebor@redhat.com>
5757
5758         PR tree-optimization/84818
5759         * builtins.c (check_access): Use warning_n.
5760
5761 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
5762
5763         PR target/83822
5764         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5765         condition.
5766         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5767         condition.
5768
5769 2018-03-30  Julia Koval  <julia.koval@intel.com>
5770
5771         PR target/84413
5772         * x86-tune.def (movx, partial_reg_dependency): Enable for
5773         m_SKYLAKE_AVX512.
5774
5775 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
5776
5777         PR inline-asm/84985
5778         * lra-constraints.c (process_alt_operands): Move setting
5779         this_alternative_matches below.
5780
5781 2018-03-29  Martin Liska  <mliska@suse.cz>
5782
5783         PR lto/84995.
5784         * doc/invoke.texi: Document how LTO works with debug info.
5785         Describe auto-load support of binutils.  Mention 'x86-64'
5786         as valid option value of -march option.
5787
5788 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
5789
5790         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5791
5792         PR c/85094
5793         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5794         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5795         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5796         checking.
5797
5798 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
5799
5800         PR target/84912
5801         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5802         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5803         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5804         for RS6000_BTM_POWERPC64.
5805         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5806         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5807         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5808         definition.
5809         (DIVDE): Use it.
5810         (DIVDEU): Likewise.
5811
5812 2018-03-28 Carl Love  <cel@us.ibm.com>
5813
5814         Revert
5815         2017-09-27  Carl Love  <cel@us.ibm.com>
5816
5817         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5818         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5819         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5820         fctiw instruction.
5821
5822 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5823
5824         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5825         instead of __vector bool.
5826         (_mm_max_pu8): Likewise.
5827         (_mm_min_pi16): Likewise.
5828
5829 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
5830
5831         PR target/84912
5832         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5833         (DIVWEUO): Likewise.
5834         (DIVDEO): Likewise.
5835         (DIVDEUO): Likewise.
5836         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5837         DIVWEUO and DIVDEUO.
5838         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5839         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5840         (div_extend): Likewise.
5841         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5842         builtin function.
5843         (__builtin_divweuo): Likewise.
5844         (__builtin_divdeo): Likewise.
5845         (__builtin_divdeuo): Likewise.
5846
5847 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
5848
5849         PR target/85095
5850         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5851         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5852
5853         PR tree-optimization/82004
5854         * gimple-match-head.c (optimize_pow_to_exp): New function.
5855         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5856         Don't fold to exp if optimize_pow_to_exp is false.
5857
5858 2018-03-28  Martin Liska  <mliska@suse.cz>
5859
5860         PR other/84819
5861         * calls.c (initialize_argument_information): Fix trailing space.
5862         * common.opt: Fix typo and provide better explanation for
5863         -fsanitize-coverage option.
5864         * config/i386/i386.opt: Fix typo.
5865
5866 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
5867             Martin Liska  <mliska@suse.cz>
5868
5869         PR sanitizer/85081
5870         * gimplify.c (asan_poison_variable): Don't do the check for
5871         gimplify_omp_ctxp here.
5872         (gimplify_decl_expr): Do it here.
5873         (gimplify_target_expr): Likewise.
5874
5875 2018-03-28  Martin Liska  <mliska@suse.cz>
5876
5877         PR target/84988
5878         * config/i386/i386.c (ix86_function_arg_advance): Do not call
5879         chkp_type_bounds_count if MPX is not enabled.
5880
5881 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
5882
5883         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5884
5885 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
5886
5887         PR target/84914
5888         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5889         function to create the function decl for complex long double
5890         multiply and divide for -mabi=ieeelongdouble.
5891         (init_float128_ieee): Call it.
5892
5893 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
5894
5895         PR target/85044
5896         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5897         -fcf-protection=branch -mibt.
5898         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5899
5900 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5901
5902         PR target/81863
5903         * config/arm/arm.c (arm_valid_symbolic_address): Handle
5904         arm_word_relocations.
5905
5906 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
5907
5908         PR target/85056
5909         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5910         extern array declarations.
5911
5912 2018-03-27  Richard Biener  <rguenther@suse.de>
5913
5914         PR middle-end/84067
5915         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5916         explicit single_use checks.
5917
5918 2018-03-27  Richard Biener  <rguenther@suse.de>
5919
5920         PR tree-optimization/85082
5921         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5922         Valueize the VUSE.
5923
5924 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5925
5926         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5927         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5928         Turn on fasynchronous-unwind-tables and funwind-tables.
5929
5930 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
5931
5932         PR target/85073
5933         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5934         (*bmi_blsr_<mode>_ccz): Ditto.
5935
5936 2018-03-26  Tom de Vries  <tom@codesourcery.com>
5937
5938         PR tree-optimization/85063
5939         * omp-general.c (offloading_function_p): New function.  Factor out
5940         of ...
5941         * omp-offload.c (pass_omp_target_link::gate): ... here.
5942         * omp-general.h (offloading_function_p): Declare.
5943         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5944         with attribute omp declare target for offloading functions.
5945
5946 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
5947
5948         PR tree-optimization/84005
5949         * tree-data-ref.h (get_base_for_alignment): Declare.
5950         * tree-data-ref.c (get_base_for_alignment_1): New function.
5951         (get_base_for_alignment): Likewise.
5952         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5953         get_base_for_alignment to find a suitable base object, instead
5954         of always using drb->base_address.
5955
5956 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
5957
5958         PR inline-asm/85022
5959         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5960         known size by default.
5961
5962 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
5963
5964         PR inline-asm/85030
5965         * lra-constraints.c (process_alt_operands): Don't match BLKmode
5966         and non BLKmode operands.
5967
5968 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5969
5970         PR target/85026
5971         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5972         Clean up attributes.
5973
5974 2018-03-23  Richard Biener  <rguenther@suse.de>
5975
5976         PR debug/85020
5977         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5978         we are going to emit early debug for LTO.
5979
5980 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
5981
5982         PR inline-asm/85034
5983         * function.c (match_asm_constraints_1): Don't optimize if input
5984         doesn't satisfy general_operand predicate for output's mode.
5985
5986         PR inline-asm/85022
5987         * alias.c (write_dependence_p): Don't require for x_canonicalized
5988         non-VOIDmode if x has VOIDmode.
5989
5990         PR sanitizer/85029
5991         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5992         just don't try to optimize it rather than assert it never happens.
5993
5994 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5995
5996         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5997         macro expansions for definition of ST_INTERNAL_<mode> and
5998         LD_INTERNAL_<mode> builtins.
5999         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
6000         Remove prototype.
6001         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
6002         function.
6003         (altivec_expand_st_builtin): Likewise.
6004         (altivec_expand_builtin): Remove calls to deleted functions.
6005         (rs6000_address_for_altivec): Delete this function.
6006         * config/rs6000/vector.md: Remove expands for
6007         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
6008
6009 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
6010
6011         PR target/84826
6012         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
6013         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
6014         re-computing once computed.
6015         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
6016         (arm_init_machine_status): Initialize
6017         machine->static_chain_stack_bytes.
6018
6019 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
6020
6021         PR target/84760
6022         * doc/extend.texi: Add four new prototypes for vec_ld.
6023         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
6024         definitions for more logical presentation.
6025         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
6026         entries for V1TI variants of __builtin_altivec_ld builtin.
6027         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
6028         handling of V1TI variant of LVX icode pattern.
6029         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
6030         (rs6000_gimple_fold_builtin): Likewise.
6031         (altivec_init_builtins): Add code to define
6032         __builtin_altivec_lvx_v1ti function.
6033
6034 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
6035
6036         PR inline-asm/84941
6037         * function.c (match_asm_constraints_1): Don't do the optimization
6038         if input isn't a REG, SUBREG, MEM or constant.
6039
6040 2018-03-22  Tom de Vries  <tom@codesourcery.com>
6041
6042         PR tree-optimization/84956
6043         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
6044         bb_has_abnormal_pred.
6045
6046 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
6047
6048         PR sanitizer/85018
6049         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
6050         DECL_INITIAL (decl) to decl at the end.
6051         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
6052         adjust the comment.
6053
6054 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
6055
6056         * doc/extend.texi (__builtin_tgmath): Document when complex
6057         integer types are treated as _Complex _Float64.
6058
6059 2018-03-21  Tom de Vries  <tom@codesourcery.com>
6060
6061         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
6062
6063 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
6064
6065         PR tree-optimization/84960
6066         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
6067         if it is ENTRY block, move them into single succ of ENTRY in that case.
6068
6069 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
6070
6071         PR tree-optimization/84811
6072         * poly-int.h (poly_span_traits): Remove the T3 parameter and
6073         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
6074         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
6075         (known_subrange_p): Update accordingly.  Cast each value involved
6076         in the size comparison, rather than casting the result of the
6077         subtraction.
6078
6079 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
6080
6081         PR tree-optimization/84982
6082         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
6083         by flipping the least significant bit rather than all bits from
6084         bitpos to bitpos + bitsize - 1.
6085
6086 2018-03-21  Nathan Sidwell  <nathan@acm.org>
6087
6088         * doc/extend.texi (Deprecated Features): Remove mention of
6089         long-deleted deprecations.
6090
6091 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
6092
6093         PR jit/84288
6094         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
6095         * configure: Regenerate.
6096
6097 2018-03-21  Tom de Vries  <tom@codesourcery.com>
6098
6099         PR tree-optimization/83126
6100         * tree-parloops.c (num_phis): New function.
6101         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
6102
6103 2018-03-21  Nathan Sidwell  <nathan@acm.org>
6104
6105         * doc/extend.texi (Deprecated Features): Update deprecated flags,
6106         mention anon-struct/union members and trailing attributes.
6107
6108 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
6109
6110         PR tree-optimization/84969
6111         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
6112         builtin memset partitions if they set different rhs values.
6113
6114 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
6115
6116         PR rtl-optimization/84989
6117         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
6118         VEC_DUPLICATE with scalar result mode.
6119
6120 2018-03-21  Martin Liska  <mliska@suse.cz>
6121
6122         PR ipa/84963
6123         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
6124         not intended return statement.
6125
6126 2018-03-21  Martin Liska  <mliska@suse.cz>
6127
6128         PR target/84988
6129         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
6130         (chkp_find_bound_slots_1): Limit number of iterations.
6131
6132 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
6133
6134         PR target/84838
6135         * Minor grammar fixes for x86 options.
6136
6137 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
6138
6139         PR debug/84875
6140         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
6141         holding REG_CFA_RESTORE notes, instead turn them into a USE.
6142
6143 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
6144
6145         PR target/83789
6146         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
6147         (altivec_lvx_<mode>_1op): Likewise.
6148         (altivec_stvx_<mode>_2op): Likewise.
6149         (altivec_stvx_<mode>_1op): Likewise.
6150         (altivec_lvx_<VM2:mode>): New define_expand.
6151         (altivec_stvx_<VM2:mode>): Likewise.
6152         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
6153         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6154         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
6155         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6156         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
6157         (rs6000_gen_lvx): Likewise.
6158         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
6159         (altivec_expand_stv_builtin): Likewise.
6160         (altivec_expand_builtin): Likewise.
6161         * config/rs6000/vector.md: Likewise.
6162
6163 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6164
6165         PR target/82518
6166         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
6167         BYTES_BIG_ENDIAN.
6168
6169 2018-03-20  Richard Biener  <rguenther@suse.de>
6170
6171         PR target/84986
6172         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
6173         sign-conversions as zero, fall back to standard scalar_stmt
6174         cost for the rest.
6175
6176 2018-03-20  Martin Liska  <mliska@suse.cz>
6177
6178         PR ipa/84825
6179         * predict.c (rebuild_frequencies): Handle case when we have
6180         PROFILE_ABSENT, but flag_guess_branch_prob is false.
6181
6182 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
6183
6184         PR target/84990
6185         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
6186         flag_section_anchors.
6187         * varasm.c (use_blocks_for_decl_p): Remove hack for
6188         dw2_force_const_mem.
6189
6190         PR target/84845
6191         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
6192         to ...
6193         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
6194         be created, use lowpart_subreg of operands[0] rather than operands[0]
6195         itself.
6196         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
6197         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
6198         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
6199         and n constraint instead of aarch64_shift_imm_di and Usd.
6200         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
6201         (*aarch64_<optab>_reg_minus<mode>3): ... this.
6202
6203 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
6204
6205         PR target/82989
6206         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
6207         to favor GPR over NEON registers.
6208         (<shift>di3_neon): Likewise.
6209
6210 2018-03-20  Tom de Vries  <tom@codesourcery.com>
6211
6212         PR target/84952
6213         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
6214         (nvptx_process_pars): Emit bar.sync asap and alap.
6215
6216 2018-03-20  Tom de Vries  <tom@codesourcery.com>
6217
6218         PR target/84954
6219         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
6220         seen_label if seen_label is already set.
6221
6222 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
6223
6224         PR target/84945
6225         * config/i386/i386.c (fold_builtin_cpu): For features above 31
6226         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6227         Use 1U instead of 1.  Formatting fixes.
6228
6229         PR c/84953
6230         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6231         instead of TREE_TYPE (s1) for the return value.
6232
6233 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
6234
6235         PR tree-optimization/84946
6236         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6237         bitsize + bitsize in poly_uint64 rather than poly_int64.
6238
6239         PR sanitizer/78651
6240         * dwarf2asm.c: Include fold-const.c.
6241         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6242         of decl rather than decl itself.
6243
6244         PR rtl-optimization/84643
6245         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6246
6247 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
6248
6249         PR sanitizer/78651
6250         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6251         calling assemble_variable.
6252
6253 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
6254
6255         PR target/81647
6256         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6257         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6258
6259 2018-03-19  Jim Wilson  <jimw@sifive.com>
6260
6261         PR bootstrap/84856
6262         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6263         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6264         (riscv_first_stack_step): Likewise.
6265         (riscv_option_override): Use STACK_BOUNDARY instead of
6266         MIN_STACK_BOUNDARY.
6267         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6268         MIN_STACK_BOUNDARY.
6269         (BIGGEST_ALIGNMENT): Set to 128.
6270         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6271         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6272         STACK_BOUNDARY.
6273
6274 2018-03-19  Richard Biener  <rguenther@suse.de>
6275
6276         PR tree-optimization/84933
6277         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6278         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6279
6280 2018-03-19  Richard Biener  <rguenther@suse.de>
6281
6282         PR tree-optimization/84859
6283         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6284         (cond_if_else_store_replacement): Perform sinking operation on
6285         single-store BBs regardless of MAX_STORES_TO_SINK setting.
6286         Generalize what a BB with a single eligible store is.
6287
6288 2018-03-19  Richard Biener  <rguenther@suse.de>
6289
6290         PR tree-optimization/84929
6291         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6292         chrec_is_positive against non-chrec arg.
6293
6294 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
6295
6296         PR target/84711
6297         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6298
6299 2018-03-18  Martin Liska  <mliska@suse.cz>
6300
6301         PR rtl-optimization/84635
6302         * regrename.c (build_def_use): Use matches_mode only when
6303         matches >= 0.
6304
6305 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
6306
6307         PR tree-optimization/84913
6308         * tree-vect-loop.c (vectorizable_reduction): Don't try to
6309         vectorize chains of COND_EXPRs.
6310
6311 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
6312
6313         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6314
6315 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
6316
6317         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6318
6319 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
6320
6321         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6322
6323 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
6324             Kito Cheng  <kito.cheng@gmail.com>
6325
6326         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6327         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6328         (nds32_adjust_reg_alloc_order): New function.
6329         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6330
6331 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
6332
6333         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6334         nds32_print_operand, nds32_print_operand_address): Use
6335         HOST_WIDE_INT_PRINT_DEC instead.
6336
6337 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
6338
6339         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6340
6341 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
6342
6343         PR target/84902
6344         * config/i386/i386.c (initial_ix86_tune_features,
6345         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6346         unsigned long long.
6347         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6348         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6349         rather than 1u << ix86_tune.  Formatting fix.
6350         (ix86_option_override_internal): Change ix86_arch_mask from
6351         unsigned int to unsigned HOST_WIDE_INT, initialize to
6352         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6353         (ix86_function_specific_restore): Likewise.
6354
6355 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
6356
6357         PR target/84899
6358         * postreload.c (reload_combine_recognize_pattern): Perform
6359         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6360         truncate_int_for_mode the result for the destination's mode.
6361
6362         PR c/84909
6363         * hsa-gen.c (mem_type_for_type): Fix comment typo.
6364         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6365         Likewise.
6366         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6367         Likewise.
6368
6369 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
6370
6371         PR target/84876
6372         * lra-assigns.c (lra_split_hard_reg_for): Don't use
6373         regno_allocno_class_array and sorted_pseudos.
6374         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6375         insns where regno is used.
6376
6377 2018-03-16  Martin Liska  <mliska@suse.cz>
6378
6379         PR ipa/84833
6380         * multiple_target.c (create_dispatcher_calls): Redirect
6381         reference in the symbol table.
6382
6383 2018-03-16  Martin Liska  <mliska@suse.cz>
6384
6385         PR ipa/84722
6386         * multiple_target.c (create_dispatcher_calls): Redirect also
6387         an alias.
6388
6389 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
6390
6391         PR c++/79937
6392         PR c++/82410
6393         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6394         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6395         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6396
6397 2018-03-16  Julia Koval  <julia.koval@intel.com>
6398
6399         * doc/invoke.texi (Skylake Server): Add CLWB.
6400         Cannonlake): Remove CLWB.
6401
6402 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
6403
6404         PR tree-optimization/84841
6405         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6406         1 << 3.
6407         (FLOAT_ONE_CONST_TYPE): Define.
6408         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6409         (sort_by_operand_rank): Put entries with higher constant_type last
6410         rather than first to match comments.
6411
6412 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
6413
6414         * config/nios2/nios2.md (movsi_internal): Fix thinko in
6415         split predicate.
6416
6417 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
6418
6419         PR c++/79085
6420         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6421         check and use address of target always.
6422
6423 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
6424
6425         PR target/84574
6426         * config/i386/i386.c (indirect_thunk_needed): Update comments.
6427         (indirect_thunk_bnd_needed): Likewise.
6428         (indirect_thunks_used): Likewise.
6429         (indirect_thunks_bnd_used): Likewise.
6430         (indirect_return_needed): New.
6431         (indirect_return_bnd_needed): Likewise.
6432         (output_indirect_thunk_function): Add a bool argument for
6433         function return.
6434         (output_indirect_thunk_function): Don't generate alias for
6435         function return thunk.
6436         (ix86_code_end): Call output_indirect_thunk_function to generate
6437         function return thunks.
6438         (ix86_output_function_return): Set indirect_return_bnd_needed
6439         and indirect_return_needed instead of indirect_thunk_bnd_needed
6440         and indirect_thunk_needed.
6441
6442 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
6443
6444         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6445         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6446         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6447
6448 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
6449             Paul Hua <paul.hua.gm@gmail.com>
6450
6451         PR c/84852
6452         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6453
6454 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
6455
6456         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6457         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6458         resp. SFmode cases.
6459
6460 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
6461
6462         PR target/84711
6463         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6464         instead of GET_MODE_SIZE when comparing Units.
6465
6466 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
6467
6468         PR target/68256
6469         * varasm.c (hash_section): Return an unchangeble hash value
6470         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6471         Return !aarch64_can_use_per_function_literal_pools_p ().
6472
6473 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
6474
6475         PR target/84860
6476         * optabs.c (emit_conditional_move): Pass address of cmode's copy
6477         rather than address of cmode as last argument to prepare_cmp_insn.
6478
6479 2018-03-15  Julia Koval  <julia.koval@intel.com>
6480
6481         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6482         F_AVX512VNNI, F_AVX512BITALG): New.
6483
6484 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
6485
6486         PR target/83451
6487         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
6488         insn for floating-point loads and stores.
6489
6490 2018-03-14  Carl Love  <cel@us.ibm.com>
6491
6492         * config/rs6000/rs6000-c.c: Add macro definitions for
6493         ALTIVEC_BUILTIN_VEC_PERMXOR.
6494         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6495         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6496         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6497         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6498         UNSPEC_VPERMXOR.
6499         * config/doc/extend.texi: Add prototypes for vec_permxor.
6500
6501 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
6502
6503         PR c/84852
6504         * diagnostic-show-locus.c (class layout_point): Convert m_line
6505         from int to linenum_type.
6506         (line_span::comparator): Use linenum "compare" function when
6507         comparing line numbers.
6508         (test_line_span): New function.
6509         (layout_range::contains_point): Convert param "row" from int to
6510         linenum_type.
6511         (layout_range::intersects_line_p): Likewise.
6512         (layout::will_show_line_p): Likewise.
6513         (layout::print_source_line): Likewise.
6514         (layout::should_print_annotation_line_p): Likewise.
6515         (layout::print_annotation_line): Likewise.
6516         (layout::print_leading_fixits): Likewise.
6517         (layout::annotation_line_showed_range_p): Likewise.
6518         (struct line_corrections): Likewise for field m_row.
6519         (line_corrections::line_corrections): Likewise for param "row".
6520         (layout::print_trailing_fixits): Likewise.
6521         (layout::get_state_at_point): Likewise.
6522         (layout::get_x_bound_for_row): Likewise.
6523         (layout::print_line): Likewise.
6524         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
6525         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6526         * input.c (selftest::test_linenum_comparisons): New function.
6527         (selftest::input_c_tests): Call it.
6528         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6529         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6530         * selftest.h (ASSERT_GT): New macro.
6531         (ASSERT_GT_AT): New macro.
6532         (ASSERT_LT): New macro.
6533         (ASSERT_LT_AT): New macro.
6534
6535 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
6536
6537         PR rtl-optimization/84780
6538         * combine.c (distribute_links): Don't make a link based on pc_rtx.
6539
6540 2018-03-14  Martin Liska  <mliska@suse.cz>
6541
6542         * tree.c (record_node_allocation_statistics): Use
6543         get_stats_node_kind.
6544         (get_stats_node_kind): New function extracted from
6545         record_node_allocation_statistics.
6546         (free_node): Use get_stats_node_kind.
6547
6548 2018-03-14  Richard Biener  <rguenther@suse.de>
6549
6550         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6551         that the value-set of ANTIC_IN doesn't grow.
6552
6553         Revert
6554         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6555         member.
6556         (BB_VISITED_WITH_VISITED_SUCCS): New define.
6557         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6558
6559 2018-03-14  Julia Koval  <julia.koval@intel.com>
6560
6561         * config.gcc (icelake-client, icelake-server): New.
6562         (icelake): Remove.
6563         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6564         (initial_ix86_arch_features): Ditto.
6565         (PTA_SKYLAKE): Add SGX.
6566         (PTA_ICELAKE): Remove.
6567         (PTA_ICELAKE_CLIENT): New.
6568         (PTA_ICELAKE_SERVER): New.
6569         (ix86_option_override_internal): Split up icelake on icelake client and
6570         icelake server.
6571         (get_builtin_code_for_version): Ditto.
6572         (fold_builtin_cpu): Ditto.
6573         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6574         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6575         * config/i386/i386.h (processor_type): Ditto.
6576         * doc/invoke.texi: Ditto.
6577
6578 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
6579
6580         PR sanitizer/83392
6581         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6582         INTEGER_CST offset, add it together with bitpos / 8 and
6583         sign extend based on POINTER_SIZE.
6584
6585         PR target/84844
6586         Revert
6587         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
6588
6589         PR target/78090
6590         * config/i386/constraints.md (Yc): New register constraint.
6591         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6592         Use Yc constraint for alternative 2 of operand 0.  Remove
6593         preferred_for_speed attribute.
6594
6595 2018-03-14  Richard Biener  <rguenther@suse.de>
6596
6597         PR tree-optimization/84830
6598         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6599         with the old one to avoid oscillations.
6600
6601 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
6602
6603         PR target/83712
6604         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6605         pseudos.
6606         (assign_by_spills): Return a flag of reload assignment failure.
6607         Do not process the reload assignment failures.  Do not spill other
6608         reload pseudos if they has the same reg class.  Update n if
6609         necessary.
6610         (lra_assign): Add a return arg.  Set up from the result of
6611         assign_by_spills call.
6612         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6613         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
6614         usage_insns if it is not NULL.
6615         (spill_hard_reg_in_range): New function.
6616         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6617         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6618         function prototypes.
6619         (lra_assign): Change prototype.
6620         * lra.c (lra): Add code to deal with fails by splitting hard reg
6621         live ranges.
6622
6623 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
6624
6625         * config/riscv/riscv.opt (mrelax): New option.
6626         * config/riscv/riscv.c (riscv_file_start): Emit ".option
6627         "norelax" when riscv_mrelax is disabled.
6628         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6629
6630 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
6631
6632         PR target/84743
6633         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6634         reassociation for int modes.
6635
6636 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
6637
6638         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6639         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6640         for big-endian.
6641         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6642         * config/aarch64/aarch64-sve.md
6643         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6644         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6645         (*extend<mode><Vwide>2): Rename to...
6646         (aarch64_sve_extend<mode><Vwide>2): ...this.
6647         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6648         renaming the old pattern to...
6649         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
6650         unsigned packs.
6651         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6652         define_expand, renaming the old pattern to...
6653         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6654         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6655         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6656         account when deciding which SVE instruction the optab should use.
6657         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6658
6659 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
6660
6661         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6662         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6663         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6664         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6665         (tlsdesc_small_<mode>): Turn a define_expand and use
6666         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
6667         (tlsdesc_small_advsimd_<mode>): ...this.
6668         (tlsdesc_small_sve_<mode>): New pattern.
6669
6670 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
6671
6672         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6673         (UNSPEC_UMUL_HIGHPART): New constants.
6674         (MUL_HIGHPART): New int iteraor.
6675         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6676         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6677         define_expand.
6678         (*<su>mul<mode>3_highpart): New define_insn.
6679
6680 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
6681
6682         PR lto/84805
6683         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6684         incomplete types.
6685
6686 2018-03-13  Martin Liska  <mliska@suse.cz>
6687
6688         PR ipa/84658.
6689         * (sem_item_optimizer::sem_item_optimizer): Initialize new
6690         vector.
6691         (sem_item_optimizer::~sem_item_optimizer): Release it.
6692         (sem_item_optimizer::merge_classes): Register variable aliases.
6693         (sem_item_optimizer::fixup_pt_set): New function.
6694         (sem_item_optimizer::fixup_points_to_sets): Likewise.
6695         * ipa-icf.h: Declare new variables and functions.
6696
6697 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
6698
6699         PR middle-end/84834
6700         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6701         integer_pow2p@2 and test integer_pow2p in condition.
6702         (A < 0 ? C : 0): Similarly for @1.
6703
6704         PR middle-end/84831
6705         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6706         characters starting at p contain '\0' character, don't look beyond
6707         that.
6708
6709         PR target/84827
6710         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6711         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6712
6713         PR target/84828
6714         * reg-stack.c (change_stack): Change update_end var from int to
6715         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6716         also call set_block_for_insn on the newly added insns and rescan.
6717
6718         PR target/84786
6719         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6720         on the last operand.
6721
6722         PR c++/84704
6723         * tree.c (stabilize_reference_1): Return save_expr (e) for
6724         STATEMENT_LIST even if it doesn't have side-effects.
6725
6726 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
6727
6728         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6729
6730 2018-03-12  Renlin Li  <renlin.li@arm.com>
6731
6732         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6733         aarch64_output_scalar_simd_mov_immediate.
6734
6735 2018-03-12  Martin Sebor  <msebor@redhat.com>
6736
6737         PR tree-optimization/83456
6738         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6739         for perfectly overlapping calls to memcpy.
6740         (gimple_fold_builtin_memory_chk): Same.
6741         (gimple_fold_builtin_strcpy): Handle no-warning.
6742         (gimple_fold_builtin_stxcpy_chk): Same.
6743         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6744
6745 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
6746
6747         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6748         parameter.  Use it for SFmode.
6749         (rs6000_function_arg_advance_1): Adjust.
6750         (rs6000_function_arg): Adjust.
6751         (rs6000_gimplify_va_arg): Pass false for that new parameter.
6752
6753 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
6754
6755         PR rtl-optimization/84169
6756         PR rtl-optimization/84780
6757         * combine.c (can_combine_p): Check for a 2-insn combination whether
6758         the destination register is used between the two insns, too.
6759
6760 2018-03-12  Richard Biener  <rguenther@suse.de>
6761
6762         PR tree-optimization/84803
6763         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6764         for refs DR analysis didn't process.
6765
6766 2018-03-12  Richard Biener  <rguenther@suse.de>
6767
6768         PR tree-optimization/84777
6769         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6770         force-vectorize loops ignore whether we are optimizing for size.
6771
6772 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6773
6774         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6775         (TARGET_MD_ASM_ADJUST): Define.
6776
6777 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
6778             Kito Cheng  <kito.cheng@gmail.com>
6779             Chung-Ju Wu  <jasonwucj@gmail.com>
6780
6781         * config/nds32/nds32.c (nds32_compute_stack_frame,
6782         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6783         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6784         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6785         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6786         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6787         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6788         * config/nds32/nds32.md (prologue, epilogue): Use macro
6789         NDS32_V3PUSH_AVAILABLE_P to do checking.
6790
6791 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
6792
6793         PR debug/58150
6794         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6795         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6796         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
6797         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
6798         addition of most attributes on !orig_type_die or the attribute not
6799         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6800
6801 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
6802             Chung-Ju Wu  <jasonwucj@gmail.com>
6803
6804         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6805         __NDS32_VH__ macro.
6806         * config/nds32/nds32.opt (mvh): New option.
6807
6808 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
6809             Chung-Ju Wu  <jasonwucj@gmail.com>
6810
6811         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6812         function.
6813         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6814         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6815         definition.
6816
6817 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
6818             Chung-Ju Wu  <jasonwucj@gmail.com>
6819
6820         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6821         function.
6822         * config/nds32/nds32-multiple.md (strlensi): New pattern.
6823         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6824
6825 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
6826             Kito Cheng  <kito.cheng@gmail.com>
6827             Chung-Ju Wu  <jasonwucj@gmail.com>
6828
6829         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6830         UNSPEC_FFMISM and UNSPEC_FLMISM.
6831         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6832         for ffb, ffmism and flmism.
6833         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6834         (unspec_ffmism): Ditto.
6835         (unspec_flmism): Ditto.
6836         (nds32_expand_builtin_impl): Check if string extension is available.
6837         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6838         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6839
6840 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
6841
6842         Reverting patch:
6843         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
6844
6845         PR target/83712
6846         * lra-assigns.c (assign_by_spills): Return a flag of reload
6847         assignment failure.  Do not process the reload assignment
6848         failures.  Do not spill other reload pseudos if they has the same
6849         reg class.
6850         (lra_assign): Add a return arg.  Set up from the result of
6851         assign_by_spills call.
6852         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6853         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
6854         usage_insns if it is not NULL.
6855         (spill_hard_reg_in_range): New function.
6856         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6857         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6858         function prototypes.
6859         (lra_assign): Change prototype.
6860         * lra.c (lra): Add code to deal with fails by splitting hard reg
6861         live ranges.
6862
6863 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
6864
6865         PR target/84807
6866         * config/i386/i386.opt: Replace Enforcment with Enforcement.
6867
6868 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
6869
6870         PR debug/84620
6871         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6872         (dw_val_node): Add val_symbolic_view.
6873         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6874         (symview_upper_bound): New.
6875         (new_line_info_table): Initialize symviews_since_reset.
6876         (dwarf2out_source_line): Count symviews_since_reset and set
6877         symview_upper_bound.
6878         (dw_val_equal_p): Handle symview.
6879         (add_AT_symview): New.
6880         (print_dw_val): Handle symview.
6881         (attr_checksum, attr_checksum_ordered): Likewise.
6882         (same_dw_val_p, size_of_die): Likewise.
6883         (value_format, output_die): Likewise.
6884         (add_high_low_attributes): Use add_AT_symview for entry_view.
6885         (dwarf2out_finish): Reset symview_upper_bound, clear
6886         zero_view_p.
6887
6888 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
6889
6890         PR target/83969
6891         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6892         Add strict argument and use it.
6893         (rs6000_split_multireg_move): Update for new strict argument.
6894         (mem_operand_gpr): Disallow all non-offsettable addresses.
6895         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6896
6897 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
6898
6899         PR target/84772
6900         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6901         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6902         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6903
6904         PR c++/84767
6905         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6906         decl, use remap_type if we want to use the type.
6907
6908 2018-03-09  Martin Sebor  <msebor@redhat.com>
6909
6910         PR tree-optimization/84526
6911         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6912         Remove dead code.
6913         (builtin_access::generic_overlap): Be prepared to handle non-array
6914         base objects.
6915
6916 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
6917
6918         PR rtl-optimization/84682
6919         * lra-constraints.c (process_address_1): Check is_address flag
6920         for address constraints.
6921         (process_alt_operands): Likewise.
6922         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6923         preprocess_constraints.
6924         * recog.h (preprocess_constraints): Add oploc parameter.
6925         Adjust callers.
6926         * recog.c (preprocess_constraints): Test address_operand for
6927         CT_ADDRESS constraints.
6928
6929 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
6930
6931         PR target/83712
6932         * lra-assigns.c (assign_by_spills): Return a flag of reload
6933         assignment failure.  Do not process the reload assignment
6934         failures.  Do not spill other reload pseudos if they has the same
6935         reg class.
6936         (lra_assign): Add a return arg.  Set up from the result of
6937         assign_by_spills call.
6938         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6939         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
6940         usage_insns if it is not NULL.
6941         (spill_hard_reg_in_range): New function.
6942         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6943         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6944         function prototypes.
6945         (lra_assign): Change prototype.
6946         * lra.c (lra): Add code to deal with fails by splitting hard reg
6947         live ranges.
6948
6949 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6950
6951         PR target/83193
6952         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6953         Accept complain bool parameter.  Only emit errors if it is true.
6954         (arm_parse_cpu_option_name): Likewise.
6955         (arm_target_thumb_only): Adjust callers of the above.
6956         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6957         prototype to take a default true bool parameter.
6958         (arm_parse_arch_option_name): Likewise.
6959
6960 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
6961             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
6962
6963         PR jit/64089
6964         PR jit/84288
6965         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6966         * configure: Regenerate.
6967         * configure.ac ("linker --version-script option"): New.
6968         ("linker soname option"): New.
6969
6970 2018-03-09  Richard Biener  <rguenther@suse.de>
6971
6972         PR tree-optimization/84775
6973         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6974         immediate uses of predicate stmts and mark them modified.
6975
6976         Revert
6977         PR tree-optimization/84178
6978         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6979         to caller.
6980         (version_loop_for_if_conversion): Delay update_ssa call.
6981         (tree_if_conversion): Delay update_ssa until after predicate
6982         insertion.
6983
6984 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
6985
6986         PR target/84763
6987         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6988         when the function accesses prior frames.
6989
6990 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
6991
6992         PR debug/84456
6993         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6994         gen_llsym, otherwise call maybe_gen_llsym.
6995
6996         PR inline-asm/84742
6997         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6998         has ',' character inside of it.
6999
7000 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7001
7002         PR target/84748
7003         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
7004         as clobbering CC_REGNUM.
7005
7006 2018-03-08  Richard Biener  <rguenther@suse.de>
7007
7008         PR middle-end/84552
7009         * tree-scalar-evolution.c: Include tree-into-ssa.h.
7010         (follow_copies_to_constant): Do not follow SSA names registered
7011         for update.
7012
7013 2018-03-08  Richard Biener  <rguenther@suse.de>
7014
7015         PR tree-optimization/84178
7016         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7017         to caller.
7018         (version_loop_for_if_conversion): Delay update_ssa call.
7019         (tree_if_conversion): Delay update_ssa until after predicate
7020         insertion.
7021
7022 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
7023
7024         PR tree-optimization/84178
7025         * tree-if-conv.c (release_bb_predicate): Remove the
7026         the assertion that the stmts have NULL use_ops.
7027         Discard the statements, asserting that they haven't
7028         yet been added to a BB.
7029
7030 2018-03-08  Richard Biener  <rguenther@suse.de>
7031
7032         PR tree-optimization/84746
7033         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
7034         (phi_translate): Pass in destination ANTIC_OUT set.
7035         (phi_translate_1): Likewise.  For a simplified result lookup
7036         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
7037         (phi_translate_set): Adjust.
7038         (do_pre_regular_insertion): Likewise.
7039         (do_pre_partial_partial_insertion): Likewise.
7040
7041 2018-03-08  Martin Liska  <mliska@suse.cz>
7042
7043         PR gcov-profile/84735
7044         * doc/gcov.texi: Document usage of profile files.
7045         * gcov-io.h: Document changes in the format.
7046
7047 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
7048
7049         PR debug/84404
7050         PR debug/84408
7051         * dwarf2out.c (struct dw_line_info_table): Update comments for
7052         view == -1.
7053         (FORCE_RESET_NEXT_VIEW): New.
7054         (FORCE_RESETTING_VIEW_P): New.
7055         (RESETTING_VIEW_P): Check for -1 too.
7056         (ZERO_VIEW_P): Likewise.
7057         (new_line_info_table): Force-reset next view.
7058         (dwarf2out_begin_function): Likewise.
7059         (dwarf2out_source_line): Simplify zero_view_p initialization.
7060         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
7061         view directly.  Omit view when omitting .loc at line 0.
7062
7063 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
7064
7065         PR tree-optimization/84740
7066         * tree-switch-conversion.c (process_switch): Call build_constructors
7067         only if info.phi_count is non-zero.
7068
7069         PR tree-optimization/84739
7070         * tree-tailcall.c (find_tail_calls): Check call arguments against
7071         DECL_ARGUMENTS (current_function_decl) rather than
7072         DECL_ARGUMENTS (func) when checking for tail recursion.
7073
7074 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
7075
7076         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
7077         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
7078         Volker Reichelt's entry and add entries for people that perform
7079         GCC fuzzy testing and report numerous bugs.
7080
7081 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
7082
7083         PR target/82411
7084         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
7085         readonly data in sdata, if that is disabled.
7086         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
7087         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
7088         -mreadonly-in-sdata option.
7089
7090 2018-03-07  Martin Sebor  <msebor@redhat.com>
7091
7092         PR tree-optimization/84468
7093         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
7094         basic block when looking for nul assignment.
7095
7096 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
7097
7098         PR target/84277
7099         * except.h (output_function_exception_table): Adjust prototype.
7100         * except.c (output_function_exception_table): Remove FNNAME parameter
7101         and add SECTION parameter.  Ouput one part of the table at a time.
7102         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
7103         the first part of the exception table and emit unwind directives.
7104         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
7105         (i386_pe_seh_cold_init): Likewise.
7106         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
7107         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
7108         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
7109         (ix86_output_call_insn): Emit a nop in one more case for SEH.
7110         * config/i386/winnt.c: Include except.h.
7111         (struct seh_frame_state): Add reg_offset, after_prologue and
7112         in_cold_section fields.
7113         (i386_pe_seh_end_prologue): Set seh->after_prologue.
7114         (i386_pe_seh_cold_init): New function.
7115         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
7116         to seh->in_cold_section.
7117         (seh_emit_push): Record the offset of the push.
7118         (seh_emit_save): Record the offet of the save.
7119         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
7120         Test seh->after_prologue to disregard the epilogue.
7121         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
7122         (i386_pe_end_cold_function): New function.
7123
7124 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
7125
7126         PR fortran/84565
7127         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
7128         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
7129
7130         PR c++/84704
7131         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
7132         on tmp_var.
7133         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
7134         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
7135
7136         PR middle-end/84723
7137         * multiple_target.c: Include tree-inline.h and intl.h.
7138         (expand_target_clones): Diagnose and fail if node->definition and
7139         !tree_versionable_function_p (node->decl).
7140
7141 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
7142
7143         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
7144         sprint_ul.
7145         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
7146         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
7147         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
7148
7149 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
7150
7151         PR target/84710
7152         * combine.c (try_combine): Use reg_or_subregno instead of handling
7153         just paradoxical SUBREGs and REGs.
7154
7155 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
7156
7157         * config/arc/arc.c (arc_finalize_pic): Remove function.
7158         (arc_must_save_register): We use single base PIC register, remove
7159         checks to save/restore the PIC register.
7160         (arc_expand_prologue): Likewise.
7161         * config/arc/arc-protos.h (arc_set_default_type_attributes):
7162         Remove.
7163         (arc_verify_short): Likewise.
7164         (arc_attr_type): Likewise.
7165         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
7166         (walk_stores): Likewise.
7167         (arc_address_cost): Make it static.
7168         (arc_verify_short): Likewise.
7169         (branch_dest): Likewise.
7170         (arc_attr_type): Likewise.
7171         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
7172         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
7173         (arc_final_prescan_insn): Remove inserting the nops due to
7174         hardware hazards.  It is done in reorg step.
7175         (insn_length_variant_t): Remove.
7176         (insn_length_parameters_t): Likewise.
7177         (arc_insn_length_parameters): Likewise.
7178         (arc_get_insn_variants): Likewise.
7179         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
7180
7181 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
7182
7183         PR inline-asm/84683
7184         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7185         assertion failure.
7186
7187         PR tree-optimization/84687
7188         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
7189         on new_node->decl.
7190         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
7191
7192 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
7193
7194         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
7195         Rename to ppc_speculation_barrier.
7196         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
7197         __builtin_ppc_speculation_barrier.
7198
7199 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
7200
7201         PR target/84700
7202         * combine.c (combine_simplify_rtx): Don't try to simplify if
7203         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
7204         are equal to x.
7205
7206 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
7207
7208         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
7209         to 32 bytes when compiling for POWER9.
7210
7211 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
7212
7213         PR target/84564
7214         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
7215         regparm >= 3 with no arg reg available also for calls with
7216         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
7217
7218         PR target/84524
7219         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
7220         orig,vex.
7221         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
7222
7223 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
7224
7225         PR target/84264
7226         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7227
7228 2018-03-05  Richard Biener  <rguenther@suse.de>
7229
7230         PR tree-optimization/84486
7231         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7232         When inserting a __builtin_assume_aligned call set the LHS
7233         SSA name alignment info accordingly.
7234
7235 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
7236
7237         PR tree-optimization/84114
7238         * config/aarch64/aarch64.c (aarch64_reassociation_width)
7239         Avoid reassociation of FLOAT_MODE addition.
7240
7241 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
7242
7243         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7244         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7245         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7246         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7247         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7248         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7249         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7250         and -mwbnoinvd.
7251         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7252         __builtin_ia32_wbinvd): New builtins.
7253         (SPECIAL_ARGS2): New.
7254         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7255         (SPECIAL_ARGS2): New.
7256         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7257         (ix86_valid_target_attribute_inner_p): Ditto.
7258         (ix86_init_mmx_sse_builtins): Add special_args2.
7259         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7260         TARGET_WBNOINVD_P): New.
7261         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7262         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7263         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7264         * config/i386/immintrin.h (_wbinvd): New intrinsic.
7265         * config/i386/pconfigintrin.h: New file.
7266         * config/i386/wbnoinvdintrin.h: Ditto.
7267         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
7268         wbnoinvdintrin.h.
7269         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7270
7271 2018-03-05  Richard Biener  <rguenther@suse.de>
7272
7273         PR tree-optimization/84670
7274         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7275         member.
7276         (BB_VISITED_WITH_VISITED_SUCCS): New define.
7277         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7278         (compute_antic_aux): Only assert the number of values in ANTIC_IN
7279         doesn't grow if all successors (recursively) were visited at least
7280         once.
7281
7282 2018-03-05  Richard Biener  <rguenther@suse.de>
7283
7284         PR tree-optimization/84650
7285         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7286         if executed in the loop pipeline.
7287
7288 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
7289
7290         * doc/configfiles.texi (Configuration Files): Move info about
7291         conditionalizing $target-protos.h to...
7292         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
7293         differs from $target-protos.h.
7294
7295 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
7296             Chung-Ju Wu  <jasonwucj@gmail.com>
7297
7298         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7299         * config/nds32/nds32-multiple.md (setmemsi): Define.
7300         * config/nds32/nds32-memory-manipulation.c
7301         (nds32_gen_dup_4_byte_to_word_value): New.
7302         (emit_setmem_word_loop): New.
7303         (emit_setmem_byte_loop): New.
7304         (nds32_expand_setmem_loop): New.
7305         (nds32_expand_setmem_loop_v3m): New.
7306         (nds32_expand_setmem_unroll): New.
7307         (nds32_expand_setmem): New.
7308
7309 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
7310             Chung-Ju Wu  <jasonwucj@gmail.com>
7311
7312         * config/nds32/nds32-memory-manipulation.c
7313         (nds32_emit_load_store): New.
7314         (nds32_emit_post_inc_load_store): New.
7315         (nds32_emit_mem_move): New.
7316         (nds32_emit_mem_move_block): New.
7317         (nds32_expand_movmemsi_loop_unknown_size): New.
7318         (nds32_expand_movmemsi_loop_known_size): New.
7319         (nds32_expand_movmemsi_loop): New.
7320         (nds32_expand_movmemsi_unroll): New.
7321         (nds32_expand_movmemqi): Rename ...
7322         (nds32_expand_movmemsi): ... to this.
7323         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7324         (movmemsi): ... to this.
7325         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7326         (nds32_expand_movmemsi): ... to this.
7327
7328 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
7329             Monk Chiang  <sh.chiang04@gmail.com>
7330             Chung-Ju Wu  <jasonwucj@gmail.com>
7331
7332         * config/nds32/nds32-protos.h
7333         (nds32_expand_load_multiple): New arguments.
7334         (nds32_expand_store_multiple): Ditto.
7335         (nds32_valid_multiple_load_store): Rename ...
7336         (nds32_valid_multiple_load_store_p): ... to this.
7337         * config/nds32/nds32-memory-manipulation.c
7338         (nds32_expand_load_multiple): Refine implementation.
7339         (nds32_expand_store_multiple): Ditto.
7340         * config/nds32/nds32-multiple.md
7341         (load_multiple): Update nds32_expand_load_multiple interface.
7342         (store_multiple): Update nds32_expand_store_multiple interface.
7343         * config/nds32/nds32-predicates.c
7344         (nds32_valid_multiple_load_store): Rename ...
7345         (nds32_valid_multiple_load_store_p): ... to this and refine
7346         implementation.
7347         * config/nds32/predicates.md
7348         (nds32_load_multiple_and_update_address_operation): New predicate.
7349         (nds32_store_multiple_and_update_address_operation): New predicate.
7350
7351 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
7352             Chung-Ju Wu  <jasonwucj@gmail.com>
7353
7354         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7355         (combo): New attribute.
7356         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7357
7358 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
7359
7360         * config/nds32/nds32.opt: Change -mcmodel= default value.
7361
7362 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
7363             Monk Chiang  <sh.chiang04@gmail.com>
7364             Chung-Ju Wu  <jasonwucj@gmail.com>
7365
7366         * config/nds32/constants.md (unspec_element): New enum.
7367         * config/nds32/constraints.md (Umw): New constraint.
7368         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7369         * config/nds32/nds32-intrinsic.md: Likewise.
7370         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7371         (nds32_valid_smw_lwm_base_p): New.
7372         (nds32_output_smw_single_word): New.
7373         (nds32_output_lmw_single_word): New.
7374         (nds32_expand_unaligned_load): New.
7375         (nds32_expand_unaligned_store): New.
7376         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7377         (nds32_output_smw_single_word): Declare.
7378         (nds32_output_lmw_single_word): Declare.
7379         (nds32_expand_unaligned_load): Declare.
7380         (nds32_expand_unaligned_store): Declare.
7381         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7382         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7383         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7384         NDS32_BUILTIN_UASTORE_DW.
7385         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7386         predicate.
7387
7388 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
7389             Kito Cheng  <kito.cheng@gmail.com>
7390             Chung-Ju Wu  <jasonwucj@gmail.com>
7391
7392         * config/nds32/nds32-intrinsic.c
7393         (nds32_expand_builtin_null_ftype_reg): Delete.
7394         (nds32_expand_builtin_reg_ftype_imm): Ditto.
7395         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7396         (nds32_read_argument): New.
7397         (nds32_legitimize_target): Ditto.
7398         (nds32_legitimize_argument): Ditto.
7399         (nds32_check_constant_argument): Ditto.
7400         (nds32_expand_unop_builtin): Ditto.
7401         (nds32_expand_unopimm_builtin): Ditto.
7402         (nds32_expand_binop_builtin): Ditto.
7403         (nds32_builtin_decl_impl): Ditto.
7404         (builtin_description): Ditto.
7405         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7406         (nds32_init_builtins_impl): Ditto.
7407         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7408         (nds32_builtin_decl): New.
7409         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7410         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7411
7412 2018-03-02  Jeff Law  <law@redhat.com>
7413
7414         * reorg.c (stop_search_p): Handle DEBUG_INSN.
7415         (redundant_insn, fill_simple_delay_slots): Likewise.
7416         (fill_slots_from_thread): Likewise.
7417         * resource.c (mark_referenced_resources): Likewise.
7418         (mark_set_resources, find_dead_or_set_registers): Likewise.
7419
7420 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
7421
7422         * substring-locations.h (format_warning_va): Formatting fix for
7423         ATTRIBUTE_GCC_DIAG.
7424         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7425         argument.
7426         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7427         * substring-locations.c: Include intl.h.
7428         (format_warning_va): Turned into small wrapper around
7429         format_warning_n_va, renamed to ...
7430         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7431         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7432         use ngettext.
7433         (format_warning_at_substring_n): New function.
7434         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7435         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
7436         format_warning_at_substring with just a shorter name instead of
7437         const function pointer.
7438         (fmtwarn_n): New function.
7439         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7440         appropriate, get rid of all the fmtstr temporaries, move conditionals
7441         with G_() wrapped string literals directly into fmtwarn arguments,
7442         cast dir.len to (int), formatting fixes.
7443
7444 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
7445
7446         * doc/invoke.texi: Remove "Cilk Plus" references.
7447
7448 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
7449             Richard Biener  <rguenther@suse.de>
7450
7451         PR ipa/84628
7452         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7453         for error or warning attributes if CALL_FROM_THUNK_P is set.
7454         Formatting fixes.
7455
7456 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
7457
7458         PR target/56540
7459         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7460         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7461
7462         PR target/56540
7463         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7464         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7465
7466         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7467         instead of -1U in last predictors element's probability member.
7468
7469 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
7470
7471         PR ipa/83983
7472         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7473         arguments if they are comparable.
7474
7475 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
7476
7477         PR tree-optimization/84634
7478         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7479         masks and masked_loop_p with a single loop_masks, making sure it's
7480         null for bb vectorization.
7481
7482 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
7483
7484         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7485         (vect_analyze_data_ref_access): Use loop->safe_len rather than
7486         loop->force_vectorize to check whether there is no alias.
7487
7488 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
7489
7490         PR target/84614
7491         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7492         prototypes.
7493         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7494         comments.
7495         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7496         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7497         instead of a loop around prev_real_insn.
7498         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7499         prev_real_insn.
7500
7501         PR inline-asm/84625
7502         * config/i386/i386.c (ix86_print_operand): Use conditional
7503         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7504         zero vector.
7505
7506 2018-03-02  Richard Biener  <rguenther@suse.de>
7507
7508         PR tree-optimization/84427
7509         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7510         (bitmap_set_subtract_values): Rewrite to handle multiple
7511         exprs per value.
7512         (clean): Likewise.
7513         (prune_clobbered_mems): Likewise.
7514         (phi_translate): Take edge instead of pred/phiblock.
7515         (phi_translate_1): Likewise.
7516         (phi_translate_set): Likewise.  Insert all translated
7517         exprs for a value into the set, keeping possibly multiple
7518         expressions per value.
7519         (compute_antic_aux): Adjust for phi_translate changes.
7520         When intersecting union the expressions and prune those
7521         not in the final value set, keeping possibly multiple
7522         expressions per value.  Do not use value-insertion
7523         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7524         all expressions.  Add verification that the value-sets
7525         only shrink during iteration.
7526         (compute_partial_antic_aux): Adjust for the phi_translate changes.
7527         (do_pre_regular_insertion): Likewise.
7528         (do_pre_partial_partial_insertion): Likewise.
7529
7530 2018-03-02  Richard Biener  <rguenther@suse.de>
7531
7532         PR target/82005
7533         * config/darwin.c (saved_debug_info_level): New static global.
7534         (darwin_asm_lto_start): Disable debug info generation for LTO out.
7535         (darwin_asm_lto_end): Restore debug info generation settings.
7536
7537 2018-03-01  Martin Liska  <mliska@suse.cz>
7538
7539         PR sanitizer/82484
7540         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7541         volatile arguments.
7542
7543 2018-03-01  Richard Biener  <rguenther@suse.de>
7544
7545         PR debug/84645
7546         * dwarf2out.c (gen_variable_die): Properly handle late VLA
7547         type annotation with LTO when debug was disabled at compile-time.
7548
7549 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
7550
7551         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7552         XINT with INTVAL.
7553         (mips_final_postscan_insn): Likewise.
7554
7555 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
7556
7557         PR rtl-optimization/84528
7558         * alias.c (init_alias_target): Add commentary.
7559         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7560         a unique base value if the frame pointer is not eliminated
7561         to the stack pointer.
7562
7563 2018-03-01  Tom de Vries  <tom@codesourcery.com>
7564
7565         PR rtl-optimization/83327
7566         * lra-int.h (hard_regs_spilled_into): Declare.
7567         * lra.c (hard_regs_spilled_into): Define.
7568         (init_reg_info): Init hard_regs_spilled_into.
7569         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7570         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7571         (process_bb_lives): Handle hard_regs_spilled_into.
7572         (lra_create_live_ranges_1): Before doing liveness propagation, clear
7573         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7574
7575 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
7576
7577         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7578         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7579         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7580         * config/rs6000/aix72.h: New file.
7581
7582 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
7583
7584         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7585         instead of warning_at with conditional singular and plural messages
7586         where possible.
7587
7588         PR target/52991
7589         * stor-layout.c (update_alignment_for_field): For
7590         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7591         && !DECL_PACKED (field), do the alignment update, just use
7592         only desired_align instead of MAX (type_align, desired_align)
7593         as the alignment.
7594         (place_field): Don't do known_align < desired_align handling
7595         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7596         is non-NULL, instead do it after rli->prev_field handling and
7597         only if not within a bitfield word.  For DECL_PACKED (field)
7598         use type_align of BITS_PER_UNIT.
7599
7600 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
7601
7602         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7603         superfluous parentheses and trailing spaces.
7604
7605 2018-02-28  Richard Biener  <rguenther@suse.de>
7606
7607         PR tree-optimization/84584
7608         * graphite-scop-detection.c (scop_detection::add_scop): Discard
7609         SCoPs with fake exit edge.
7610
7611 2018-02-28  Martin Liska  <mliska@suse.cz>
7612
7613         PR testsuite/84597
7614         * timevar.c (timer::print): Fix format to properly print 100%
7615         values.
7616
7617 2018-02-28  Richard Biener  <rguenther@suse.de>
7618
7619         PR middle-end/84607
7620         * genmatch.c (capture_info::walk_match): Do not mark
7621         captured expressions without operands as expr_p given
7622         they act more like predicates and should be subject to
7623         "lost tail" side-effect preserving.
7624
7625 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
7626
7627         PR rtl-optimization/81611
7628         * auto-inc-dec.c (attempt_change): Move dead note from
7629         mem_insn if it's the next use of regno
7630         (find_address): Take address use of reg holding
7631         non-incremented value.  Add parm to limit search to the named
7632         reg only.
7633         (merge_in_block): Attempt to use a mem insn that is the next
7634         use of the original regno.
7635
7636 2018-02-27  Martin Sebor  <msebor@redhat.com>
7637
7638         PR c++/83871
7639         * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7640         * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7641
7642 2018-02-27  Martin Sebor  <msebor@redhat.com>
7643
7644         PR translation/84207
7645         * diagnostic-core.h (warning_n, error_n, inform_n): Change
7646         n argument to unsigned HOST_WIDE_INT.
7647         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7648         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
7649         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7650         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7651
7652 2018-02-27  Richard Biener  <rguenther@suse.de>
7653
7654         PR tree-optimization/84512
7655         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7656         Do not use the estimate returned from record_stmt_cost for
7657         the scalar iteration cost but sum properly using add_stmt_cost.
7658
7659 2018-02-27  Richard Biener  <rguenther@suse.de>
7660
7661         PR tree-optimization/84466
7662         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7663         Adjust last change to less strictly validate use operands.
7664
7665 2018-02-27  Martin Liska  <mliska@suse.cz>
7666
7667         PR gcov-profile/84548
7668         * gcov.c (process_file): Allow partial overlap and consider it
7669         also as group functions.
7670         (output_lines): Properly calculate range of lines for a group.
7671
7672 2018-02-27  Martin Liska  <mliska@suse.cz>
7673
7674         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7675         'ggc' suffixes.  Change first column width.
7676         (timer::print): Fix formatting of the column.
7677
7678 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
7679
7680         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7681         preserve inline entry blocks for the sake of debug inline
7682         entry point markers alone.
7683         (remove_unused_locals): Suggest in comments a better place to
7684         force the preservation of inline entry blocks that are
7685         otherwise unused, but do not preserve them.
7686
7687 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
7688
7689         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7690
7691 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
7692
7693         PR target/84039
7694         * config/i386/constraints.md (Bs): Replace
7695         ix86_indirect_branch_register with
7696         TARGET_INDIRECT_BRANCH_REGISTER.
7697         (Bw): Likewise.
7698         * config/i386/i386.md (indirect_jump): Likewise.
7699         (tablejump): Likewise.
7700         (*sibcall_memory): Likewise.
7701         (*sibcall_value_memory): Likewise.
7702         Peepholes of indirect call and jump via memory: Likewise.
7703         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7704         (*sibcall_value_GOT_32): Likewise.
7705         * config/i386/predicates.md (indirect_branch_operand): Likewise.
7706         (GOT_memory_operand): Likewise.
7707         (call_insn_operand): Likewise.
7708         (sibcall_insn_operand): Likewise.
7709         (GOT32_symbol_operand): Likewise.
7710         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7711
7712 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
7713
7714         PR rtl-optimization/83496
7715         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7716         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
7717         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7718         redundant insn, if any.
7719         (relax_delay_slots): Likewise.
7720         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7721
7722 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
7723
7724         PR tree-optimization/83965
7725         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7726         that grouped statements are part of a reduction chain.  Return
7727         true if the statement is not marked as a reduction itself but
7728         is part of a group.
7729         (vect_recog_dot_prod_pattern): Don't check whether the statement
7730         is part of a group here.
7731         (vect_recog_sad_pattern): Likewise.
7732         (vect_recog_widen_sum_pattern): Likewise.
7733
7734 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
7735
7736         PR debug/84545
7737         * final.c (rest_of_clean_state): Also look for calls inside sequences.
7738
7739 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
7740
7741         PR target/84530
7742         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7743         the bool argument.
7744         (ix86_output_indirect_function_return): New prototype.
7745         (ix86_split_simple_return_pop_internal): Likewise.
7746         * config/i386/i386.c (indirect_return_via_cx): New.
7747         (indirect_return_via_cx_bnd): Likewise.
7748         (indirect_thunk_name): Handle return va CX_REG.
7749         (output_indirect_thunk_function): Create alias for
7750         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7751         (ix86_output_indirect_jmp): Remove the bool argument.
7752         (ix86_output_indirect_function_return): New function.
7753         (ix86_split_simple_return_pop_internal): Likewise.
7754         * config/i386/i386.md (*indirect_jump): Don't pass false
7755         to ix86_output_indirect_jmp.
7756         (*tablejump_1): Likewise.
7757         (simple_return_pop_internal): Change it to define_insn_and_split.
7758         Call ix86_split_simple_return_pop_internal to split it for
7759         -mfunction-return=.
7760         (simple_return_indirect_internal): Call
7761         ix86_output_indirect_function_return instead of
7762         ix86_output_indirect_jmp.
7763
7764 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
7765
7766         PR bootstrap/84405
7767         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7768         memset and value initialization afterwards.
7769
7770 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
7771
7772         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7773
7774 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
7775
7776         PR target/84521
7777         * common/config/aarch64/aarch64-common.c
7778         (aarch_option_optimization_table[]): Switch
7779         off fomit-frame-pointer
7780
7781 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
7782             Chung-Ju Wu  <jasonwucj@gmail.com>
7783
7784         * config/nds32/nds32-multiple.md (load_multiple): Disallow
7785         volatile memory.
7786         (store_multiple): Ditto.
7787
7788 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
7789
7790         * config.gcc: Add --with-cpu support for nds32 target.
7791         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7792         * config/nds32/nds32.opt: Add -mcpu= option.
7793
7794 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
7795
7796         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7797         isel=yes): Warn for these deprecated options.
7798
7799 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
7800
7801         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7802         ISA_2_5_MASKS_EMBEDDED.
7803
7804 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
7805
7806         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7807         p->max as pointers rather than using iterative_hash_expr.
7808
7809 2018-02-23  Carl Love  <cel@us.ibm.com>
7810
7811         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7812         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7813         BU_P8V_OVERLOAD_2.
7814         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7815         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7816         P8V_BUILTIN_VEC_VUNSIGNED2.
7817
7818 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
7819
7820         PR target/81572
7821         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7822         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7823         LRA_UNKNOWN_ALT.
7824         * lra-constraints.c (curr_insn_transform): Set up
7825         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
7826         LRA_UNKNOWN_ALT.
7827         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7828         * lra-eliminations.c (spill_pseudos): Ditto.
7829         (process_insn_for_elimination): Ditto.
7830         * lra-lives.c (reg_early_clobber_p): Use the new macros.
7831         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7832         LRA_NON_CLOBBERED_ALT.
7833
7834 2018-02-22  Martin Sebor  <msebor@redhat.com>
7835
7836         PR tree-optimization/84480
7837         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7838         to maybe_diag_stxncpy_trunc.  Call it.
7839         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7840         from gimple_fold_builtin_strcpy.  Print inlining stack.
7841         (handle_builtin_stxncpy): Print inlining stack.
7842         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7843
7844 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
7845
7846         PR target/84176
7847         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7848         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7849         and -fcheck-pointer-bounds are used together.
7850         (indirect_thunk_prefix): New enum.
7851         (indirect_thunk_need_prefix): New function.
7852         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
7853         "_nt" instead of "_bnd" for NOTRACK prefix.
7854         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7855         (output_indirect_thunk_function): Likewise.
7856         (): Likewise.
7857         (ix86_code_end): Update output_indirect_thunk_function calls.
7858         (ix86_output_indirect_branch_via_reg): Replace
7859         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7860         (ix86_output_indirect_branch_via_push): Likewise.
7861         (ix86_output_function_return): Likewise.
7862         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7863         incompatible with -fcf-protection=branch and
7864         -fcheck-pointer-bounds.
7865
7866 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
7867
7868         PR target/83335
7869         * config/aarch64/aarch64.c (aarch64_print_address_internal):
7870         Change gcc_assert call to output_operand_lossage.
7871
7872 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
7873
7874         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7875
7876 2018-02-22  DJ Delorie  <dj@redhat.com>
7877             Sebastian Perta  <sebastian.perta@renesas.com>
7878             Oleg Endo  <olegendo@gcc.gnu.org>
7879
7880         * config/rx/rx.c (rx_rtx_costs): New function.
7881         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7882
7883 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
7884
7885         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7886
7887 2018-02-22  Martin Liska  <mliska@suse.cz>
7888
7889         PR driver/83193
7890         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7891         Add "native" as a possible value.
7892
7893 2018-02-22  Martin Liska  <mliska@suse.cz>
7894
7895         PR driver/83193
7896         * config/i386/i386.c (ix86_option_override_internal):
7897         Add "native" as a possible value for -march and -mtune.
7898
7899 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
7900
7901         PR target/84502
7902         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7903         to all type variants.
7904
7905         PR tree-optimization/84503
7906         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7907         width as info->bitpos + info->bitsize - start.
7908         (merged_store_group::merge_overlapping): Simplify width computation.
7909         (check_no_overlap): New function.
7910         (imm_store_chain_info::try_coalesce_bswap): Compute expected
7911         start + width and last_order of the group, fail if check_no_overlap
7912         fails.
7913         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7914         to group if check_no_overlap fails.
7915
7916 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
7917
7918         * config/rs6000/altivec.md: Delete contraint arguments to
7919         define_expand, define_split, and define_peephole2, and in
7920         define_insn_and_split if always unused.
7921         * config/rs6000/darwin.md: Ditto.
7922         * config/rs6000/dfp.md: Ditto.
7923         * config/rs6000/rs6000.md: Ditto.
7924         * config/rs6000/sync.md: Ditto.
7925         * config/rs6000/vector.md: Ditto.
7926         * config/rs6000/vsx.md: Ditto.
7927
7928 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
7929
7930         * config/rs6000/altivec.md: Write output control strings as braced
7931         blocks instead of double-quoted strings.
7932         * config/rs6000/darwin.md: Ditto.
7933         * config/rs6000/rs6000.md: Ditto.
7934         * config/rs6000/vector.md: Ditto.
7935         * config/rs6000/vsx.md: Ditto.
7936
7937 2018-02-21  Jason Merrill  <jason@redhat.com>
7938
7939         PR c++/84314 - ICE with templates and fastcall attribute.
7940         * attribs.c (build_type_attribute_qual_variant): Remove assert.
7941
7942 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
7943
7944         * ipa-cp.c (determine_versionability): Fix comment typos.
7945
7946 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
7947
7948         PR c/84229
7949         * ipa-cp.c (determine_versionability): Do not version functions caling
7950         va_arg_pack.
7951
7952 2018-02-21  Martin Liska  <mliska@suse.cz>
7953
7954         PR driver/83193
7955         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7956         Add "native" as a possible value.
7957         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
7958         the macro when native cpu detection is available.
7959
7960 2018-02-21  Martin Liska  <mliska@suse.cz>
7961
7962         PR driver/83193
7963         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7964         Add "native" as a possible value.
7965         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7966         when native cpu detection is available.
7967
7968 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
7969             Martin Sebor  <msebor@redhat.com>
7970
7971         PR tree-optimization/84478
7972         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7973         false.
7974         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7975         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7976         support which is conservatively correct, for 2 only stay conservative
7977         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
7978         argument to the 2 argument get_range_strlen, adjust 6 arg
7979         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7980         false.
7981         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7982         (gimple_fold_builtin_strlen): Pass true as last argument to
7983         get_range_strlen.
7984
7985 2018-02-20  Martin Sebor  <msebor@redhat.com>
7986
7987         PR middle-end/84095
7988         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7989         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
7990         (builtin_memref::builtin_memref): Factor out parts into
7991         set_base_and_offset and call it.
7992
7993 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
7994
7995         PR middle-end/84406
7996         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7997         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7998         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
7999         search at the associated MODE_INT.
8000
8001 2018-02-20  Jeff Law  <law@redhat.com>
8002
8003         PR middle-end/82123
8004         PR tree-optimization/81592
8005         PR middle-end/79257
8006         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
8007         for range data rather than using global data.
8008         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
8009         range data rather than using global data.
8010         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
8011         pass it to children as needed.
8012         (struct directive::fmtresult): Similarly.
8013         (struct directive::set_width): Similarly.
8014         (struct directive::set_precision): Similarly.
8015         (format_integer, format_directive, parse_directive): Similarly.
8016         (format_none): Accept unnamed vr_values parameter.
8017         (format_percent, format_floating, format_character): Similarly.
8018         (format_string, format_plain): Similarly.
8019         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
8020         the EVRP range analyzer for range data rather than using global data.
8021         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
8022         gimple-ssa-evrp-analyze.h
8023         (class sprintf_dom_walker): Add after_dom_children member function.
8024         Add evrp_range_analyzer member.
8025         (sprintf_dom_walker::before_dom_children): Call into the EVRP
8026         range analyzer as needed.
8027         (sprintf_dom_walker::after_dom_children): New member function.
8028         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
8029         if not optimizing.
8030         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
8031         (evrp_range_analyzer::pop_to_marker): Likewise.
8032
8033 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
8034
8035         PR tree-optimization/84419
8036         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
8037         with the required type if its current type is compatible but
8038         different.
8039
8040 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
8041
8042         PR middle-end/82004
8043         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
8044         after vectorization.
8045
8046 2018-02-20  Martin Liska  <mliska@suse.cz>
8047
8048         PR driver/83193
8049         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
8050         possible values if we don't have a hint.
8051
8052 2018-02-20  Martin Liska  <mliska@suse.cz>
8053
8054         PR c/84310
8055         PR target/79747
8056         * final.c (shorten_branches): Build align_tab array with one
8057         more element.
8058         * opts.c (finish_options): Add alignment option limit check.
8059         (MAX_CODE_ALIGN): Likewise.
8060         (MAX_CODE_ALIGN_VALUE): Likewise.
8061         * doc/invoke.texi: Document maximum allowed option value for
8062         all -falign-* options.
8063
8064 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
8065
8066         PR target/84146
8067         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
8068         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
8069         * var-tracking.c (emit_note_insn_var_location): Remove all references
8070         to NOTE_INSN_CALL_ARG_LOCATION.
8071         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
8072         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
8073         Use copy_rtx_if_shared.
8074         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
8075         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
8076         (dwarf2out_var_location): Remove handling of
8077         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
8078         on call_insn.
8079         * final.c (final_scan_insn): Remove all references to
8080         NOTE_INSN_CALL_ARG_LOCATION.
8081         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
8082         before dumping final insns.
8083         * except.c (emit_note_eh_region_end): Remove all references to
8084         NOTE_INSN_CALL_ARG_LOCATION.
8085         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
8086         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
8087         * config/arc/arc.c (hwloop_optimize): Likewise.
8088         * config/arm/arm.c (create_fix_barrier): Likewise.
8089         * config/s390/s390.c (s390_chunkify_start): Likewise.
8090         * config/sh/sh.c (find_barrier): Likewise.
8091         * config/i386/i386.c (rest_of_insert_endbranch,
8092         ix86_seh_fixup_eh_fallthru): Likewise.
8093         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
8094         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
8095         * config/frv/frv.c (frv_function_prologue): Likewise.
8096         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
8097         reg note.
8098         (note_outside_basic_block_p): Remove all references to
8099         NOTE_INSN_CALL_ARG_LOCATION.
8100         * gengtype.c (adjust_field_rtx_def): Likewise.
8101         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
8102         Likewise.
8103         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
8104         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
8105
8106         PR c++/84444
8107         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
8108         is ADDR_EXPR.
8109
8110         PR tree-optimization/84452
8111         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
8112         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
8113         is NULL.
8114
8115 2018-02-19  Martin Liska  <mliska@suse.cz>
8116
8117         PR sanitizer/82183
8118         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
8119
8120 2018-02-19  Martin Liska  <mliska@suse.cz>
8121             Richard Sandiford  <richard.sandiford@linaro.org>
8122
8123         PR tree-optimization/82491
8124         * gimple-fold.c (get_base_constructor): Make earlier bail out
8125         to prevent ubsan.
8126
8127 2018-02-19  Carl Love  <cel@us.ibm.com>
8128
8129         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
8130         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
8131         BU_P8V_OVERLOAD_1.
8132         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
8133         P8V_BUILTIN_VEC_NEG.
8134
8135 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
8136
8137         * config/rl78/rl78.md (movdf): New define expand.
8138
8139 2018-02-19  Martin Liska  <mliska@suse.cz>
8140
8141         PR other/80589
8142         * doc/invoke.texi: Fix typo.
8143         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
8144
8145 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
8146
8147         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8148         handle rs6000_single_float and rs6000_double_float specially for
8149         e500 family CPUs.
8150
8151 2018-02-16  Jeff Law  <law@redhat.com>
8152
8153         * config/rx/rx.c (add_pop_cfi_notes): New function.;
8154         (pop_regs): Use it.
8155
8156 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
8157
8158         PR ipa/84425
8159         * ipa-inline.c (inline_small_functions): Fix a typo.
8160
8161 2018-02-16  Nathan Sidwell  <nathan@acm.org>
8162
8163         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
8164
8165 2018-02-16  Carl Love  <cel@us.ibm.com>
8166
8167         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
8168         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
8169         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
8170         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
8171         expansion to P8V_BUILTIN_VEC_FLOAT2.
8172
8173 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
8174
8175         PR rtl-optimization/70023
8176         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
8177         src_regno into account.
8178
8179 2018-02-16  Carl Love  <cel@us.ibm.com>
8180
8181         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
8182         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
8183         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
8184         * config/rs6000/rs6000.c: Remove case statements for
8185         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
8186         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
8187         and P9V_BUILTIN_VEC_VINSERT4B.
8188         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
8189         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
8190         * config/rs6000/vsx.md:
8191         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
8192         vec_insert4b.
8193
8194 2018-02-16  Carl Love  <cel@us.ibm.com>
8195
8196         * config/rs6000/altivec.h: Add builtin names vec_extract4b
8197         vec_insert4b.
8198         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
8199         definitions.
8200         * config/rs6000/rs6000-c.c: Add the definitions for
8201         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
8202         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
8203         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
8204         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
8205         definition for insert4b and define insn *insert3b_internal.
8206         * doc/extend.texi: Add documentation for vec_extract4b.
8207
8208 2018-02-16  Nathan Sidwell  <nathan@acm.org>
8209
8210         * doc/extend.texi (Backwards Compatibility): Mention friend
8211         injection.  Note for-scope is deprecated.
8212         * doc/invoke.texi (-ffriend-injection): Deprecate.
8213
8214 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
8215
8216         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
8217         that moved to I2, also allow destinations that are a paradoxical
8218         subreg (instead of a normal reg).
8219
8220 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
8221
8222         PR target/83831
8223         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8224         to QImode.
8225
8226 2018-02-16  Richard Biener  <rguenther@suse.de>
8227
8228         PR tree-optimization/84037
8229         PR tree-optimization/84016
8230         PR target/82862
8231         * config/i386/i386.c (ix86_builtin_vectorization_cost):
8232         Adjust vec_construct for the fact we need additional higher latency
8233         128bit inserts for AVX256 and AVX512 vector builds.
8234         (ix86_add_stmt_cost): Scale vector construction cost for
8235         elementwise loads.
8236
8237 2018-02-16  Richard Biener  <rguenther@suse.de>
8238
8239         PR tree-optimization/84417
8240         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8241         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8242         (non_rewritable_lvalue_p): Likewise, use poly-ints.
8243
8244 2018-02-16  Martin Liska  <mliska@suse.cz>
8245
8246         PR sanitizer/84307
8247         * internal-fn.def (ASAN_CHECK): Set proper flags.
8248         (ASAN_MARK): Likewise.
8249
8250 2018-02-16  Julia Koval  <julia.koval@intel.com>
8251
8252         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8253         from PTA_CANNONLAKE.
8254
8255 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
8256
8257         PR target/84272
8258         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8259         Use ++iter rather than iter++ for std::list iterators.
8260         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
8261         defer deleting them until all nodes in the forest are processed.  Do
8262         free even leaf nodes.  Change to_process into auto_vec.
8263
8264         PR bootstrap/84405
8265         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8266         * vec.h (vec_default_construct): Use memset instead of placement new
8267         if BROKEN_VALUE_INITIALIZATION is defined.
8268         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8269         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8270         is defined.
8271
8272         PR rtl-optimization/83723
8273         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8274         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
8275         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
8276         recursive calls.
8277         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8278         callers.
8279         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8280
8281 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
8282
8283         PR rtl-optimization/81443
8284         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8285         from inner REGs to paradoxical SUBREGs.
8286
8287 2018-02-16  Richard Biener  <rguenther@suse.de>
8288
8289         PR tree-optimization/84399
8290         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8291         For operands we can analyze at their definition make sure we can
8292         analyze them at each use as well.
8293
8294 2018-02-16  Richard Biener  <rguenther@suse.de>
8295
8296         PR tree-optimization/84190
8297         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8298         volatile accesses if the decl isn't volatile.
8299
8300 2018-02-15  Jason Merrill  <jason@redhat.com>
8301
8302         PR c++/84314 - ICE with templates and fastcall attribute.
8303         * attribs.c (build_type_attribute_qual_variant): Don't clobber
8304         TYPE_CANONICAL on an existing type.
8305
8306 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
8307
8308         PR tree-optimization/84383
8309         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8310         dstoff nor call operand_equal_p if dstbase is NULL.
8311
8312         PR tree-optimization/84334
8313         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8314         also a CONSTANT_CLASS_P, punt.
8315
8316 2018-02-14  Jim Wilson  <jimw@sifive.com>
8317
8318         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8319         first SMALL_OPERAND check.  New local min_second_step.  Move assert
8320         to where locals are set.  Add TARGET_RVC support.
8321         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8322
8323 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
8324
8325         * doc/invoke.texi: Correct -Wformat-overflow code sample.
8326
8327 2018-02-14  Martin Sebor  <msebor@redhat.com>
8328
8329         PR tree-optimization/83698
8330         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8331         arrays constrain the offset range to their bounds.
8332         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8333         (builtin_access::overlap): Avoid setting the size of overlap if it's
8334         already been set.
8335         (maybe_diag_overlap): Also consider arrays when deciding what values
8336         of offsets to include in diagnostics.
8337
8338 2018-02-14  Martin Sebor  <msebor@redhat.com>
8339
8340         PR c/84108
8341         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8342         that correspond to the kind of a declaration.
8343
8344 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
8345
8346         PR target/83984
8347         * config/pa/pa.md: Load address of PIC label using the linkage table
8348         if the label is nonlocal.
8349
8350 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8351
8352         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8353         warning message if user requests -maltivec=be.
8354         * doc/invoke.texi: Document deprecation of -maltivec=be.
8355
8356 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
8357
8358         PR target/84220
8359         * config/rs6000/rs6000-c.c: Update definitions for
8360         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8361         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8362
8363 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
8364
8365         PR target/84239
8366         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8367         add _get_ssp intrinsics. Remove argument from
8368         __builtin_ia32_rdssp[d|q].
8369         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8370         * config/i386/i386-builtin.def: Remove argument from
8371         __builtin_ia32_rdssp[d|q].
8372         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8373         ix86_expand_special_args_builtin for _rdssp[d|q].
8374         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8375         Clear register before usage.
8376         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8377         Add documentation for new _get_ssp and _inc_ssp intrinsics.
8378
8379 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
8380
8381         PR tree-optimization/84357
8382         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8383         operand 1 of an ARRAY_REF too.
8384
8385 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
8386
8387         PR target/83831
8388         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8389         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8390         declarations.
8391         (set_of_reg): New struct.
8392         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8393         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8394         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8395         functions.
8396         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8397         Split into bitclr, bitset, bitinvert patterns if appropriate.
8398         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8399         use rx_fuse_in_memory_bitop.
8400         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8401         to named insn, correct maximum insn length.
8402
8403 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
8404
8405         PR target/79242
8406         * machmode.def: Define a complex mode for PARTIAL_INT.
8407         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8408         MODE_PARTIAL_INT.
8409         * doc/rtl.texi: Document CSPImode.
8410         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8411         handling.
8412         (msp430_hard_regno_nregs_with_padding): Likewise.
8413
8414 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
8415
8416         PR target/84279
8417         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8418
8419 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
8420
8421         PR rtl-optimization/84169
8422         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
8423         we generated a parallel as new i3 and we split that to new i2 and i3
8424         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
8425         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
8426         those to i2, not i1.  Partially rewrite this scan code.
8427
8428 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
8429
8430         PR c/82210
8431         * stor-layout.c (place_field): For variable length fields, adjust
8432         offset_align afterwards not just based on the field's alignment,
8433         but also on the size.
8434
8435         PR middle-end/84309
8436         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8437         of exps and logs in the use_exp2 case.
8438
8439 2018-02-13  Jeff Law  <law@redhat.com>
8440
8441         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8442         entry for "vector".
8443
8444         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8445         ARGS as unused.
8446
8447 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
8448
8449         PR debug/84342
8450         PR debug/84319
8451         * common.opt (gas-loc-support, gas-locview-support): New.
8452         (ginline-points, ginternal-reset-location-views): New.
8453         * doc/invoke.texi: Document them.  Use @itemx where intended.
8454         (gvariable-location-views): Adjust.
8455         * target.def (reset_location_view): New.
8456         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8457         (TARGET_RESET_LOCATION_VIEW): New.
8458         * doc/tm.texi: Rebuilt.
8459         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8460         (dwarf2out_default_as_locview_support): New.
8461         (output_asm_line_debug_info): Use option variables.
8462         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8463         (output_loc_list): Likewise.
8464         (add_high_low_attributes): Check option variables.
8465         Don't output entry view attribute in strict mode.
8466         (gen_inlined_subroutine_die): Check option variables.
8467         (dwarf2out_inline_entry): Likewise.
8468         (init_sections_and_labels): Likewise.
8469         (dwarf2out_early_finish): Likewise.
8470         (maybe_reset_location_view): New, from...
8471         (dwarf2out_var_location): ... here.  Call it.
8472         * debug.h (dwarf2out_default_as_loc_support): Declare.
8473         (dwarf2out_default_as_locview_support): Declare.
8474         * hooks.c (hook_int_rtx_insn_0): New.
8475         * hooks.h (hook_int_rtx_insn_0): Declare.
8476         * toplev.c (process_options): Take -gas-loc-support and
8477         -gas-locview-support from dwarf2out.  Enable
8478         -gvariable-location-views by default only with locview
8479         assembler support.  Enable -ginternal-reset-location-views by
8480         default only if the target defines the corresponding hook.
8481         Enable -ginline-points by default if location views are
8482         enabled; force it disabled if statement frontiers are
8483         disabled.
8484         * tree-inline.c (expand_call_inline): Check option variables.
8485         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8486
8487 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
8488
8489         PR tree-optimization/84321
8490         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8491         handling.  Also check whether the anti-range contains any values
8492         that satisfy the mask; switch to a VR_RANGE if not.
8493
8494 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
8495
8496         PR sanitizer/84340
8497         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8498
8499 2018-02-13  Martin Jambor  <mjambor@suse.cz>
8500
8501         PR c++/83990
8502         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8503         of call statements, also set location of a load to a temporary.
8504
8505 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
8506
8507         * config/rl78/rl78.c (add_vector_labels): New function.
8508         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8509         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8510         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8511         which checks that no arguments are passed.
8512         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8513         * doc/extend.texi: Documentation for the new attribute.
8514
8515 2018-02-13  Andreas Schwab  <schwab@suse.de>
8516
8517         * config/riscv/linux.h (CPP_SPEC): Define.
8518
8519 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
8520
8521         PR target/84335
8522         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8523         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8524         OPTION_MASK_ISA_AES as first argument to def_builtin_const
8525         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8526         instead of OPTION_MASK_ISA_PCLMUL as first argument to
8527         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8528         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8529         temporarily for AES and PCLMUL builtins.
8530
8531         PR tree-optimization/84339
8532         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8533         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8534         Formatting fixes.
8535
8536         PR middle-end/84309
8537         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8538         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8539         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8540         inline function.
8541         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8542         inline function.
8543         * omp-simd-clone.h: New file.
8544         * omp-simd-clone.c: Include omp-simd-clone.h.
8545         (expand_simd_clones): No longer static.
8546         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8547         cgraph.h and omp-simd-clone.h.
8548         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8549         (vect_recog_widen_shift_pattern): Formatting fix.
8550         (vect_pattern_recog_1): Don't check optab for calls.
8551
8552         PR target/84336
8553         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8554         operands[2] into a REG before using gen_lowpart on it.
8555
8556 2018-02-12  Jeff Law  <law@redhat.com>
8557
8558         PR target/83760
8559         * config/sh/sh.c (find_barrier): Consider a sibling call
8560         a barrier as well.
8561
8562         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8563         successfully back substituting a reg.
8564
8565 2018-02-12  Richard Biener  <rguenther@suse.de>
8566
8567         PR tree-optimization/84037
8568         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8569         parameter, move visited init to caller.
8570         (vect_slp_analyze_operations): Separate cost from validity
8571         check, initialize visited once for all instances.
8572         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8573         for all instances.
8574         * tree-vect-stmts.c (vect_model_simple_cost): Make early
8575         out an assert.
8576         (vect_model_promotion_demotion_cost): Likewise.
8577         (vectorizable_bswap): Guard cost modeling with !slp_node
8578         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8579         SLP stmts.
8580         (vectorizable_call): Likewise.
8581         (vectorizable_conversion): Likewise.
8582         (vectorizable_assignment): Likewise.
8583         (vectorizable_shift): Likewise.
8584         (vectorizable_operation): Likewise.
8585         (vectorizable_store): Likewise.
8586         (vectorizable_load): Likewise.
8587         (vectorizable_condition): Likewise.
8588         (vectorizable_comparison): Likewise.
8589
8590 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
8591
8592         PR sanitizer/84307
8593         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8594         (ASAN_MARK): Fix fnspec to account for return value, change pointer
8595         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8596
8597 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
8598
8599         PR middle-end/83665
8600         * params.def (inline-min-speedup): Increase from 8 to 15.
8601         (max-inline-insns-auto): Decrease from 40 to 30.
8602         * ipa-split.c (consider_split): Add some buffer for function to
8603         be considered inlining candidate.
8604         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8605         default values.
8606
8607 2018-02-12  Richard Biener  <rguenther@suse.de>
8608
8609         PR tree-optimization/84037
8610         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8611         matched stmts if we cannot swap the non-matched ones.
8612
8613 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
8614
8615         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8616         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8617         _mm_maskz_scalef_round_ss): New intrinsics.
8618         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8619         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8620         __builtin_ia32_scalefss_round): Remove.
8621         (__builtin_ia32_scalefsd_mask_round,
8622         __builtin_ia32_scalefss_mask_round): New intrinsics.
8623         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8624         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8625         ((match_operand:VF_128 2 "<round_nimm_predicate>"
8626         "<round_constraint>")): Changed to ...
8627         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8628         "<round_scalar_constraint>")): ... this.
8629         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8630         %0, %1, %2<round_op3>}"): Changed to ...
8631         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8632         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8633         %2<round_scalar_mask_op3>}"): ... this.
8634         * config/i386/subst.md (round_scalar_nimm_predicate): New.
8635
8636 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
8637
8638         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8639         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8640         (_mm_maskz_sqrt_round_ss): New intrinsics.
8641         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8642         (__builtin_ia32_sqrtsd_mask_round)
8643         (__builtin_ia32_sqrtss_mask_round): New builtins.
8644         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8645         (__builtin_ia32_sqrtss_round): Remove.
8646         (__builtin_ia32_sqrtsd_mask_round)
8647         (__builtin_ia32_sqrtss_mask_round): New builtins.
8648         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8649         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8650         ((match_operand:VF_128 1 "vector_operand"
8651         "xBm,<round_constraint>")): Changed to ...
8652         ((match_operand:VF_128 1 "vector_operand"
8653         "xBm,<round_scalar_constraint>")): ... this.
8654         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8655         %0, %2, %<iptr>1<round_op3>}): Changed to ...
8656         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8657         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8658         %<iptr>1<round_scalar_mask_op3>}): ... this.
8659         ((set_attr "prefix" "<round_prefix>")): Changed to ...
8660         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8661
8662 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
8663
8664         PR target/84266
8665         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8666         Cast vec_cmpeq result to correct type.
8667         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8668         Cast vec_cmpgt result to correct type.
8669
8670 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
8671
8672         * final.c (final_scan_insn_1): Renamed from...
8673         (final_scan_insn): ... this.  New wrapper, to recover
8674         seen from the outermost call in recursive ones.
8675         * config/sparc/sparc.c (output_return): Drop seen from call.
8676         (output_sibcall): Likewise.
8677         * config/visium/visium.c (output_branch): Likewise.
8678
8679 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
8680
8681         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8682         function label.
8683
8684 2018-02-10  Alan Modra  <amodra@gmail.com>
8685
8686         PR target/84300
8687         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8688         Specify LR as an input.
8689
8690 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
8691
8692         PR sanitizer/83987
8693         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8694         remove_member_access_dummy_vars): New functions.
8695         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8696         lower_omp_1, execute_lower_omp): Use them.
8697
8698         PR rtl-optimization/84308
8699         * shrink-wrap.c (spread_components): Release todo vector.
8700
8701 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
8702
8703         PR rtl-optimization/57193
8704         * ira-color.c (struct allocno_color_data): Add member
8705         conflict_allocno_hard_prefs.
8706         (update_conflict_allocno_hard_prefs): New.
8707         (bucket_allocno_compare_func): Add a preference based on
8708         conflict_allocno_hard_prefs.
8709         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8710         (color_allocnos): Remove a dead code.  Initiate
8711         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
8712
8713 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
8714
8715         PR target/84226
8716         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8717         constraint from =wa to wa.  Avoid a subreg on the output operand,
8718         instead use a pseudo and subreg it in a move.
8719         (p9_xxbrd_<mode>): Changed to ...
8720         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8721         (p9_xxbrd_v2df): New expander.
8722         (p9_xxbrw_<mode>): Changed to ...
8723         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8724         (p9_xxbrw_v4sf): New expander.
8725
8726 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
8727
8728         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8729
8730 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
8731
8732         PR target/83926
8733         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8734         multiply in 32-bit mode.
8735         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8736         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8737         mode.
8738
8739 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
8740
8741         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8742         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8743         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8744         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8745
8746 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
8747
8748         PR lto/84213
8749         * dwarf2out.c (is_trivial_indirect_ref): New function.
8750         (dwarf2out_late_global_decl): Do not generate a location
8751         attribute for variables that have a non-trivial DECL_VALUE_EXPR
8752         and that are not defined in the current unit.
8753
8754 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
8755
8756         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8757         instead of a libcall for UNORDERED.
8758
8759 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
8760
8761         PR target/82641
8762         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8763         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8764
8765 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
8766
8767         PR target/PR84295
8768         * config/s390/s390.c (s390_set_current_function): Invoke
8769         s390_indirect_branch_settings also if fndecl didn't change.
8770
8771 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
8772
8773         * config/rs6000/rs6000.md (blockage): Set length to zero.
8774
8775 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
8776
8777         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8778
8779 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
8780
8781         PR sanitizer/84285
8782         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8783         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8784         -static-lib*san.
8785
8786         PR debug/84252
8787         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8788         PARALLEL incoming that failed vt_get_decl_and_offset check.
8789
8790         PR middle-end/84237
8791         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8792         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8793         TREE_READONLY bit.
8794         (get_variable_section): For decls in named .bss* sections pass true as
8795         second argument to bss_initializer_p.
8796
8797 2018-02-09  Marek Polacek  <polacek@redhat.com>
8798             Jakub Jelinek  <jakub@redhat.com>
8799
8800         PR c++/83659
8801         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8802         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
8803         Sync some changes from cxx_fold_indirect_ref.
8804
8805 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
8806
8807         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8808         markers.
8809         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8810         (BLOCK_INLINE_ENTRY_LABEL): New.
8811         (dwarf2out_var_location): Disregard inline entry markers.
8812         (inline_entry_data): New struct.
8813         (inline_entry_data_hasher): New hashtable type.
8814         (inline_entry_data_hasher::hash): New.
8815         (inline_entry_data_hasher::equal): New.
8816         (inline_entry_data_table): New variable.
8817         (add_high_low_attributes): Add DW_AT_entry_pc and
8818         DW_AT_GNU_entry_view attributes if a pending entry is found
8819         in inline_entry_data_table.  Add old entry_pc attribute only
8820         if debug nonbinding markers are disabled.
8821         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8822         markers are enabled.
8823         (block_within_block_p, dwarf2out_inline_entry): New.
8824         (dwarf2out_finish): Check that no entries remained in
8825         inline_entry_data_table.
8826         * final.c (reemit_insn_block_notes): Handle inline entry notes.
8827         (final_scan_insn, notice_source_line): Likewise.
8828         (rest_of_clean_state): Skip inline entry markers.
8829         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8830         markers.
8831         * gimple.c (gimple_build_debug_inline_entry): New.
8832         * gimple.h (enum gimple_debug_subcode): Add
8833         GIMPLE_DEBUG_INLINE_ENTRY.
8834         (gimple_build_debug_inline_entry): Declare.
8835         (gimple_debug_inline_entry_p): New.
8836         (gimple_debug_nonbind_marker_p): Adjust.
8837         * insn-notes.def (INLINE_ENTRY): New.
8838         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8839         inline entry marker notes.
8840         (print_insn): Likewise.
8841         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8842         (INSN_DEBUG_MARKER_KIND): Likewise.
8843         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8844         * tree-inline.c (expand_call_inline): Build and insert
8845         debug_inline_entry stmt.
8846         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8847         inline entry blocks early, if nonbind markers are enabled.
8848         (dump_scope_block): Dump fragment info.
8849         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8850         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8851         (gimple_build_debug_inline_entry): New.
8852         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8853         Enable/disable inline entry points too.
8854         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8855         (DEBUG_INSN): Describe inline entry markers.
8856
8857         * common.opt (gvariable-location-views): New.
8858         (gvariable-location-views=incompat5): New.
8859         * config.in: Rebuilt.
8860         * configure: Rebuilt.
8861         * configure.ac: Test assembler for view support.
8862         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8863         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8864         * dwarf2out.c (var_loc_view): New typedef.
8865         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8866         (dwarf2out_locviews_in_attribute): New.
8867         (dwarf2out_locviews_in_loclist): New.
8868         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8869         (enum dw_line_info_opcode): Add LI_adv_address.
8870         (struct dw_line_info_table): Add view.
8871         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8872         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8873         (zero_view_p): New variable.
8874         (ZERO_VIEW_P): New macro.
8875         (output_asm_line_debug_info): New.
8876         (struct var_loc_node): Add view.
8877         (add_AT_view_list, AT_loc_list): New.
8878         (add_var_loc_to_decl): Add view param.  Test it against last.
8879         (new_loc_list): Add view params.  Record them.
8880         (AT_loc_list_ptr): Handle loc and view lists.
8881         (view_list_to_loc_list_val_node): New.
8882         (print_dw_val): Handle dw_val_class_view_list.
8883         (size_of_die): Likewise.
8884         (value_format): Likewise.
8885         (loc_list_has_views): New.
8886         (gen_llsym): Set vl_symbol too.
8887         (maybe_gen_llsym, skip_loc_list_entry): New.
8888         (dwarf2out_maybe_output_loclist_view_pair): New.
8889         (output_loc_list): Output view list or entries too.
8890         (output_view_list_offset): New.
8891         (output_die): Handle dw_val_class_view_list.
8892         (output_dwarf_version): New.
8893         (output_compilation_unit_header): Use it.
8894         (output_skeleton_debug_sections): Likewise.
8895         (output_rnglists, output_line_info): Likewise.
8896         (output_pubnames, output_aranges): Update version comments.
8897         (output_one_line_info_table): Output view numbers in asm comments.
8898         (dw_loc_list): Determine current endview, pass it to new_loc_list.
8899         Call maybe_gen_llsym.
8900         (loc_list_from_tree_1): Adjust.
8901         (add_AT_location_description): Create view list attribute if
8902         needed, check it's absent otherwise.
8903         (convert_cfa_to_fb_loc_list): Adjust.
8904         (maybe_emit_file): Call output_asm_line_debug_info for test.
8905         (dwarf2out_var_location): Reset views as needed.  Precompute
8906         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
8907         attribute.  Set view.
8908         (new_line_info_table): Reset next view.
8909         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8910         (dwarf2out_source_line): Likewise.  Output view resets and labels to
8911         the assembler, or select appropriate line info opcodes.
8912         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8913         (optimize_string_length): Catch it.  Adjust.
8914         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
8915         dw_val_class_view_list, and remove it if no longer needed.
8916         (hash_loc_list): Hash view numbers.
8917         (loc_list_hasher::equal): Compare them.
8918         (optimize_location_lists): Check whether a view list symbol is
8919         needed, and whether the locview attribute is present, and
8920         whether they match.  Remove the locview attribute if no longer
8921         needed.
8922         (index_location_lists): Call skip_loc_list_entry for test.
8923         (dwarf2out_finish): Call output_asm_line_debug_info for test.
8924         Use output_dwarf_version.
8925         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8926         (struct dw_val_node): Add val_view_list.
8927         * final.c (SEEN_NEXT_VIEW): New.
8928         (set_next_view_needed): New.
8929         (clear_next_view_needed): New.
8930         (maybe_output_next_view): New.
8931         (final_start_function): Rename to...
8932         (final_start_function_1): ... this.  Take pointer to FIRST,
8933         add SEEN parameter.  Emit param bindings in the initial view.
8934         (final_start_function): Reintroduce SEEN-less interface.
8935         (final): Rename to...
8936         (final_1): ... this.  Take SEEN parameter.  Output final pending
8937         next view at the end.
8938         (final): Reintroduce seen-less interface.
8939         (final_scan_insn): Output pending next view before switching
8940         sections or ending a block.  Mark the next view as needed when
8941         outputting variable locations.  Notify debug backend of section
8942         changes, and of location view changes.
8943         (rest_of_handle_final): Adjust.
8944         * toplev.c (process_options): Autodetect value for debug variable
8945         location views option.  Warn on incompat5 without -gdwarf-5.
8946         * doc/invoke.texi (gvariable-location-views): New.
8947         (gvariable-location-views=incompat5): New.
8948         (gno-variable-location-views): New.
8949
8950 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
8951
8952         PR tree-optimization/84136
8953         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8954         that the result of find_edge is non-NULL.
8955
8956 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
8957
8958         PR target/83008
8959         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8960         storing integer register in SImode.  Fix cost of 256 and 512
8961         byte aligned SSE register store.
8962
8963 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
8964
8965         * config/i386/i386.c (ix86_multiplication_cost): Fix
8966         multiplication cost for TARGET_AVX512DQ.
8967
8968 2018-02-08  Marek Polacek  <polacek@redhat.com>
8969
8970         PR tree-optimization/84238
8971         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8972         get_range_strlen.
8973
8974 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
8975
8976         PR tree-optimization/84265
8977         * tree-vect-stmts.c (vectorizable_store): Don't treat
8978         VMAT_CONTIGUOUS accesses as grouped.
8979         (vectorizable_load): Likewise.
8980
8981 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
8982
8983         PR tree-optimization/81635
8984         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8985         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8986         (test_round_for_mask): New functions.
8987         (wide_int_cc_tests): Call test_round_for_mask.
8988         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8989         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8990         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8991         range returned by get_range_info.
8992
8993 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
8994
8995         PR ipa/81360
8996         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8997         * symtab.c: Include builtins.h
8998         (symtab_node::output_to_lto_symbol_table_p): Move here
8999         from lto-streamer-out.c:output_symbol_p.
9000         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
9001         (output_symbol_p): Move all logic to symtab.c
9002         (produce_symtab): Update.
9003
9004 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
9005
9006         * config/s390/s390-opts.h (enum indirect_branch): Define.
9007         * config/s390/s390-protos.h (s390_return_addr_from_memory)
9008         (s390_indirect_branch_via_thunk)
9009         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
9010         (enum s390_indirect_branch_type): Define.
9011         * config/s390/s390.c (struct s390_frame_layout, struct
9012         machine_function): Remove.
9013         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
9014         (indirect_branch_table_label_no, indirect_branch_table_name):
9015         Define variables.
9016         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
9017         (enum s390_indirect_branch_option): Define.
9018         (s390_return_addr_from_memory): New function.
9019         (s390_handle_string_attribute): New function.
9020         (s390_attribute_table): Add new attribute handler.
9021         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
9022         (s390_indirect_branch_via_thunk): New function.
9023         (s390_indirect_branch_via_inline_thunk): New function.
9024         (s390_function_ok_for_sibcall): When jumping via thunk disallow
9025         sibling call optimization for non z10 compiles.
9026         (s390_emit_call): Force indirect branch target to be a single
9027         register.  Add r1 clobber for non-z10 compiles.
9028         (s390_emit_epilogue): Emit return jump via return_use expander.
9029         (s390_reorg): Handle JUMP_INSNs as execute targets.
9030         (s390_option_override_internal): Perform validity checks for the
9031         new command line options.
9032         (s390_indirect_branch_attrvalue): New function.
9033         (s390_indirect_branch_settings): New function.
9034         (s390_set_current_function): Invoke s390_indirect_branch_settings.
9035         (s390_output_indirect_thunk_function):  New function.
9036         (s390_code_end): Implement target hook.
9037         (s390_case_values_threshold): Implement target hook.
9038         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
9039         macros.
9040         * config/s390/s390.h (struct s390_frame_layout)
9041         (struct machine_function): Move here from s390.c.
9042         (TARGET_INDIRECT_BRANCH_NOBP_RET)
9043         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
9044         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
9045         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
9046         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
9047         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
9048         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
9049         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
9050         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
9051         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
9052         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
9053         (mnemonic attribute): Add values which aren't recognized
9054         automatically.
9055         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
9056         pattern for branch conversion.  Fix mnemonic attribute.
9057         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
9058         indirect branch via thunk if requested.
9059         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
9060         ("*indirect_jump"): Disable for branch conversion using out of
9061         line thunks.
9062         ("indirect_jump_via_thunk<mode>_z10")
9063         ("indirect_jump_via_thunk<mode>")
9064         ("indirect_jump_via_inlinethunk<mode>_z10")
9065         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
9066         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
9067         ("casesi_jump_via_inlinethunk<mode>_z10")
9068         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
9069         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
9070         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
9071         ("*indirect2_jump"): Disable for branch conversion.
9072         ("casesi_jump"): Turn into expander and expand patterns for branch
9073         conversion.
9074         ("return_use"): New expander.
9075         ("*return"): Emit return via thunk and rename it to ...
9076         ("*return<mode>"): ... this one.
9077         * config/s390/s390.opt: Add new options and and enum for the
9078         option values.
9079
9080 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
9081
9082         * lra-constraints.c (match_reload): Unconditionally use
9083         gen_lowpart_SUBREG, rather than selecting between that
9084         and equivalent gen_rtx_SUBREG code.
9085
9086 2018-02-08  Richard Biener  <rguenther@suse.de>
9087
9088         PR tree-optimization/84233
9089         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
9090         changed flag instead of boguously re-using phi_inserted.
9091
9092 2018-02-08  Martin Jambor  <mjambor@suse.cz>
9093
9094         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
9095         static local variables.
9096
9097 2018-02-08  Richard Biener  <rguenther@suse.de>
9098
9099         PR tree-optimization/84278
9100         * tree-vect-stmts.c (vectorizable_store): When looking for
9101         smaller vector types to perform grouped strided loads/stores
9102         make sure the mode is supported by the target.
9103         (vectorizable_load): Likewise.
9104
9105 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
9106
9107         * config/aarch64/aarch64.c (aarch64_components_for_bb):
9108         Increase LDP/STP opportunities by adding adjacent callee-saves.
9109
9110 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
9111
9112         PR rtl-optimization/84068
9113         PR rtl-optimization/83459
9114         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
9115
9116 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
9117
9118         PR tree-optimization/84224
9119         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
9120         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
9121         non-zero arguments.
9122
9123 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
9124
9125         PR target/84113
9126         * config/rs6000/altivec.md (*restore_world): Remove LR use.
9127         * config/rs6000/predicates.md (restore_world_operation): Adjust op
9128         count, remove one USE.
9129
9130 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
9131
9132         * doc/install.texi (Configuration): Document the
9133         --with-long-double-format={ibm,ieee} PowerPC configuration
9134         options.
9135
9136         PR target/84154
9137         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
9138         Convert from define_expand to be define_insn_and_split.  Rework
9139         float/double/_Float128 conversions to QI/HI/SImode to work with
9140         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
9141         conversions to QI/HImode types did a store and then a load to
9142         truncate the value.  For conversions to VSX registers, don't split
9143         the insn, instead emit the code directly.  Use the code iterator
9144         any_fix to combine signed and unsigned conversions.
9145         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
9146         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
9147         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9148         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
9149         (fix_<mode>di2_hw): Likewise.
9150         (fixuns_<mode>di2_hw): Likewise.
9151         (fix_<mode>si2_hw): Likewise.
9152         (fixuns_<mode>si2_hw): Likewise.
9153         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
9154         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9155         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
9156         fix<uns>_trunc<SFDF:mode>si2_p8.
9157         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
9158         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
9159         (fix<uns>_<mode>_mem): Likewise.
9160         (fctiw<u>z_<mode>_mem): Likewise.
9161         (fix<uns>_<mode>_mem): Likewise.
9162         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
9163         the register allocator from doing a direct move to the GPRs to do
9164         a store, and instead use the ISA 3.0 store byte/half-word from
9165         vector register instruction.  For IEEE 128-bit floating point,
9166         also optimize stores of 32-bit ints.
9167         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
9168
9169 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
9170
9171         * genextract.c (push_pathstr_operand): New function to support
9172         [a-zA-Z].
9173         (walk_rtx): Call push_pathstr_operand.
9174         (print_path): Support [a-zA-Z].
9175
9176 2018-02-07  Richard Biener  <rguenther@suse.de>
9177
9178         PR tree-optimization/84037
9179         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
9180         (cse_and_gimplify_to_preheader): Declare.
9181         (vect_get_place_in_interleaving_chain): Likewise.
9182         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9183         ivexpr_map.
9184         (_loop_vec_info::~_loop_vec_info): Delete it.
9185         (cse_and_gimplify_to_preheader): New function.
9186         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
9187         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
9188         (vectorizable_load): Likewise.  For grouped stores always base
9189         the IV on the first element.
9190         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
9191         condition before gimplifying.
9192
9193 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
9194
9195         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
9196         *DIV_EXPR and *MOD_EXPR.
9197
9198 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
9199
9200         PR target/84248
9201         * config/i386/i386.c (ix86_option_override_internal): Mask out
9202         the CF_SET bit when checking -fcf-protection.
9203
9204 2018-02-07  Tom de Vries  <tom@codesourcery.com>
9205
9206         PR libgomp/84217
9207         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
9208         enough.
9209
9210 2018-02-07  Richard Biener  <rguenther@suse.de>
9211
9212         PR tree-optimization/84204
9213         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
9214         this place.
9215
9216         PR tree-optimization/84205
9217         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
9218         special-case isl_ast_op_zdiv_r.
9219
9220         PR tree-optimization/84223
9221         * graphite-scop-detection.c (gather_bbs::before_dom_children):
9222         Only add conditions from within the region.
9223         (gather_bbs::after_dom_children): Adjust.
9224
9225 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
9226
9227         PR target/84209
9228         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9229         * config/avr/avr.md: Only post-reload split REG-REG moves if
9230         either register is GENERAL_REG_P.
9231
9232 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
9233
9234         PR tree-optimization/84235
9235         * tree-ssa-scopedtables.c
9236         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9237         if the subtraction is performed in floating point type where NaNs are
9238         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9239         build 1.  Formatting fix.
9240
9241 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
9242
9243         PR target/84146
9244         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9245         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9246         and skip it regardless of bb boundaries.  Use CALL_P macro,
9247         don't test INSN_P (insn) together with CALL_P or JUMP_P check
9248         unnecessarily, formatting fix.
9249
9250 2018-02-06  Michael Collison  <michael.collison@arm.com>
9251
9252         * config/arm/thumb2.md:
9253         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9254         (*thumb_mov_notscc): Ditto.
9255
9256 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
9257
9258         PR target/84154
9259         * config/rs6000/rs6000.md (su code attribute): Use "u" for
9260         unsigned_fix, not "s".
9261
9262 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9263
9264         * configure.ac (gcc_fn_eh_frame_ro): New function.
9265         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9266         correct .eh_frame permissions.
9267         * configure: Regenerate.
9268
9269 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
9270
9271         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9272         irrelevant options.
9273
9274 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9275
9276         * config/rs6000/rs6000.c (rs6000_option_override_internal):
9277         Display warning message for -mno-speculate-indirect-jumps.
9278
9279 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
9280
9281         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9282         Undocumented.
9283         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9284
9285 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
9286
9287         PR tree-optimization/84225
9288         * tree-eh.c (find_trapping_overflow): Only call
9289         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9290
9291 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
9292
9293         PR target/84145
9294         * config/i386/i386.c: Reimplement the check of possible options
9295         -mibt/-mshstk conbination. Change error messages.
9296         * doc/invoke.texi: Fix a typo: remove extra '='.
9297
9298 2018-02-06  Marek Polacek  <polacek@redhat.com>
9299
9300         PR tree-optimization/84228
9301         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9302
9303 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
9304
9305         PR target/82641
9306         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9307         emitted arch directives.
9308         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9309         __ARM_FEATURE_COPROC before changing architectures.
9310
9311 2018-02-06  Richard Biener  <rguenther@suse.de>
9312
9313         * config/i386/i386.c (print_reg): Fix typo.
9314         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9315
9316 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
9317
9318         * configure: Regenerate.
9319
9320 2018-02-05  Martin Sebor  <msebor@redhat.com>
9321
9322         PR tree-optimization/83369
9323         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9324         inlining context.
9325
9326 2018-02-05  Martin Liska  <mliska@suse.cz>
9327
9328         * doc/invoke.texi: Cherry-pick upstream r323995.
9329
9330 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
9331
9332         * ira.c (ira_init_register_move_cost): Adjust comment.
9333
9334 2018-02-05  Martin Liska  <mliska@suse.cz>
9335
9336         PR gcov-profile/84137
9337         * doc/gcov.texi: Fix typo in documentation.
9338
9339 2018-02-05  Martin Liska  <mliska@suse.cz>
9340
9341         PR gcov-profile/83879
9342         * doc/gcov.texi: Document necessity of --dynamic-list-data when
9343         using dlopen functionality.
9344
9345 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
9346
9347         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9348         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9349         _mm_maskz_range_ss, _mm_mask_range_round_ss,
9350         _mm_maskz_range_round_ss): New intrinsics.
9351         (__builtin_ia32_rangesd128_round)
9352         (__builtin_ia32_rangess128_round): Remove.
9353         (__builtin_ia32_rangesd128_mask_round,
9354         __builtin_ia32_rangess128_mask_round): New builtins.
9355         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9356         __builtin_ia32_rangess128_round): Remove.
9357         (__builtin_ia32_rangesd128_mask_round,
9358         __builtin_ia32_rangess128_mask_round): New builtins.
9359         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9360         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9361         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9362         "<round_saeonly_constraint>")): Changed to ...
9363         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9364         "<round_saeonly_scalar_constraint>")): ... this.
9365         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9366         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9367         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9368         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9369         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9370
9371 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
9372
9373         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9374         options.
9375         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9376         Remove all values except native, 8540 and 8548.
9377
9378 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
9379
9380         * config/i386/i386.c (ix86_output_function_return): Pass
9381         INVALID_REGNUM, instead of -1, as invalid register number to
9382         indirect_thunk_name and output_indirect_thunk.
9383
9384 2018-02-02  Julia Koval  <julia.koval@intel.com>
9385
9386         * config.gcc: Add -march=icelake.
9387         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9388         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9389         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9390         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9391         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9392         (processor_target_table): Add icelake.
9393         (ix86_option_override_internal): Handle new PTAs.
9394         (get_builtin_code_for_version): Handle icelake.
9395         (M_INTEL_COREI7_ICELAKE): New.
9396         (fold_builtin_cpu): Handle icelake.
9397         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9398         * doc/invoke.texi: Add -march=icelake.
9399
9400 2018-02-02  Julia Koval  <julia.koval@intel.com>
9401
9402         * config/i386/i386.c (ix86_option_override_internal): Change flags type
9403         to wide_int_bitmask.
9404         * wide-int-bitmask.h: New.
9405
9406 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
9407
9408         PR target/84066
9409         * config/i386/i386.md: Replace Pmode with word_mode in
9410         builtin_setjmp_setup and builtin_longjmp to support x32.
9411
9412 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
9413
9414         PR target/56010
9415         PR target/83743
9416         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9417         #include "opts.h".
9418         (rs6000_supported_cpu_names): New static variable.
9419         (linux_cpu_translation_table): Likewise.
9420         (elf_platform) <cpu>: Define new static variable and use it.
9421         Translate kernel AT_PLATFORM name to canonical name if needed.
9422         Error if platform name is unknown.
9423
9424 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
9425
9426         PR target/84089
9427         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9428
9429 2018-02-01  Jeff Law  <law@redhat.com>
9430
9431         PR target/84128
9432         * config/i386/i386.c (release_scratch_register_on_entry): Add new
9433         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
9434         the scratch if RELEASE_VIA_POP is false.
9435         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9436         If we have to save a temporary register, decrement SIZE appropriately.
9437         Pass new arguments to release_scratch_register_on_entry.
9438         (ix86_adjust_stack_and_probe): Likewise.
9439         (ix86_emit_probe_stack_range): Pass new arguments to
9440         release_scratch_register_on_entry.
9441
9442 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
9443
9444         PR rtl-optimization/84157
9445         * combine.c (change_zero_ext): Use REG_P predicate in
9446         front of HARD_REGISTER_P predicate.
9447
9448 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
9449
9450         * config/avr/avr.c (avr_option_override): Move disabling of
9451         -fdelete-null-pointer-checks to...
9452         * common/config/avr/avr-common.c (avr_option_optimization_table):
9453         ...here.
9454
9455 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
9456
9457         PR tree-optimization/81635
9458         * tree-data-ref.c (split_constant_offset_1): For types that
9459         wrap on overflow, try to use range info to prove that wrapping
9460         cannot occur.
9461
9462 2018-02-01  Renlin Li  <renlin.li@arm.com>
9463
9464         PR target/83370
9465         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9466         TAILCALL_ADDR_REGS.
9467         (aarch64_register_move_cost): Likewise.
9468         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9469         TAILCALL_ADDR_REGS.
9470         (REG_CLASS_NAMES): Likewise.
9471         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9472         TAILCALL_ADDR_REGS. Remove IP registers.
9473         * config/aarch64/aarch64.md (Ucs): Update register constraint.
9474
9475 2018-02-01  Richard Biener  <rguenther@suse.de>
9476
9477         * domwalk.h (dom_walker::dom_walker): Add additional constructor
9478         for specifying RPO order and allow NULL for that.
9479         * domwalk.c (dom_walker::dom_walker): Likewise.
9480         (dom_walker::walk): Handle NULL RPO order.
9481         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9482         in RPO order.
9483         (rewrite_update_dom_walker): Likewise.
9484         (mark_def_dom_walker): Likewise.
9485
9486 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
9487
9488         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9489         (aarch64_maybe_expand_sve_subreg_move): Declare.
9490         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9491         * config/aarch64/predicates.md (aarch64_any_register_operand): New
9492         predicate.
9493         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9494         that are semantically a reverse operation.
9495         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9496         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9497         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9498         functions.
9499         (aarch64_can_change_mode_class): For big-endian, forbid changes
9500         between two SVE modes if they have different element sizes.
9501
9502 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
9503
9504         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9505         the TImode handling for big-endian targets.
9506
9507 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
9508
9509         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9510         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
9511         not just bytes.
9512         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9513         Remove BSWAP handing for big-endian targets and use the form of
9514         LD1RQ appropariate for the mode.
9515
9516 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
9517
9518         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9519         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9520         duplicated element.
9521
9522 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
9523
9524         PR tearget/83845
9525         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9526         check for operands that need to go through aarch64_sve_reload_be.
9527
9528 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
9529
9530         PR tree-optimization/81661
9531         PR tree-optimization/84117
9532         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9533         * tree-eh.c: Include gimplify.h.
9534         (find_trapping_overflow, replace_trapping_overflow,
9535         rewrite_to_non_trapping_overflow): New functions.
9536         * tree-vect-loop.c: Include tree-eh.h.
9537         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9538         * tree-data-ref.c: Include tree-eh.h.
9539         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9540
9541 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
9542
9543         PR rtl-optimization/84123
9544         * combine.c (change_zero_ext): Check if hard register satisfies
9545         can_change_dest_mode before calling gen_lowpart_SUBREG.
9546
9547 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
9548
9549         PR target/82444
9550         * ira.c (ira_init_register_move_cost): Remove assert.
9551
9552 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
9553
9554         PR rtl-optimization/84071
9555         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9556         * doc/tm.texi: Regenerate.
9557
9558 2018-01-31  Richard Biener  <rguenther@suse.de>
9559
9560         PR tree-optimization/84132
9561         * tree-data-ref.c (analyze_miv_subscript): Properly
9562         check whether evolution_function_is_affine_multivariate_p
9563         before calling gcd_of_steps_may_divide_p.
9564
9565 2018-01-31  Julia Koval  <julia.koval@intel.com>
9566
9567         PR target/83618
9568         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9569         * config/i386/i386.md (rdpid_rex64) New.
9570         (rdpid): Make 32bit only.
9571
9572 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
9573
9574         PR lto/84105
9575         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9576         an IDENTIFIER_NODE for FUNCTION_TYPE's.
9577
9578 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
9579
9580         Revert
9581         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
9582
9583         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9584
9585 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
9586
9587         PR rtl-optimization/84071
9588         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9589         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9590
9591 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
9592
9593         * config/arc/arc.c (arc_handle_aux_attribute): New function.
9594         (arc_attribute_table): Add 'aux' attribute.
9595         (arc_in_small_data_p): Consider aux like variables.
9596         (arc_is_aux_reg_p): New function.
9597         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9598         (arc_get_aux_arg): New function.
9599         (prepare_move_operands): Handle aux-register access.
9600         (arc_handle_aux_attribute): New function.
9601         * doc/extend.texi (ARC Variable attributes): Add subsection.
9602
9603 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
9604
9605         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9606         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9607         (arc_attribute_table): Add 'uncached' attribute.
9608         (arc_print_operand): Print '.di' flag for uncached memory
9609         accesses.
9610         (arc_in_small_data_p): Do not consider for small data the uncached
9611         types.
9612         (arc_is_uncached_mem_p): New function.
9613         * config/arc/predicates.md (compact_store_memory_operand): Check
9614         for uncached memory accesses.
9615         (nonvol_nonimm_operand): Likewise.
9616         * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9617
9618 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
9619
9620         PR c/84100
9621         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9622         falign-loops=): Add Optimization flag.
9623
9624 2018-01-30  Jeff Law  <law@redhat.com>
9625
9626         PR target/84064
9627         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9628         INT_REGISTERS_SAVED.  Check it prior to calling
9629         get_scratch_register_on_entry.
9630         (ix86_adjust_stack_and_probe): Similarly.
9631         (ix86_emit_probe_stack_range): Similarly.
9632         (ix86_expand_prologue): Corresponding changes.
9633
9634 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9635
9636         PR target/40411
9637         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9638         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9639
9640 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
9641
9642         PR target/84112
9643         * lra-constraints.c (curr_insn_transform): Process AND in the
9644         address.
9645
9646 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
9647
9648         PR rtl-optimization/83986
9649         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9650         dependence against last_pending_memory_flush in addition to
9651         pending_jump_insns.
9652
9653 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
9654
9655         PR tree-optimization/81611
9656         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9657         copies.
9658
9659 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
9660
9661         PR target/83758
9662         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9663         a reg rtx.
9664
9665 2018-01-30  Richard Biener  <rguenther@suse.de>
9666             Jakub Jelinek  <jakub@redhat.com>
9667
9668         PR tree-optimization/84111
9669         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9670         inner loops added during recursion, as they don't have up-to-date
9671         SSA form.
9672
9673 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
9674
9675         PR ipa/81360
9676         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9677         (can_inline_edge_by_limits_p): ... here.
9678         (can_early_inline_edge_p, check_callers,
9679         update_caller_keys, update_callee_keys, recursive_inlining,
9680         add_new_edges_to_heap, speculation_useful_p,
9681         inline_small_functions,
9682         inline_small_functions, flatten_function,
9683         inline_to_all_callers_1): Update.
9684
9685 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
9686
9687         * profile-count.c (profile_count::combine_with_ipa_count): Handle
9688         zeros correctly.
9689
9690 2018-01-30  Richard Biener  <rguenther@suse.de>
9691
9692         PR tree-optimization/83008
9693         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9694         invariant and constant vector uses in stmts when they need
9695         more than one stmt.
9696
9697 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9698
9699         PR bootstrap/84017
9700         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9701         * configure: Regenerate.
9702
9703 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
9704
9705         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9706         pattern.
9707         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9708         Use gen_rtx_REG rather than gen_lowpart.
9709
9710 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
9711
9712         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9713         rather than 0 when creating partial subregs.
9714
9715 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
9716
9717         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9718         of usage.
9719
9720 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
9721
9722         PR target/81550
9723         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9724         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9725         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9726         flags.  This restores the settings used before the 2017-07-24.
9727         Turning off pre increment/decrement/modify allows IVOPTS to
9728         optimize DF/SF loops where the index is an int.
9729
9730 2018-01-29  Richard Biener  <rguenther@suse.de>
9731             Kelvin Nilsen  <kelvin@gcc.gnu.org>
9732
9733         PR bootstrap/80867
9734         * tree-vect-stmts.c (vectorizable_call): Don't call
9735         targetm.vectorize_builtin_md_vectorized_function if callee is
9736         NULL.
9737
9738 2018-01-22  Carl Love  <cel@us.ibm.com>
9739
9740         * doc/extend.tex: Fix typo in second arg in
9741         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9742
9743 2018-01-29  Richard Biener  <rguenther@suse.de>
9744
9745         PR tree-optimization/84086
9746         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9747         (flush_ssaname_freelist): When SSA names were released reset
9748         the SCEV hash table.
9749
9750 2018-01-29  Richard Biener  <rguenther@suse.de>
9751
9752         PR tree-optimization/84057
9753         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9754         removed paths when removing edges.
9755
9756 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
9757
9758         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9759         -mfunction-return=@var{choice}.
9760
9761 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9762
9763         PR diagnostic/84034
9764         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9765         Handle CR like TAB.
9766         (layout::print_source_line): Likewise.
9767         (test_get_line_width_without_trailing_whitespace): Add test cases.
9768
9769 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
9770
9771         PR middle-end/84040
9772         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9773         debug insns.
9774
9775 2018-01-26  Jim Wilson  <jimw@sifive.com>
9776
9777         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9778
9779         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9780         specified.
9781
9782 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9783
9784         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9785         and CMP + SUB-immediate -> SUBS.
9786
9787 2018-01-26  Martin Sebor  <msebor@redhat.com>
9788
9789         PR tree-optimization/83896
9790         * tree-ssa-strlen.c (get_string_len): Rename...
9791         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
9792         Avoid assuming length is constant.
9793         (handle_char_store): Use HOST_WIDE_INT for string length.
9794
9795 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
9796
9797         PR target/81763
9798         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9799         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9800
9801 2018-01-26  Richard Biener  <rguenther@suse.de>
9802
9803         PR rtl-optimization/84003
9804         * dse.c (record_store): Only record redundant stores when
9805         the earlier store aliases at least all accesses the later one does.
9806
9807 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
9808
9809         PR rtl-optimization/83985
9810         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9811         REG_CFA_RESTORE insns.
9812         (delete_unmarked_insns): Don't ignore separate shrink wrapping
9813         REG_CFA_RESTORE insns here.
9814
9815         PR c/83989
9816         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9817         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9818
9819 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
9820
9821         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9822         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9823         (arc_init): Likewise.
9824         (arc_override_options): Likewise.
9825         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9826         value.
9827         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9828         support.
9829         * config/arc/arc.h (TARGET_DBNZ): Define.
9830         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9831         properly set the tune attribute.
9832         (dbnz): Use TARGET_DBNZ guard.
9833         * config/arc/arc.opt (mtune): Add core3 option.
9834
9835 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
9836
9837         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9838         recognize new pic like addresses.
9839         (arc_delegitimize_address): Clean up.
9840
9841 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
9842
9843         * config/arc/arc-arches.def: Option mrf16 valid for all
9844         architectures.
9845         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9846         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9847         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9848         * config/arc/arc-tables.opt: Regenerate.
9849         * config/arc/arc.c (arc_conditional_register_usage): Handle
9850         reduced register file case.
9851         (arc_file_start): Set must have build attributes.
9852         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9853         mrf16 option value.
9854         * config/arc/arc.opt (mrf16): Add new option.
9855         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9856         * config/arc/genmultilib.awk: Handle new mrf16 option.
9857         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9858         * config/arc/t-multilib: Regenerate.
9859         * doc/invoke.texi (ARC Options): Document mrf16 option.
9860
9861 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
9862
9863         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9864         * config/arc/arc.c (arc_handle_secure_attribute): New function.
9865         (arc_attribute_table): Add 'secure_call' attribute.
9866         (arc_print_operand): Print secure call operand.
9867         (arc_function_ok_for_sibcall): Don't optimize tail calls when
9868         secure.
9869         (arc_is_secure_call_p): New function.  * config/arc/arc.md
9870         (call_i): Add support for sjli instruction.
9871         (call_value_i): Likewise.
9872         * config/arc/constraints.md (Csc): New constraint.
9873
9874 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
9875             John Eric Martin  <John.Martin@emmicro-us.com>
9876
9877         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9878         * config/arc/arc.c (_arc_jli_section): New struct.
9879         (arc_jli_section): New type.
9880         (rc_jli_sections): New static variable.
9881         (arc_handle_jli_attribute): New function.
9882         (arc_attribute_table): Add jli_always and jli_fixed attribute.
9883         (arc_file_end): New function.
9884         (TARGET_ASM_FILE_END): Define.
9885         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9886         (arc_add_jli_section): New function.
9887         (jli_call_scan): Likewise.
9888         (arc_reorg): Call jli_call_scan.
9889         (arc_output_addsi): Remove 'S' from printing asm operand.
9890         (arc_is_jli_call_p): New function.
9891         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9892         operand.
9893         (movhi_insn): Likewise.
9894         (movsi_insn): Likewise.
9895         (movsi_set_cc_insn): Likewise.
9896         (loadqi_update): Likewise.
9897         (load_zeroextendqisi_update): Likewise.
9898         (load_signextendqisi_update): Likewise.
9899         (loadhi_update): Likewise.
9900         (load_zeroextendhisi_update): Likewise.
9901         (load_signextendhisi_update): Likewise.
9902         (loadsi_update): Likewise.
9903         (loadsf_update): Likewise.
9904         (movsicc_insn): Likewise.
9905         (bset_insn): Likewise.
9906         (bxor_insn): Likewise.
9907         (bclr_insn): Likewise.
9908         (bmsk_insn): Likewise.
9909         (bicsi3_insn): Likewise.
9910         (cmpsi_cc_c_insn): Likewise.
9911         (movsi_ne): Likewise.
9912         (movsi_cond_exec): Likewise.
9913         (clrsbsi2): Likewise.
9914         (norm_f): Likewise.
9915         (normw): Likewise.
9916         (swap): Likewise.
9917         (divaw): Likewise.
9918         (flag): Likewise.
9919         (sr): Likewise.
9920         (kflag): Likewise.
9921         (ffs): Likewise.
9922         (ffs_f): Likewise.
9923         (fls): Likewise.
9924         (call_i): Remove 'S' asm letter, add jli instruction.
9925         (call_value_i): Likewise.
9926         * config/arc/arc.op (mjli-always): New option.
9927         * config/arc/constraints.md (Cji): New constraint.
9928         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9929         operand.
9930         (subsf3_fpx): Likewise.
9931         (mulsf3_fpx): Likewise.
9932         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9933         asm operand.
9934         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9935         function attrbutes.
9936         * doc/invoke.texi (ARC): Document mjli-always option.
9937
9938 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
9939
9940         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9941         avoid addition with 0 and use incw and decw where possible.
9942
9943 2018-01-26  Richard Biener  <rguenther@suse.de>
9944
9945         PR tree-optimization/81082
9946         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9947         association if it requires casting to unsigned.
9948         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9949         from fold_plusminus_mult_expr to catch important cases late when
9950         range info is available.
9951
9952 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9953
9954         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9955         * configure.ac (hidden_linkonce): New test.
9956         * configure: Regenerate.
9957         * config.in: Regenerate.
9958
9959 2018-01-26  Julia Koval  <julia.koval@intel.com>
9960
9961         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9962         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9963         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9964         _mm_mask_bitshuffle_epi64_mask): Fix type.
9965         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9966         USI_FTYPE_V4DI_V4DI_USI): Remove.
9967         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9968         __builtin_ia32_vpshufbitqmb256_mask,
9969         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9970         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9971         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9972
9973 2018-01-26  Alan Modra  <amodra@gmail.com>
9974
9975         PR target/84033
9976         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9977         UNSPEC_VBPERMQ.  Sort other unspecs.
9978
9979 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
9980
9981         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9982
9983 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
9984
9985         PR middle-end/83055
9986         * predict.c (drop_profile): Do not push/pop cfun; update also
9987         node->count.
9988         (handle_missing_profiles): Fix logic looking for zero profiles.
9989
9990 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
9991
9992         PR middle-end/83977
9993         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9994         on functions with #pragma omp declare simd or functions with simd
9995         attribute.
9996         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9997         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9998         Remove trailing \n from warning_at calls.
9999
10000 2018-01-25  Tom de Vries  <tom@codesourcery.com>
10001
10002         PR target/84028
10003         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10004         for neutered workers.
10005
10006 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
10007
10008         PR target/68467
10009         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
10010         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
10011
10012 2018-01-24  Jeff Law  <law@redhat.com>
10013
10014         PR target/83994
10015         * i386.c (get_probe_interval): Move to earlier point.
10016         (ix86_compute_frame_layout): If -fstack-clash-protection and
10017         the frame is larger than the probe interval, then use pushes
10018         to save registers rather than reg->mem moves.
10019         (ix86_expand_prologue): Remove conditional for int_registers_saved
10020         assertion.
10021
10022 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
10023
10024         PR target/84014
10025         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
10026         min/max for never referenced object.
10027
10028 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
10029
10030         PR middle-end/83977
10031         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
10032         here.
10033         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
10034         attributes from DECL_ATTRIBUTES (decl) without affecting
10035         DECL_ATTRIBUTES (current_function_decl).
10036         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
10037         functions with non-NULL DECL_ABSTRACT_ORIGIN.
10038
10039 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
10040
10041         PR tree-optimization/83979
10042         * fold-const.c (fold_comparison): Use constant_boolean_node
10043         instead of boolean_{true,false}_node.
10044
10045 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
10046
10047         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
10048         with zero counts.
10049
10050 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10051
10052         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10053         Simplify the clause that sets the length attribute.
10054         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10055         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
10056         clause that sets the length attribute.
10057         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10058
10059 2018-01-24  Tom de Vries  <tom@codesourcery.com>
10060
10061         PR target/83589
10062         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
10063         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
10064         Add strict parameter.
10065         (prevent_branch_around_nothing): Insert dummy insn between branch to
10066         label and label with no ptx insn inbetween.
10067         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
10068
10069 2018-01-24  Tom de Vries  <tom@codesourcery.com>
10070
10071         PR target/81352
10072         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10073         for neutered threads in warp.
10074         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
10075
10076 2018-01-24  Richard Biener  <rguenther@suse.de>
10077
10078         PR tree-optimization/83176
10079         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
10080         operands.
10081
10082 2018-01-24  Richard Biener  <rguenther@suse.de>
10083
10084         PR tree-optimization/82819
10085         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
10086         code generating pluses that are no-ops in the target precision.
10087
10088 2018-01-24  Richard Biener  <rguenther@suse.de>
10089
10090         PR middle-end/84000
10091         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
10092
10093 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
10094
10095         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
10096         to merge probabilities.
10097         * predict.c (probably_never_executed): Also mark as cold functions
10098         with global 0 profile and guessed local profile.
10099         * profile-count.c (profile_probability::combine_with_count): New
10100         member function.
10101         * profile-count.h (profile_probability::operator*,
10102         profile_probability::operator*=, profile_probability::operator/,
10103         profile_probability::operator/=): Reduce precision to adjusted
10104         and set value to guessed on contradictory divisions.
10105         (profile_probability::combine_with_freq): Remove.
10106         (profile_probability::combine_wiht_count): Declare.
10107         (profile_count::force_nonzero):: Set to adjusted.
10108         (profile_count::probability_in):: Set quality to adjusted.
10109         * tree-ssa-tail-merge.c (replace_block_by): Use
10110         combine_with_count.
10111
10112 2018-01-23  Andrew Waterman  <andrew@sifive.com>
10113             Jim Wilson  <jimw@sifive.com>
10114
10115         * config/riscv/riscv.c (riscv_stack_boundary): New.
10116         (riscv_option_override): Set riscv_stack_boundary.  Handle
10117         riscv_preferred_stack_boundary_arg.
10118         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
10119         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
10120         (STACK_BOUNDARY): Set to riscv_stack_boundary.
10121         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
10122         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
10123         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
10124
10125 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
10126
10127         PR target/83905
10128         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
10129         of struct ix86_frame.
10130         (ix86_expand_epilogue): Likewise.  Add a local variable for
10131         the reg_save_offset field in struct ix86_frame.
10132
10133 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
10134
10135         PR tree-optimization/82604
10136         * tree-loop-distribution.c (enum partition_kind): New enum item
10137         PKIND_PARTIAL_MEMSET.
10138         (partition_builtin_p): Support above new enum item.
10139         (generate_code_for_partition): Ditto.
10140         (compute_access_range): Differentiate cases that equality can be
10141         proven at all loops, the innermost loops or no loops.
10142         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
10143         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
10144         (finalize_partitions, distribute_loop): Don't fuse partition of
10145         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
10146         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
10147         parloop is enabled.
10148
10149 2018-01-23  Martin Liska  <mliska@suse.cz>
10150
10151         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
10152         order to ignore the predictor.
10153         (PRED_POLYMORPHIC_CALL): Likewise.
10154         (PRED_RECURSIVE_CALL): Likewise.
10155
10156 2018-01-23  Martin Liska  <mliska@suse.cz>
10157
10158         * tree-profile.c (tree_profiling): Print function header to
10159         aware reader which function we are working on.
10160         * value-prof.c (gimple_find_values_to_profile): Do not print
10161         not interesting value histograms.
10162
10163 2018-01-23  Martin Liska  <mliska@suse.cz>
10164
10165         * profile-count.h (enum profile_quality): Add
10166         profile_uninitialized as the first value. Do not number values
10167         as they are zero based.
10168         (profile_count::verify): Update sanity check.
10169         (profile_probability::verify): Likewise.
10170
10171 2018-01-23  Nathan Sidwell  <nathan@acm.org>
10172
10173         * doc/invoke.texi (ffor-scope): Deprecate.
10174
10175 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
10176
10177         PR tree-optimization/83510
10178         * domwalk.c (set_all_edges_as_executable): New function.
10179         (dom_walker::dom_walker): Convert bool param
10180         "skip_unreachable_blocks" to enum reachability.  Move setup of
10181         edge flags to set_all_edges_as_executable and only do it when
10182         reachability is REACHABLE_BLOCKS.
10183         * domwalk.h (enum dom_walker::reachability): New enum.
10184         (dom_walker::dom_walker): Convert bool param
10185         "skip_unreachable_blocks" to enum reachability.
10186         (set_all_edges_as_executable): New decl.
10187         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
10188         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
10189         "reachability".
10190         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
10191         but converting true to REACHABLE_BLOCKS.
10192         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
10193         * tree-vrp.c
10194         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
10195         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
10196         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
10197         REACHABLE_BLOCKS.
10198         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
10199         if check_all_array_refs will be called.
10200
10201 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
10202
10203         * tree.c (selftest::test_location_wrappers): Add more test
10204         coverage.
10205
10206 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
10207
10208         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
10209         (selftest::test_bit_in_range): Likewise.
10210
10211 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
10212
10213         PR testsuite/83888
10214         * doc/sourcebuild.texi (vect_float): Say that the selector
10215         only describes the situation when -funsafe-math-optimizations is on.
10216         (vect_float_strict): Document.
10217
10218 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
10219
10220         PR tree-optimization/83965
10221         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10222         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10223         instead of checking only for a reduction.
10224         (vect_recog_widen_sum_pattern): Likewise.
10225
10226 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
10227
10228         * predict.c (probably_never_executed): Only use precise profile info.
10229         (compute_function_frequency): Skip after inlining hack since we now
10230         have quality checking.
10231
10232 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
10233
10234         * profile-count.h (profile_probability::very_unlikely,
10235         profile_probability::unlikely, profile_probability::even): Set
10236         precision to guessed.
10237
10238 2018-01-23  Richard Biener  <rguenther@suse.de>
10239
10240         PR tree-optimization/83963
10241         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10242         Properly terminate dominator walk when crossing the exit edge not
10243         when visiting its source block.
10244
10245 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
10246
10247         PR c++/83918
10248         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10249         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10250
10251 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
10252
10253         PR tree-optimization/83957
10254         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
10255         semicolon after for body surrounded by braces.
10256
10257         PR tree-optimization/83081
10258         * profile-count.h (profile_probability::split): New method.
10259         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10260         Use profile_probability::split.
10261         (do_compare_rtx_and_jump): Fix adjustment of probabilities
10262         when splitting a single conditional jump into 2.
10263
10264 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
10265
10266         PR tree-optimization/69452
10267         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10268         decl.
10269
10270 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10271
10272         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10273         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10274         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10275
10276 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10277
10278         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
10279         declaration.
10280         * config/rl78/rl78.md (movdi): New define_expand.
10281         * config/rl78/rl78.c (rl78_split_movdi): New function.
10282
10283 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
10284
10285         PR target/83862
10286         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10287         no longer used.
10288         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10289         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10290         128-bit to produce an UNSPEC move to get the double word with the
10291         signbit and then a shift directly to do signbit.
10292         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10293         implementation with a new version that just does either a direct
10294         move or a regular move.  Move memory interface to separate insns.
10295         Move insns so they are next to the expander.
10296         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10297         with signbit move.  Split big and little endian case.
10298         (signbit<mode>2_dm_mem_le): Likewise.
10299         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10300         (signbit<mode>2_dm2): Likewise.
10301
10302 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10303
10304         * config/rl78/rl78.md (anddi3): New define_expand.
10305
10306 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10307
10308         * config/rl78/rl78.md (umindi3): New define_expand.
10309
10310 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10311
10312         * config/rl78/rl78.md (smindi3): New define_expand.
10313
10314 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10315
10316         * config/rl78/rl78.md (smaxdi3): New define_expand.
10317
10318 2018-01-22  Carl Love  <cel@us.ibm.com>
10319
10320         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10321         LVX_V1TI): Add macro expansion.
10322         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10323         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10324         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10325         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10326         Change check to determine if the instruction is a byte reversing
10327         entry.  Fix typo in comment.
10328         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10329         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10330         Add def_builtin calls for new builtins.
10331         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10332         Add define_insn expansion.
10333
10334 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10335
10336         * config/rl78/rl78.md (umaxdi3): New define_expand.
10337
10338 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
10339
10340         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10341         for non-QImode registers.
10342
10343 2018-01-22  Richard Biener  <rguenther@suse.de>
10344
10345         PR tree-optimization/83963
10346         * graphite-scop-detection.c (scop_detection::get_sese): Delay
10347         including the loop exit block.
10348         (scop_detection::merge_sese): Likewise.
10349         (scop_detection::add_scop): Do it here instead.
10350
10351 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10352
10353         * doc/sourcebuild.texi (arm_softfloat): Document.
10354
10355 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
10356
10357         PR gcc/77734
10358         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10359         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10360         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10361
10362 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10363             David Edelsohn  <dje.gcc@gmail.com>
10364
10365         PR target/83946
10366         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10367         Change "crset eq" to "crset 2".
10368         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10369         (*call_indirect_aix<mode>_nospec): Likewise.
10370         (*call_value_indirect_aix<mode>_nospec): Likewise.
10371         (*call_indirect_elfv2<mode>_nospec): Likewise.
10372         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10373         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10374         change assembly output from . to $.
10375         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10376         (indirect_jump<mode>_nospec): Change assembly output from . to $.
10377         (*tablejump<mode>_internal1_nospec): Likewise.
10378
10379 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
10380
10381         PR target/80870
10382         * config/sh/sh_optimize_sett_clrt.cc:
10383         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10384
10385 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
10386
10387         PR tree-optimization/83940
10388         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10389         offset_dt to vect_constant_def rather than vect_unknown_def_type.
10390         (vect_check_load_store_mask): Add a mask_dt_out parameter and
10391         use it to pass back the definition type.
10392         (vect_check_store_rhs): Likewise rhs_dt_out.
10393         (vect_build_gather_load_calls): Add a mask_dt argument and use
10394         it instead of a call to vect_is_simple_use.
10395         (vectorizable_store): Update calls to vect_check_load_store_mask
10396         and vect_check_store_rhs.  Use the dt returned by the latter instead
10397         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
10398         instead of calls to vect_is_simple_use.  Pass the scalar rather
10399         than the vector operand to vect_is_simple_use when handling
10400         second and subsequent copies of an rhs value.
10401         (vectorizable_load): Update calls to vect_check_load_store_mask
10402         and vect_build_gather_load_calls.  Use the cached mask_dt and
10403         gs_info.offset_dt instead of calls to vect_is_simple_use.
10404
10405 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
10406
10407         PR middle-end/83945
10408         * tree-emutls.c: Include gimplify.h.
10409         (lower_emutls_2): New function.
10410         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10411         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10412         it before further processing.
10413
10414         PR target/83930
10415         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10416         UINTVAL (trueop1) instead of INTVAL (op1).
10417
10418 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
10419
10420         PR debug/81570
10421         PR debug/83728
10422         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10423         INCOMING_FRAME_SP_OFFSET if not defined.
10424         (scan_trace): Add ENTRY argument.  If true and
10425         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10426         emit a note to adjust the CFA offset.
10427         (create_cfi_notes): Adjust scan_trace callers.
10428         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10429         INCOMING_FRAME_SP_OFFSET in the CIE.
10430         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10431         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10432         Likewise.
10433         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10434         * doc/tm.texi: Regenerated.
10435
10436 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
10437
10438         PR rtl-optimization/83147
10439         * lra-constraints.c (remove_inheritance_pseudos): Use
10440         lra_substitute_pseudo_within_insn.
10441
10442 2018-01-19  Tom de Vries  <tom@codesourcery.com>
10443             Cesar Philippidis  <cesar@codesourcery.com>
10444
10445         PR target/83920
10446         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10447
10448 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
10449
10450         PR target/83790
10451         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10452         spaces for function labels.
10453
10454 2018-01-19  Martin Liska  <mliska@suse.cz>
10455
10456         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10457         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10458         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10459         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10460         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10461         (PRED_CONST_RETURN): Change from 69 to 65.
10462         (PRED_NULL_RETURN): Change from 91 to 71.
10463         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10464         (PRED_LOOP_GUARD): Change from 66 to 73.
10465
10466 2018-01-19  Martin Liska  <mliska@suse.cz>
10467
10468         * predict.c (predict_insn_def): Add new assert.
10469         (struct branch_predictor): Change type to signed integer.
10470         (test_prediction_value_range): Amend test to cover
10471         PROB_UNINITIALIZED.
10472         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10473         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10474         (PRED_LOOP_ITERATIONS_MAX): Likewise.
10475         (PRED_LOOP_IV_COMPARE): Likewise.
10476         * predict.h (PROB_UNINITIALIZED): Define new constant.
10477
10478 2018-01-19  Martin Liska  <mliska@suse.cz>
10479
10480         * predict.c (dump_prediction): Add new format for
10481         analyze_brprob.py script which is enabled with -details
10482         suboption.
10483         * profile-count.h (precise_p): New function.
10484
10485 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
10486
10487         PR tree-optimization/83922
10488         * tree-vect-loop.c (vect_verify_full_masking): Return false if
10489         there are no statements that need masking.
10490         (vect_active_double_reduction_p): New function.
10491         (vect_analyze_loop_operations): Use it when handling phis that
10492         are not in the loop header.
10493
10494 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
10495
10496         PR tree-optimization/83914
10497         * tree-vect-loop.c (vectorizable_induction): Don't convert
10498         init_expr or apply the peeling adjustment for inductions
10499         that are nested within the vectorized loop.
10500
10501 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10502
10503         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10504         instead of NEG.
10505
10506 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
10507
10508         PR sanitizer/81715
10509         PR testsuite/83882
10510         * function.h (gimplify_parameters): Add gimple_seq * argument.
10511         * function.c: Include gimple.h and options.h.
10512         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10513         for the added local temporaries if needed.
10514         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10515         if there are any parameter cleanups, wrap whole body into a
10516         try/finally with the cleanups.
10517
10518 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
10519
10520         PR target/82964
10521         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10522         Use GET_MODE_CLASS for scalar floating point.
10523
10524 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
10525
10526         PR ipa/82256
10527         patch by PaX Team
10528         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10529         Fix call of call_cgraph_insertion_hooks.
10530
10531 2018-01-18  Martin Sebor  <msebor@redhat.com>
10532
10533         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10534
10535 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
10536
10537         PR ipa/83619
10538         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10539         frequencies.
10540
10541 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
10542
10543         PR other/70268
10544         * common.opt: (-ffile-prefix-map): New option.
10545         * opts.c (common_handle_option): Defer it.
10546         * opts-global.c (handle_common_deferred_options): Handle it.
10547         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10548         * file-prefix-map.h: New file.
10549         (remap_debug_filename, add_debug_prefix_map): ...here.
10550         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10551         * final.c (debug_prefix_map, add_debug_prefix_map
10552         remap_debug_filename): Move to...
10553         * file-prefix-map.c: New file.
10554         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10555         generalize, get rid of alloca(), use strrchr() instead of strchr().
10556         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10557         Implement in terms of add_prefix_map().
10558         (remap_macro_filename, remap_debug_filename): Implement in term of
10559         remap_filename().
10560         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10561         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10562         * dbxout.c: Include file-prefix-map.h.
10563         * varasm.c: Likewise.
10564         * vmsdbgout.c: Likewise.
10565         * xcoffout.c: Likewise.
10566         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10567         * doc/cppopts.texi (-fmacro-prefix-map): Document.
10568         * doc/invoke.texi (-ffile-prefix-map): Document.
10569         (-fdebug-prefix-map): Update description.
10570
10571 2018-01-18  Martin Liska  <mliska@suse.cz>
10572
10573         * config/i386/i386.c (indirect_thunk_name): Document that also
10574         lfence is emitted.
10575         (output_indirect_thunk): Document why both instructions
10576         (pause and lfence) are generated.
10577
10578 2018-01-18  Richard Biener  <rguenther@suse.de>
10579
10580         PR tree-optimization/83887
10581         * graphite-scop-detection.c
10582         (scop_detection::get_nearest_dom_with_single_entry): Remove.
10583         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10584         (scop_detection::merge_sese): Re-implement with a flood-fill
10585         algorithm that properly finds a SESE region if it exists.
10586
10587 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
10588
10589         PR c/61240
10590         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10591         pointer_diff optimizations use view_convert instead of convert.
10592
10593 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10594
10595         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10596         Generate different code for -mno-speculate-indirect-jumps.
10597         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10598         (*call_indirect_aix<mode>): Disable for
10599         -mno-speculate-indirect-jumps.
10600         (*call_indirect_aix<mode>_nospec): New define_insn.
10601         (*call_value_indirect_aix<mode>): Disable for
10602         -mno-speculate-indirect-jumps.
10603         (*call_value_indirect_aix<mode>_nospec): New define_insn.
10604         (*sibcall_nonlocal_sysv<mode>): Generate different code for
10605         -mno-speculate-indirect-jumps.
10606         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10607
10608 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
10609
10610         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10611         long double type, set the flags for noting the default long double
10612         type, even if we don't pass or return a long double type.
10613
10614 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
10615
10616         PR ipa/83051
10617         * ipa-inline.c (flatten_function): Do not overwrite final inlining
10618         failure.
10619
10620 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
10621
10622         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10623         support for merge[hl].
10624         (fold_mergehl_helper): New helper function.
10625         (tree-vector-builder.h): New #include for tree_vector_builder usage.
10626         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10627         (altivec_vmrglw_direct): Add xxmrglw insn.
10628
10629 2018-01-17  Andrew Waterman  <andrew@sifive.com>
10630
10631         * config/riscv/riscv.c (riscv_conditional_register_usage): If
10632         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10633
10634 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
10635
10636         PR lto/83121
10637         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10638         call the lto_location_cache before reading the
10639         DECL_SOURCE_LOCATION of the types.
10640
10641 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
10642             Richard Sandiford  <richard.sandiford@linaro.org>
10643
10644         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10645         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10646         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10647         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10648         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10649         Add declaration.
10650         * config/aarch64/constraints.md (aarch64_movti_operand):
10651         Limit immediates.
10652         * config/aarch64/predicates.md (Uti): Add new constraint.
10653
10654 2018-01-17 Carl Love  <cel@us.ibm.com>
10655
10656         * config/rs6000/vsx.md (define_expand xl_len_r,
10657         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10658         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10659         lxvll.
10660         (define_expand, define_insn): Move the shift left from  the
10661         define_insn to the define_expand for lxvl and stxvl instructions.
10662         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10663         and XL_LEN_R definitions to PURE.
10664
10665 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
10666
10667         * config/i386/i386.c (indirect_thunk_name): Declare regno
10668         as unsigned int.  Compare regno with INVALID_REGNUM.
10669         (output_indirect_thunk): Ditto.
10670         (output_indirect_thunk_function): Ditto.
10671         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
10672         in the call to output_indirect_thunk_function.
10673
10674 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
10675
10676         PR middle-end/83884
10677         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10678         rather than the size of inner_type to determine the stack slot size
10679         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10680
10681 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
10682
10683         PR target/83546
10684         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10685         to PTA_SILVERMONT.
10686
10687 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
10688
10689         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10690         endian Linux systems to optionally enable multilibs for selecting
10691         the long double type if the user configured an explicit type.
10692         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10693         have no long double multilibs if not defined.
10694         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10695         warn if the user used -mabi={ieee,ibm}longdouble and we built
10696         multilibs for long double.
10697         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10698         appropriate multilib option.
10699         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10700         multilib options.
10701         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10702         for building long double multilibs.
10703         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10704
10705 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
10706
10707         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10708         copies.
10709
10710         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10711         64 bits.
10712         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10713         128 bits.
10714
10715         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10716         variables.
10717
10718         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10719         return value.
10720
10721 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
10722
10723         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10724         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10725
10726 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10727
10728         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10729         different rtl trees depending on TARGET_64BIT.
10730         (rs6000_gen_lvx): Likewise.
10731
10732 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
10733
10734         * config/visium/visium.md (nop): Tweak comment.
10735         (hazard_nop): Likewise.
10736
10737 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10738
10739         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10740         -mspeculate-indirect-jumps.
10741         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10742         for -mno-speculate-indirect-jumps.
10743         (*call_indirect_elfv2<mode>_nospec): New define_insn.
10744         (*call_value_indirect_elfv2<mode>): Disable for
10745         -mno-speculate-indirect-jumps.
10746         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10747         (indirect_jump): Emit different RTL for
10748         -mno-speculate-indirect-jumps.
10749         (*indirect_jump<mode>): Disable for
10750         -mno-speculate-indirect-jumps.
10751         (*indirect_jump<mode>_nospec): New define_insn.
10752         (tablejump): Emit different RTL for
10753         -mno-speculate-indirect-jumps.
10754         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10755         (tablejumpsi_nospec): New define_expand.
10756         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10757         (tablejumpdi_nospec): New define_expand.
10758         (*tablejump<mode>_internal1): Disable for
10759         -mno-speculate-indirect-jumps.
10760         (*tablejump<mode>_internal1_nospec): New define_insn.
10761         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10762         option.
10763
10764 2018-01-16  Artyom Skrobov tyomitch@gmail.com
10765
10766         * caller-save.c (insert_save): Drop unnecessary parameter.  All
10767         callers updated.
10768
10769 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
10770             Richard Biener  <rguenth@suse.de>
10771
10772         PR libgomp/83590
10773         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10774         return early, inline manually is_gimple_sizepos.  Make sure if we
10775         call gimplify_expr we don't end up with a gimple constant.
10776         * tree.c (variably_modified_type_p): Don't return true for
10777         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
10778         * gimplify.h (is_gimple_sizepos): Remove.
10779
10780 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
10781
10782         PR tree-optimization/83857
10783         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10784         vectorizable_live_operation for pure SLP statements.
10785         (vectorizable_live_operation): Handle PHIs.
10786
10787 2018-01-16  Richard Biener  <rguenther@suse.de>
10788
10789         PR tree-optimization/83867
10790         * tree-vect-stmts.c (vect_transform_stmt): Precompute
10791         nested_in_vect_loop_p since the scalar stmt may get invalidated.
10792
10793 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
10794
10795         PR c/83844
10796         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10797         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10798         If off is not INTEGER_CST, issue a may not be aligned warning
10799         rather than isn't aligned.  Use isn%'t rather than isn't.
10800         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10801         into MULT_EXPR.
10802         <case MULT_EXPR>: Improve the case when bottom and one of the
10803         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10804         operand, in that case check if the other operand is multiple of
10805         bottom divided by the INTEGER_CST operand.
10806
10807 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
10808
10809         PR target/83858
10810         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10811         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10812         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10813         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10814         * config/pa/pa.c (pa_function_arg_advance): Likewise.
10815         (pa_function_arg, pa_arg_partial_bytes): Likewise.
10816         (pa_function_arg_size): New function.
10817
10818 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
10819
10820         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10821         in a separate statement.
10822
10823 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
10824
10825         PR tree-optimization/83847
10826         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10827         group gathers and scatters.
10828
10829 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
10830
10831         PR rtl-optimization/86620
10832         * params.def (max-sched-ready-insns): Bump minimum value to 1.
10833
10834         PR rtl-optimization/83213
10835         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10836         to last if both are JUMP_INSNs.
10837
10838         PR tree-optimization/83843
10839         * gimple-ssa-store-merging.c
10840         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10841         store_immediate_info for bswap/nop orig_stores.
10842
10843 2018-01-15  Andrew Waterman  <andrew@sifive.com>
10844
10845         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10846         !TARGET_MUL.
10847         <UDIV>: Increase cost if !TARGET_DIV.
10848
10849 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
10850
10851         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10852         (define_attr "cr_logical_3op"): New.
10853         (cceq_ior_compare): Adjust.
10854         (cceq_ior_compare_complement): Adjust.
10855         (*cceq_rev_compare): Adjust.
10856         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10857         (is_cracked_insn): Adjust.
10858         (insn_must_be_first_in_group): Adjust.
10859         * config/rs6000/40x.md: Adjust.
10860         * config/rs6000/440.md: Adjust.
10861         * config/rs6000/476.md: Adjust.
10862         * config/rs6000/601.md: Adjust.
10863         * config/rs6000/603.md: Adjust.
10864         * config/rs6000/6xx.md: Adjust.
10865         * config/rs6000/7450.md: Adjust.
10866         * config/rs6000/7xx.md: Adjust.
10867         * config/rs6000/8540.md: Adjust.
10868         * config/rs6000/cell.md: Adjust.
10869         * config/rs6000/e300c2c3.md: Adjust.
10870         * config/rs6000/e500mc.md: Adjust.
10871         * config/rs6000/e500mc64.md: Adjust.
10872         * config/rs6000/e5500.md: Adjust.
10873         * config/rs6000/e6500.md: Adjust.
10874         * config/rs6000/mpc.md: Adjust.
10875         * config/rs6000/power4.md: Adjust.
10876         * config/rs6000/power5.md: Adjust.
10877         * config/rs6000/power6.md: Adjust.
10878         * config/rs6000/power7.md: Adjust.
10879         * config/rs6000/power8.md: Adjust.
10880         * config/rs6000/power9.md: Adjust.
10881         * config/rs6000/rs64.md: Adjust.
10882         * config/rs6000/titan.md: Adjust.
10883
10884 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
10885
10886         * config/i386/predicates.md (indirect_branch_operand): Rewrite
10887         ix86_indirect_branch_register logic.
10888
10889 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
10890
10891         * config/i386/constraints.md (Bs): Update
10892         ix86_indirect_branch_register check.  Don't check
10893         ix86_indirect_branch_register with GOT_memory_operand.
10894         (Bw): Likewise.
10895         * config/i386/predicates.md (GOT_memory_operand): Don't check
10896         ix86_indirect_branch_register here.
10897         (GOT32_symbol_operand): Likewise.
10898
10899 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
10900
10901         * config/i386/predicates.md (constant_call_address_operand):
10902         Rewrite ix86_indirect_branch_register logic.
10903         (sibcall_insn_operand): Likewise.
10904
10905 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
10906
10907         * config/i386/constraints.md (Bs): Replace
10908         ix86_indirect_branch_thunk_register with
10909         ix86_indirect_branch_register.
10910         (Bw): Likewise.
10911         * config/i386/i386.md (indirect_jump): Likewise.
10912         (tablejump): Likewise.
10913         (*sibcall_memory): Likewise.
10914         (*sibcall_value_memory): Likewise.
10915         Peepholes of indirect call and jump via memory: Likewise.
10916         * config/i386/i386.opt: Likewise.
10917         * config/i386/predicates.md (indirect_branch_operand): Likewise.
10918         (GOT_memory_operand): Likewise.
10919         (call_insn_operand): Likewise.
10920         (sibcall_insn_operand): Likewise.
10921         (GOT32_symbol_operand): Likewise.
10922
10923 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
10924
10925         PR middle-end/83837
10926         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10927         type rather than type addr's type points to.
10928         (expand_omp_atomic_mutex): Likewise.
10929         (expand_omp_atomic): Likewise.
10930
10931 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
10932
10933         PR target/83839
10934         * config/i386/i386.c (output_indirect_thunk_function): Use
10935         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10936         for  __x86_return_thunk.
10937
10938 2018-01-15  Richard Biener  <rguenther@suse.de>
10939
10940         PR middle-end/83850
10941         * expmed.c (extract_bit_field_1): Fix typo.
10942
10943 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10944
10945         PR target/83687
10946         * config/arm/iterators.md (VF): New mode iterator.
10947         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10948         Remove integer-related logic from pattern.
10949         (neon_vabd<mode>_3): Likewise.
10950
10951 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
10952
10953         PR middle-end/82694
10954         * common.opt (fstrict-overflow): No longer an alias.
10955         (fwrapv-pointer): New option.
10956         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10957         also for pointer types based on flag_wrapv_pointer.
10958         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10959         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10960         opts->x_flag_wrapv got set.
10961         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10962         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10963         POINTER_TYPE_OVERFLOW_UNDEFINED.
10964         * match.pd: Likewise in address comparison pattern.
10965         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10966
10967 2018-01-15  Richard Biener  <rguenther@suse.de>
10968
10969         PR lto/83804
10970         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10971         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
10972         Reset type names to their identifier if their TYPE_DECL doesn't
10973         have linkage (and thus is used for ODR and devirt).
10974         (save_debug_info_for_decl): Remove.
10975         (save_debug_info_for_type): Likewise.
10976         (add_tree_to_fld_list): Adjust.
10977         * tree-pretty-print.c (dump_generic_node): Make dumping of
10978         type names more robust.
10979
10980 2018-01-15  Richard Biener  <rguenther@suse.de>
10981
10982         * BASE-VER: Bump to 8.0.1.
10983
10984 2018-01-14  Martin Sebor  <msebor@redhat.com>
10985
10986         PR other/83508
10987         * builtins.c (check_access): Avoid warning when the no-warning bit
10988         is set.
10989
10990 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
10991
10992         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10993         * ira-color (allocno_hard_regs_compare): Likewise.
10994
10995 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
10996
10997         PR target/83013
10998         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10999         Use .pushsection/.popsection.
11000
11001 2018-01-14  Martin Sebor  <msebor@redhat.com>
11002
11003         PR c++/81327
11004         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
11005
11006 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
11007
11008         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
11009         entry from extra_headers.
11010         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
11011         extra_headers, make the list bitwise identical to the i?86-*-* one.
11012
11013 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
11014
11015         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
11016         -mcmodel=large with -mindirect-branch=thunk,
11017         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
11018         -mfunction-return=thunk-extern.
11019         * doc/invoke.texi: Document -mcmodel=large is incompatible with
11020         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
11021         -mfunction-return=thunk and -mfunction-return=thunk-extern.
11022
11023 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
11024
11025         * config/i386/i386.c (print_reg): Print the name of the full
11026         integer register without '%'.
11027         (ix86_print_operand): Handle 'V'.
11028         * doc/extend.texi: Document 'V' modifier.
11029
11030 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
11031
11032         * config/i386/constraints.md (Bs): Disallow memory operand for
11033         -mindirect-branch-register.
11034         (Bw): Likewise.
11035         * config/i386/predicates.md (indirect_branch_operand): Likewise.
11036         (GOT_memory_operand): Likewise.
11037         (call_insn_operand): Likewise.
11038         (sibcall_insn_operand): Likewise.
11039         (GOT32_symbol_operand): Likewise.
11040         * config/i386/i386.md (indirect_jump): Call convert_memory_address
11041         for -mindirect-branch-register.
11042         (tablejump): Likewise.
11043         (*sibcall_memory): Likewise.
11044         (*sibcall_value_memory): Likewise.
11045         Disallow peepholes of indirect call and jump via memory for
11046         -mindirect-branch-register.
11047         (*call_pop): Replace m with Bw.
11048         (*call_value_pop): Likewise.
11049         (*sibcall_pop_memory): Replace m with Bs.
11050         * config/i386/i386.opt (mindirect-branch-register): New option.
11051         * doc/invoke.texi: Document -mindirect-branch-register option.
11052
11053 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
11054
11055         * config/i386/i386-protos.h (ix86_output_function_return): New.
11056         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
11057         set function_return_type.
11058         (indirect_thunk_name): Add ret_p to indicate thunk for function
11059         return.
11060         (output_indirect_thunk_function): Pass false to
11061         indirect_thunk_name.
11062         (ix86_output_indirect_branch_via_reg): Likewise.
11063         (ix86_output_indirect_branch_via_push): Likewise.
11064         (output_indirect_thunk_function): Create alias for function
11065         return thunk if regno < 0.
11066         (ix86_output_function_return): New function.
11067         (ix86_handle_fndecl_attribute): Handle function_return.
11068         (ix86_attribute_table): Add function_return.
11069         * config/i386/i386.h (machine_function): Add
11070         function_return_type.
11071         * config/i386/i386.md (simple_return_internal): Use
11072         ix86_output_function_return.
11073         (simple_return_internal_long): Likewise.
11074         * config/i386/i386.opt (mfunction-return=): New option.
11075         (indirect_branch): Mention -mfunction-return=.
11076         * doc/extend.texi: Document function_return function attribute.
11077         * doc/invoke.texi: Document -mfunction-return= option.
11078
11079 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
11080
11081         * config/i386/i386-opts.h (indirect_branch): New.
11082         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
11083         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
11084         with local indirect jump when converting indirect call and jump.
11085         (ix86_set_indirect_branch_type): New.
11086         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
11087         (indirectlabelno): New.
11088         (indirect_thunk_needed): Likewise.
11089         (indirect_thunk_bnd_needed): Likewise.
11090         (indirect_thunks_used): Likewise.
11091         (indirect_thunks_bnd_used): Likewise.
11092         (INDIRECT_LABEL): Likewise.
11093         (indirect_thunk_name): Likewise.
11094         (output_indirect_thunk): Likewise.
11095         (output_indirect_thunk_function): Likewise.
11096         (ix86_output_indirect_branch_via_reg): Likewise.
11097         (ix86_output_indirect_branch_via_push): Likewise.
11098         (ix86_output_indirect_branch): Likewise.
11099         (ix86_output_indirect_jmp): Likewise.
11100         (ix86_code_end): Call output_indirect_thunk_function if needed.
11101         (ix86_output_call_insn): Call ix86_output_indirect_branch if
11102         needed.
11103         (ix86_handle_fndecl_attribute): Handle indirect_branch.
11104         (ix86_attribute_table): Add indirect_branch.
11105         * config/i386/i386.h (machine_function): Add indirect_branch_type
11106         and has_local_indirect_jump.
11107         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
11108         to true.
11109         (tablejump): Likewise.
11110         (*indirect_jump): Use ix86_output_indirect_jmp.
11111         (*tablejump_1): Likewise.
11112         (simple_return_indirect_internal): Likewise.
11113         * config/i386/i386.opt (mindirect-branch=): New option.
11114         (indirect_branch): New.
11115         (keep): Likewise.
11116         (thunk): Likewise.
11117         (thunk-inline): Likewise.
11118         (thunk-extern): Likewise.
11119         * doc/extend.texi: Document indirect_branch function attribute.
11120         * doc/invoke.texi: Document -mindirect-branch= option.
11121
11122 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
11123
11124         PR ipa/83051
11125         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
11126
11127 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
11128
11129         * ipa-inline.c (want_inline_small_function_p): Return false if
11130         inlining has already failed with CIF_FINAL_ERROR.
11131         (update_caller_keys): Call want_inline_small_function_p before
11132         can_inline_edge_p.
11133         (update_callee_keys): Likewise.
11134
11135 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11136
11137         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
11138         New function.
11139         (rs6000_quadword_masked_address_p): Likewise.
11140         (quad_aligned_load_p): Likewise.
11141         (quad_aligned_store_p): Likewise.
11142         (const_load_sequence_p): Add comment to describe the outer-most loop.
11143         (mimic_memory_attributes_and_flags): New function.
11144         (rs6000_gen_stvx): Likewise.
11145         (replace_swapped_aligned_store): Likewise.
11146         (rs6000_gen_lvx): Likewise.
11147         (replace_swapped_aligned_load): Likewise.
11148         (replace_swapped_load_constant): Capitalize argument name in
11149         comment describing this function.
11150         (rs6000_analyze_swaps): Add a third pass to search for vector loads
11151         and stores that access quad-word aligned addresses and replace
11152         with stvx or lvx instructions when appropriate.
11153         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
11154         New function prototype.
11155         (rs6000_quadword_masked_address_p): Likewise.
11156         (rs6000_gen_lvx): Likewise.
11157         (rs6000_gen_stvx): Likewise.
11158         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
11159         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
11160         when memory address is aligned.
11161         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
11162         this split to select lvx instruction when memory address is aligned.
11163         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
11164         instruction when memory address is aligned.
11165         (*vsx_le_perm_load_v16qi): Likewise.
11166         (four unnamed splitters): Modify to select the stvx instruction
11167         when memory is aligned.
11168
11169 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
11170
11171         * predict.c (determine_unlikely_bbs): Handle correctly BBs
11172         which appears in the queue multiple times.
11173
11174 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11175             Alan Hayward  <alan.hayward@arm.com>
11176             David Sherwood  <david.sherwood@arm.com>
11177
11178         * tree-vectorizer.h (vec_lower_bound): New structure.
11179         (_loop_vec_info): Add check_nonzero and lower_bounds.
11180         (LOOP_VINFO_CHECK_NONZERO): New macro.
11181         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
11182         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
11183         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
11184         fields.  Make seg_len the distance travelled, not including the
11185         access size.
11186         (dr_direction_indicator): Declare.
11187         (dr_zero_step_indicator): Likewise.
11188         (dr_known_forward_stride_p): Likewise.
11189         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
11190         tree-ssanames.h.
11191         (runtime_alias_check_p): Allow runtime alias checks with
11192         variable strides.
11193         (operator ==): Compare access_size and align.
11194         (prune_runtime_alias_test_list): Rework for new distinction between
11195         the access_size and seg_len.
11196         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
11197         segment lengths.
11198         (get_segment_min_max): New function.
11199         (create_intersect_range_checks): Use it.
11200         (dr_step_indicator): New function.
11201         (dr_direction_indicator): Likewise.
11202         (dr_zero_step_indicator): Likewise.
11203         (dr_known_forward_stride_p): Likewise.
11204         * tree-loop-distribution.c (data_ref_segment_size): Return
11205         DR_STEP * (niters - 1).
11206         (compute_alias_check_pairs): Update call to the dr_with_seg_len
11207         constructor.
11208         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
11209         (vect_preserves_scalar_order_p): New function, split out from...
11210         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
11211         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
11212         (vect_vfa_access_size): New function.
11213         (vect_vfa_align): Likewise.
11214         (vect_compile_time_alias): Take access_size_a and access_b arguments.
11215         (dump_lower_bound): New function.
11216         (vect_check_lower_bound): Likewise.
11217         (vect_small_gap_p): Likewise.
11218         (vectorizable_with_step_bound_p): Likewise.
11219         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
11220         depencies if the vectorization factor is 1.  Convert the checks
11221         for nonzero steps into checks on the bounds of DR_STEP.  Try using
11222         a bunds check for variable steps if the minimum required step is
11223         relatively small. Update calls to the dr_with_seg_len
11224         constructor and to vect_compile_time_alias.
11225         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11226         function.
11227         (vect_loop_versioning): Call it.
11228         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11229         when retrying.
11230         (vect_estimate_min_profitable_iters): Account for any bounds checks.
11231
11232 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11233             Alan Hayward  <alan.hayward@arm.com>
11234             David Sherwood  <david.sherwood@arm.com>
11235
11236         * doc/sourcebuild.texi (vect_scatter_store): Document.
11237         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11238         optabs.
11239         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11240         Document.
11241         * genopinit.c (main): Add supports_vec_scatter_store and
11242         supports_vec_scatter_store_cached to target_optabs.
11243         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11244         IFN_MASK_SCATTER_STORE.
11245         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11246         functions.
11247         * internal-fn.h (internal_store_fn_p): Declare.
11248         (internal_fn_stored_value_index): Likewise.
11249         * internal-fn.c (scatter_store_direct): New macro.
11250         (expand_scatter_store_optab_fn): New function.
11251         (direct_scatter_store_optab_supported_p): New macro.
11252         (internal_store_fn_p): New function.
11253         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11254         IFN_MASK_SCATTER_STORE.
11255         (internal_fn_mask_index): Likewise.
11256         (internal_fn_stored_value_index): New function.
11257         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11258         for scatter stores.
11259         * optabs-query.h (supports_vec_scatter_store_p): Declare.
11260         * optabs-query.c (supports_vec_scatter_store_p): New function.
11261         * tree-vectorizer.h (vect_get_store_rhs): Declare.
11262         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11263         true for scatter stores.
11264         (vect_gather_scatter_fn_p): Handle scatter stores too.
11265         (vect_check_gather_scatter): Consider using scatter stores if
11266         supports_vec_scatter_store_p.
11267         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11268         scatter stores too.
11269         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11270         internal_fn_stored_value_index.
11271         (check_load_store_masking): Handle scatter stores too.
11272         (vect_get_store_rhs): Make public.
11273         (vectorizable_call): Use internal_store_fn_p.
11274         (vectorizable_store): Handle scatter store internal functions.
11275         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11276         when deciding whether the end of the group has been reached.
11277         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11278         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11279         (mask_scatter_store<mode>): New insns.
11280
11281 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11282             Alan Hayward  <alan.hayward@arm.com>
11283             David Sherwood  <david.sherwood@arm.com>
11284
11285         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11286         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11287         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11288         function.
11289         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11290         Use vect_truncate_gather_scatter_offset if we can't treat the
11291         operation as a normal gather load or scatter store.
11292         (get_group_load_store_type): Take the gather_scatter_info
11293         as argument.  Try using a gather load or scatter store for
11294         single-element groups.
11295         (get_load_store_type): Update calls to get_group_load_store_type
11296         and vect_use_strided_gather_scatters_p.
11297
11298 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11299             Alan Hayward  <alan.hayward@arm.com>
11300             David Sherwood  <david.sherwood@arm.com>
11301
11302         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11303         optional tree argument.
11304         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11305         null target hooks.
11306         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11307         but continue to use the current value as a fallback.
11308         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11309         to compare the updates.
11310         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11311         (get_load_store_type): Use it when handling a strided access.
11312         (vect_get_strided_load_store_ops): New function.
11313         (vect_get_data_ptr_increment): Likewise.
11314         (vectorizable_load): Handle strided gather loads.  Always pass
11315         a step to vect_create_data_ref_ptr and bump_vector_ptr.
11316
11317 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11318             Alan Hayward  <alan.hayward@arm.com>
11319             David Sherwood  <david.sherwood@arm.com>
11320
11321         * doc/md.texi (gather_load@var{m}): Document.
11322         (mask_gather_load@var{m}): Likewise.
11323         * genopinit.c (main): Add supports_vec_gather_load and
11324         supports_vec_gather_load_cached to target_optabs.
11325         * optabs-tree.c (init_tree_optimization_optabs): Use
11326         ggc_cleared_alloc to allocate target_optabs.
11327         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11328         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11329         functions.
11330         * internal-fn.h (internal_load_fn_p): Declare.
11331         (internal_gather_scatter_fn_p): Likewise.
11332         (internal_fn_mask_index): Likewise.
11333         (internal_gather_scatter_fn_supported_p): Likewise.
11334         * internal-fn.c (gather_load_direct): New macro.
11335         (expand_gather_load_optab_fn): New function.
11336         (direct_gather_load_optab_supported_p): New macro.
11337         (direct_internal_fn_optab): New function.
11338         (internal_load_fn_p): Likewise.
11339         (internal_gather_scatter_fn_p): Likewise.
11340         (internal_fn_mask_index): Likewise.
11341         (internal_gather_scatter_fn_supported_p): Likewise.
11342         * optabs-query.c (supports_at_least_one_mode_p): New function.
11343         (supports_vec_gather_load_p): Likewise.
11344         * optabs-query.h (supports_vec_gather_load_p): Declare.
11345         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11346         and memory_type field.
11347         (NUM_PATTERNS): Bump to 15.
11348         * tree-vect-data-refs.c: Include internal-fn.h.
11349         (vect_gather_scatter_fn_p): New function.
11350         (vect_describe_gather_scatter_call): Likewise.
11351         (vect_check_gather_scatter): Try using internal functions for
11352         gather loads.  Recognize existing calls to a gather load function.
11353         (vect_analyze_data_refs): Consider using gather loads if
11354         supports_vec_gather_load_p.
11355         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11356         (vect_get_gather_scatter_offset_type): Likewise.
11357         (vect_convert_mask_for_vectype): Likewise.
11358         (vect_add_conversion_to_patterm): Likewise.
11359         (vect_try_gather_scatter_pattern): Likewise.
11360         (vect_recog_gather_scatter_pattern): New pattern recognizer.
11361         (vect_vect_recog_func_ptrs): Add it.
11362         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11363         internal_fn_mask_index and internal_gather_scatter_fn_p.
11364         (check_load_store_masking): Take the gather_scatter_info as an
11365         argument and handle gather loads.
11366         (vect_get_gather_scatter_ops): New function.
11367         (vectorizable_call): Check internal_load_fn_p.
11368         (vectorizable_load): Likewise.  Handle gather load internal
11369         functions.
11370         (vectorizable_store): Update call to check_load_store_masking.
11371         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11372         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11373         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11374         (aarch64_gather_scale_operand_d): New predicates.
11375         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11376         (mask_gather_load<mode>): New insns.
11377
11378 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11379             Alan Hayward  <alan.hayward@arm.com>
11380             David Sherwood  <david.sherwood@arm.com>
11381
11382         * optabs.def (fold_left_plus_optab): New optab.
11383         * doc/md.texi (fold_left_plus_@var{m}): Document.
11384         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11385         * internal-fn.c (fold_left_direct): Define.
11386         (expand_fold_left_optab_fn): Likewise.
11387         (direct_fold_left_optab_supported_p): Likewise.
11388         * fold-const-call.c (fold_const_fold_left): New function.
11389         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11390         * tree-parloops.c (valid_reduction_p): New function.
11391         (gather_scalar_reductions): Use it.
11392         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11393         (vect_finish_replace_stmt): Declare.
11394         * tree-vect-loop.c (fold_left_reduction_fn): New function.
11395         (needs_fold_left_reduction_p): New function, split out from...
11396         (vect_is_simple_reduction): ...here.  Accept reductions that
11397         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11398         (vect_force_simple_reduction): Also store the reduction type in
11399         the assignment's STMT_VINFO_REDUC_TYPE.
11400         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11401         (merge_with_identity): New function.
11402         (vect_expand_fold_left): Likewise.
11403         (vectorize_fold_left_reduction): Likewise.
11404         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
11405         scalar phi in place for it.  Check for target support and reject
11406         cases that would reassociate the operation.  Defer the transform
11407         phase to vectorize_fold_left_reduction.
11408         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11409         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11410         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11411
11412 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11413
11414         * tree-if-conv.c (predicate_mem_writes): Remove redundant
11415         call to ifc_temp_var.
11416
11417 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11418             Alan Hayward  <alan.hayward@arm.com>
11419             David Sherwood  <david.sherwood@arm.com>
11420
11421         * target.def (legitimize_address_displacement): Take the original
11422         offset as a poly_int.
11423         * targhooks.h (default_legitimize_address_displacement): Update
11424         accordingly.
11425         * targhooks.c (default_legitimize_address_displacement): Likewise.
11426         * doc/tm.texi: Regenerate.
11427         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11428         as an argument, moving assert of ad->disp == ad->disp_term to...
11429         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
11430         Try calling targetm.legitimize_address_displacement before expanding
11431         the address rather than afterwards, and adjust for the new interface.
11432         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11433         Match the new hook interface.  Handle SVE addresses.
11434         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11435         new hook interface.
11436
11437 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11438
11439         * Makefile.in (OBJS): Add early-remat.o.
11440         * target.def (select_early_remat_modes): New hook.
11441         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11442         * doc/tm.texi: Regenerate.
11443         * targhooks.h (default_select_early_remat_modes): Declare.
11444         * targhooks.c (default_select_early_remat_modes): New function.
11445         * timevar.def (TV_EARLY_REMAT): New timevar.
11446         * passes.def (pass_early_remat): New pass.
11447         * tree-pass.h (make_pass_early_remat): Declare.
11448         * early-remat.c: New file.
11449         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11450         function.
11451         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11452
11453 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11454             Alan Hayward  <alan.hayward@arm.com>
11455             David Sherwood  <david.sherwood@arm.com>
11456
11457         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11458         vfm1 with a bound_epilog parameter.
11459         (vect_do_peeling): Update calls accordingly, and move the prologue
11460         call earlier in the function.  Treat the base bound_epilog as 0 for
11461         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
11462         this base when peeling for gaps.
11463         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11464         with fully-masked loops.
11465         (vect_estimate_min_profitable_iters): Handle the single peeled
11466         iteration in that case.
11467
11468 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11469             Alan Hayward  <alan.hayward@arm.com>
11470             David Sherwood  <david.sherwood@arm.com>
11471
11472         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11473         single-element interleaving even if the size is not a power of 2.
11474         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11475         accesses for single-element interleaving if the group size is
11476         not a power of 2.
11477
11478 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11479             Alan Hayward  <alan.hayward@arm.com>
11480             David Sherwood  <david.sherwood@arm.com>
11481
11482         * doc/md.texi (fold_extract_last_@var{m}): Document.
11483         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11484         * optabs.def (fold_extract_last_optab): New optab.
11485         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11486         * internal-fn.c (fold_extract_direct): New macro.
11487         (expand_fold_extract_optab_fn): Likewise.
11488         (direct_fold_extract_optab_supported_p): Likewise.
11489         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11490         * tree-vect-loop.c (vect_model_reduction_cost): Handle
11491         EXTRACT_LAST_REDUCTION.
11492         (get_initial_def_for_reduction): Do not create an initial vector
11493         for EXTRACT_LAST_REDUCTION reductions.
11494         (vectorizable_reduction): Leave the scalar phi in place for
11495         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
11496         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
11497         epilogue code for EXTRACT_LAST_REDUCTION and defer the
11498         transform phase to vectorizable_condition.
11499         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11500         split out from...
11501         (vect_finish_stmt_generation): ...here.
11502         (vect_finish_replace_stmt): New function.
11503         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11504         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11505         pattern.
11506         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11507
11508 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11509             Alan Hayward  <alan.hayward@arm.com>
11510             David Sherwood  <david.sherwood@arm.com>
11511
11512         * doc/md.texi (extract_last_@var{m}): Document.
11513         * optabs.def (extract_last_optab): New optab.
11514         * internal-fn.def (EXTRACT_LAST): New internal function.
11515         * internal-fn.c (cond_unary_direct): New macro.
11516         (expand_cond_unary_optab_fn): Likewise.
11517         (direct_cond_unary_optab_supported_p): Likewise.
11518         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11519         loops using EXTRACT_LAST.
11520         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11521         (extract_last_<mode>): ...this optab.
11522         (vec_extract<mode><Vel>): Update accordingly.
11523
11524 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11525             Alan Hayward  <alan.hayward@arm.com>
11526             David Sherwood  <david.sherwood@arm.com>
11527
11528         * target.def (empty_mask_is_expensive): New hook.
11529         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11530         * doc/tm.texi: Regenerate.
11531         * targhooks.h (default_empty_mask_is_expensive): Declare.
11532         * targhooks.c (default_empty_mask_is_expensive): New function.
11533         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11534         if the target says that empty masks are expensive.
11535         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11536         New function.
11537         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11538
11539 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11540             Alan Hayward  <alan.hayward@arm.com>
11541             David Sherwood  <david.sherwood@arm.com>
11542
11543         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11544         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11545         (vect_use_loop_mask_for_alignment_p): New function.
11546         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11547         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11548         niters_skip argument.  Make sure that the first niters_skip elements
11549         of the first iteration are inactive.
11550         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11551         Update call to vect_set_loop_masks_directly.
11552         (get_misalign_in_elems): New function, split out from...
11553         (vect_gen_prolog_loop_niters): ...here.
11554         (vect_update_init_of_dr): Take a code argument that specifies whether
11555         the adjustment should be added or subtracted.
11556         (vect_update_init_of_drs): Likewise.
11557         (vect_prepare_for_masked_peels): New function.
11558         (vect_do_peeling): Skip prologue peeling if we're using a mask
11559         instead.  Update call to vect_update_inits_of_drs.
11560         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11561         mask_skip_niters.
11562         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11563         alignment.  Do not include the number of peeled iterations in
11564         the minimum threshold in that case.
11565         (vectorizable_induction): Adjust the start value down by
11566         LOOP_VINFO_MASK_SKIP_NITERS iterations.
11567         (vect_transform_loop): Call vect_prepare_for_masked_peels.
11568         Take the number of skipped iterations into account when calculating
11569         the loop bounds.
11570         * tree-vect-stmts.c (vect_gen_while_not): New function.
11571
11572 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11573             Alan Hayward  <alan.hayward@arm.com>
11574             David Sherwood  <david.sherwood@arm.com>
11575
11576         * doc/sourcebuild.texi (vect_fully_masked): Document.
11577         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11578         default value to 0.
11579         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11580         split out from...
11581         (vect_analyze_loop_2): ...here. Don't check the vectorization
11582         factor against the number of loop iterations if the loop is
11583         fully-masked.
11584
11585 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11586             Alan Hayward  <alan.hayward@arm.com>
11587             David Sherwood  <david.sherwood@arm.com>
11588
11589         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11590         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11591         (dump_groups): Update accordingly.
11592         (iv_use::mem_type): New member variable.
11593         (address_p): New function.
11594         (record_use): Add a mem_type argument and initialize the new
11595         mem_type field.
11596         (record_group_use): Add a mem_type argument.  Use address_p.
11597         Remove obsolete null checks of base_object.  Update call to record_use.
11598         (find_interesting_uses_op): Update call to record_group_use.
11599         (find_interesting_uses_cond): Likewise.
11600         (find_interesting_uses_address): Likewise.
11601         (get_mem_type_for_internal_fn): New function.
11602         (find_address_like_use): Likewise.
11603         (find_interesting_uses_stmt): Try find_address_like_use before
11604         calling find_interesting_uses_op.
11605         (addr_offset_valid_p): Use the iv mem_type field as the type
11606         of the addressed memory.
11607         (add_autoinc_candidates): Likewise.
11608         (get_address_cost): Likewise.
11609         (split_small_address_groups_p): Use address_p.
11610         (split_address_groups): Likewise.
11611         (add_iv_candidate_for_use): Likewise.
11612         (autoinc_possible_for_pair): Likewise.
11613         (rewrite_groups): Likewise.
11614         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11615         (determine_group_iv_cost): Update after split of USE_ADDRESS.
11616         (get_alias_ptr_type_for_ptr_address): New function.
11617         (rewrite_use_address): Rewrite address uses in calls that were
11618         identified by find_address_like_use.
11619
11620 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11621             Alan Hayward  <alan.hayward@arm.com>
11622             David Sherwood  <david.sherwood@arm.com>
11623
11624         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11625         TARGET_MEM_REFs.
11626         * gimple-expr.h (is_gimple_addressable: Likewise.
11627         * gimple-expr.c (is_gimple_address): Likewise.
11628         * internal-fn.c (expand_call_mem_ref): New function.
11629         (expand_mask_load_optab_fn): Use it.
11630         (expand_mask_store_optab_fn): Likewise.
11631
11632 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11633             Alan Hayward  <alan.hayward@arm.com>
11634             David Sherwood  <david.sherwood@arm.com>
11635
11636         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11637         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11638         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11639         (cond_umax@var{mode}): Document.
11640         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11641         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11642         (cond_umin_optab, cond_umax_optab): New optabs.
11643         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11644         (COND_IOR, COND_XOR): New internal functions.
11645         * internal-fn.h (get_conditional_internal_fn): Declare.
11646         * internal-fn.c (cond_binary_direct): New macro.
11647         (expand_cond_binary_optab_fn): Likewise.
11648         (direct_cond_binary_optab_supported_p): Likewise.
11649         (get_conditional_internal_fn): New function.
11650         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11651         Cope with reduction statements that are vectorized as calls rather
11652         than assignments.
11653         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11654         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11655         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11656         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11657         (UNSPEC_COND_EOR): New unspecs.
11658         (optab): Add mappings for them.
11659         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11660         (sve_int_op, sve_fp_op): New int attributes.
11661
11662 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11663             Alan Hayward  <alan.hayward@arm.com>
11664             David Sherwood  <david.sherwood@arm.com>
11665
11666         * optabs.def (while_ult_optab): New optab.
11667         * doc/md.texi (while_ult@var{m}@var{n}): Document.
11668         * internal-fn.def (WHILE_ULT): New internal function.
11669         * internal-fn.h (direct_internal_fn_supported_p): New override
11670         that takes two types as argument.
11671         * internal-fn.c (while_direct): New macro.
11672         (expand_while_optab_fn): New function.
11673         (convert_optab_supported_p): Likewise.
11674         (direct_while_optab_supported_p): New macro.
11675         * wide-int.h (wi::udiv_ceil): New function.
11676         * tree-vectorizer.h (rgroup_masks): New structure.
11677         (vec_loop_masks): New typedef.
11678         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11679         and fully_masked_p.
11680         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11681         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11682         (vect_max_vf): New function.
11683         (slpeel_make_loop_iterate_ntimes): Delete.
11684         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11685         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11686         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11687         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11688         internal-fn.h, stor-layout.h and optabs-query.h.
11689         (vect_set_loop_mask): New function.
11690         (add_preheader_seq): Likewise.
11691         (add_header_seq): Likewise.
11692         (interleave_supported_p): Likewise.
11693         (vect_maybe_permute_loop_masks): Likewise.
11694         (vect_set_loop_masks_directly): Likewise.
11695         (vect_set_loop_condition_masked): Likewise.
11696         (vect_set_loop_condition_unmasked): New function, split out from
11697         slpeel_make_loop_iterate_ntimes.
11698         (slpeel_make_loop_iterate_ntimes): Rename to..
11699         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
11700         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11701         (vect_do_peeling): Update call accordingly.
11702         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11703         loops.
11704         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11705         mask_compare_type, can_fully_mask_p and fully_masked_p.
11706         (release_vec_loop_masks): New function.
11707         (_loop_vec_info): Use it to free the loop masks.
11708         (can_produce_all_loop_masks_p): New function.
11709         (vect_get_max_nscalars_per_iter): Likewise.
11710         (vect_verify_full_masking): Likewise.
11711         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11712         retries, and free the mask rgroups before retrying.  Check loop-wide
11713         reasons for disallowing fully-masked loops.  Make the final decision
11714         about whether use a fully-masked loop or not.
11715         (vect_estimate_min_profitable_iters): Do not assume that peeling
11716         for the number of iterations will be needed for fully-masked loops.
11717         (vectorizable_reduction): Disable fully-masked loops.
11718         (vectorizable_live_operation): Likewise.
11719         (vect_halve_mask_nunits): New function.
11720         (vect_double_mask_nunits): Likewise.
11721         (vect_record_loop_mask): Likewise.
11722         (vect_get_loop_mask): Likewise.
11723         (vect_transform_loop): Handle the case in which the final loop
11724         iteration might handle a partial vector.  Call vect_set_loop_condition
11725         instead of slpeel_make_loop_iterate_ntimes.
11726         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11727         (check_load_store_masking): New function.
11728         (prepare_load_store_mask): Likewise.
11729         (vectorizable_store): Handle fully-masked loops.
11730         (vectorizable_load): Likewise.
11731         (supportable_widening_operation): Use vect_halve_mask_nunits for
11732         booleans.
11733         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11734         (vect_gen_while): New function.
11735         * config/aarch64/aarch64.md (umax<mode>3): New expander.
11736         (aarch64_uqdec<mode>): New insn.
11737
11738 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11739             Alan Hayward  <alan.hayward@arm.com>
11740             David Sherwood  <david.sherwood@arm.com>
11741
11742         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11743         (reduc_xor_scal_optab): New optabs.
11744         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11745         (reduc_xor_scal_@var{m}): Document.
11746         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11747         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11748         internal functions.
11749         * fold-const-call.c (fold_const_call): Handle them.
11750         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11751         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11752         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11753         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11754         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11755         (UNSPEC_XORV): New unspecs.
11756         (optab): Add entries for them.
11757         (BITWISEV): New int iterator.
11758         (bit_reduc_op): New int attributes.
11759
11760 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11761             Alan Hayward  <alan.hayward@arm.com>
11762             David Sherwood  <david.sherwood@arm.com>
11763
11764         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11765         * internal-fn.def (VEC_SHL_INSERT): New internal function.
11766         * optabs.def (vec_shl_insert_optab): New optab.
11767         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11768         (duplicate_and_interleave): Likewise.
11769         * tree-vect-loop.c: Include internal-fn.h.
11770         (neutral_op_for_slp_reduction): New function, split out from
11771         get_initial_defs_for_reduction.
11772         (get_initial_def_for_reduction): Handle option 2 for variable-length
11773         vectors by loading the neutral value into a vector and then shifting
11774         the initial value into element 0.
11775         (get_initial_defs_for_reduction): Replace the code argument with
11776         the neutral value calculated by neutral_op_for_slp_reduction.
11777         Use gimple_build_vector for constant-length vectors.
11778         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11779         but the first group_size elements have a neutral value.
11780         Use duplicate_and_interleave otherwise.
11781         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11782         Update call to get_initial_defs_for_reduction.  Handle SLP
11783         reductions for variable-length vectors by creating one vector
11784         result for each scalar result, with the elements associated
11785         with other scalar results stubbed out with the neutral value.
11786         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11787         Require IFN_VEC_SHL_INSERT for double reductions on
11788         variable-length vectors, or SLP reductions that have
11789         a neutral value.  Require can_duplicate_and_interleave_p
11790         support for variable-length unchained SLP reductions if there
11791         is no neutral value, such as for MIN/MAX reductions.  Also require
11792         the number of vector elements to be a multiple of the number of
11793         SLP statements when doing variable-length unchained SLP reductions.
11794         Update call to vect_create_epilog_for_reduction.
11795         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11796         and remove initial values.
11797         (duplicate_and_interleave): Make public.
11798         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11799         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11800
11801 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11802             Alan Hayward  <alan.hayward@arm.com>
11803             David Sherwood  <david.sherwood@arm.com>
11804
11805         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11806         (can_duplicate_and_interleave_p): New function.
11807         (vect_get_and_check_slp_defs): Take the vector of statements
11808         rather than just the current one.  Remove excess parentheses.
11809         Restriction rejectinon of vect_constant_def and vect_external_def
11810         for variable-length vectors to boolean types, or types for which
11811         can_duplicate_and_interleave_p is false.
11812         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11813         (duplicate_and_interleave): New function.
11814         (vect_get_constant_vectors): Use gimple_build_vector for
11815         constant-length vectors and suitable variable-length constant
11816         vectors.  Use duplicate_and_interleave for other variable-length
11817         vectors.  Don't defer the update when inserting new statements.
11818
11819 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11820             Alan Hayward  <alan.hayward@arm.com>
11821             David Sherwood  <david.sherwood@arm.com>
11822
11823         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11824         min_profitable_iters doesn't go negative.
11825
11826 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11827             Alan Hayward  <alan.hayward@arm.com>
11828             David Sherwood  <david.sherwood@arm.com>
11829
11830         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11831         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11832         * optabs.def (vec_mask_load_lanes_optab): New optab.
11833         (vec_mask_store_lanes_optab): Likewise.
11834         * internal-fn.def (MASK_LOAD_LANES): New internal function.
11835         (MASK_STORE_LANES): Likewise.
11836         * internal-fn.c (mask_load_lanes_direct): New macro.
11837         (mask_store_lanes_direct): Likewise.
11838         (expand_mask_load_optab_fn): Handle masked operations.
11839         (expand_mask_load_lanes_optab_fn): New macro.
11840         (expand_mask_store_optab_fn): Handle masked operations.
11841         (expand_mask_store_lanes_optab_fn): New macro.
11842         (direct_mask_load_lanes_optab_supported_p): Likewise.
11843         (direct_mask_store_lanes_optab_supported_p): Likewise.
11844         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11845         parameter.
11846         (vect_load_lanes_supported): Likewise.
11847         * tree-vect-data-refs.c (strip_conversion): New function.
11848         (can_group_stmts_p): Likewise.
11849         (vect_analyze_data_ref_accesses): Use it instead of checking
11850         for a pair of assignments.
11851         (vect_store_lanes_supported): Take a masked_p parameter.
11852         (vect_load_lanes_supported): Likewise.
11853         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11854         vect_store_lanes_supported and vect_load_lanes_supported.
11855         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11856         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11857         parameter.  Don't allow gaps for masked accesses.
11858         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
11859         and vect_load_lanes_supported.
11860         (get_load_store_type): Take a masked_p parameter and update
11861         call to get_group_load_store_type.
11862         (vectorizable_store): Update call to get_load_store_type.
11863         Handle IFN_MASK_STORE_LANES.
11864         (vectorizable_load): Update call to get_load_store_type.
11865         Handle IFN_MASK_LOAD_LANES.
11866
11867 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11868             Alan Hayward  <alan.hayward@arm.com>
11869             David Sherwood  <david.sherwood@arm.com>
11870
11871         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11872         modes for SVE.
11873         * config/aarch64/aarch64-protos.h
11874         (aarch64_sve_struct_memory_operand_p): Declare.
11875         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11876         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11877         (VPRED, vpred): Handle SVE structure modes.
11878         * config/aarch64/constraints.md (Utx): New constraint.
11879         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11880         (aarch64_sve_struct_nonimmediate_operand): New predicates.
11881         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11882         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11883         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11884         structure modes.  Split into pieces after RA.
11885         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11886         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11887         New patterns.
11888         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11889         SVE structure modes.
11890         (aarch64_classify_address): Likewise.
11891         (sizetochar): Move earlier in file.
11892         (aarch64_print_operand): Handle SVE register lists.
11893         (aarch64_array_mode): New function.
11894         (aarch64_sve_struct_memory_operand_p): Likewise.
11895         (TARGET_ARRAY_MODE): Redefine.
11896
11897 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11898             Alan Hayward  <alan.hayward@arm.com>
11899             David Sherwood  <david.sherwood@arm.com>
11900
11901         * target.def (array_mode): New target hook.
11902         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11903         * doc/tm.texi: Regenerate.
11904         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11905         * hooks.c (hook_optmode_mode_uhwi_none): New function.
11906         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11907         targetm.array_mode.
11908         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
11909         type sizes.
11910
11911 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11912             Alan Hayward  <alan.hayward@arm.com>
11913             David Sherwood  <david.sherwood@arm.com>
11914
11915         * fold-const.c (fold_binary_loc): Check the argument types
11916         rather than the result type when testing for a vector operation.
11917
11918 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11919
11920         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11921         * doc/tm.texi: Regenerate.
11922
11923 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11924             Alan Hayward  <alan.hayward@arm.com>
11925             David Sherwood  <david.sherwood@arm.com>
11926
11927         * doc/invoke.texi (-msve-vector-bits=): Document new option.
11928         (sve): Document new AArch64 extension.
11929         * doc/md.texi (w): Extend the description of the AArch64
11930         constraint to include SVE vectors.
11931         (Upl, Upa): Document new AArch64 predicate constraints.
11932         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11933         enum.
11934         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11935         (msve-vector-bits=): New option.
11936         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11937         SVE when these are disabled.
11938         (sve): New extension.
11939         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11940         modes.  Adjust their number of units based on aarch64_sve_vg.
11941         (MAX_BITSIZE_MODE_ANY_MODE): Define.
11942         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11943         aarch64_addr_query_type.
11944         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11945         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11946         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11947         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11948         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11949         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11950         (aarch64_simd_imm_zero_p): Delete.
11951         (aarch64_check_zero_based_sve_index_immediate): Declare.
11952         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11953         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11954         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11955         (aarch64_sve_float_mul_immediate_p): Likewise.
11956         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11957         rather than an rtx.
11958         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11959         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11960         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11961         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11962         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11963         (aarch64_regmode_natural_size): Likewise.
11964         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11965         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11966         left one place.
11967         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11968         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11969         for VG and the SVE predicate registers.
11970         (V_ALIASES): Add a "z"-prefixed alias.
11971         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11972         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11973         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11974         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11975         (REG_CLASS_NAMES): Add entries for them.
11976         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
11977         and the predicate registers.
11978         (aarch64_sve_vg): Declare.
11979         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11980         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11981         (REGMODE_NATURAL_SIZE): Define.
11982         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11983         SVE macros.
11984         * config/aarch64/aarch64.c: Include cfgrtl.h.
11985         (simd_immediate_info): Add a constructor for series vectors,
11986         and an associated step field.
11987         (aarch64_sve_vg): New variable.
11988         (aarch64_dbx_register_number): Handle VG and the predicate registers.
11989         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11990         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11991         (VEC_ANY_DATA, VEC_STRUCT): New constants.
11992         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11993         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11994         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11995         (aarch64_get_mask_mode): New functions.
11996         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11997         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11998         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
11999         predicate modes and predicate registers.  Explicitly restrict
12000         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
12001         to store a vector mode if it is recognized by
12002         aarch64_classify_vector_mode.
12003         (aarch64_regmode_natural_size): New function.
12004         (aarch64_hard_regno_caller_save_mode): Return the original mode
12005         for predicates.
12006         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
12007         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
12008         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
12009         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
12010         functions.
12011         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
12012         does not overlap dest if the function is frame-related.  Handle
12013         SVE constants.
12014         (aarch64_split_add_offset): New function.
12015         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
12016         them aarch64_add_offset.
12017         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
12018         and update call to aarch64_sub_sp.
12019         (aarch64_add_cfa_expression): New function.
12020         (aarch64_expand_prologue): Pass extra temporary registers to the
12021         functions above.  Handle the case in which we need to emit new
12022         DW_CFA_expressions for registers that were originally saved
12023         relative to the stack pointer, but now have to be expressed
12024         relative to the frame pointer.
12025         (aarch64_output_mi_thunk): Pass extra temporary registers to the
12026         functions above.
12027         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
12028         IP0 and IP1 values for SVE frames.
12029         (aarch64_expand_vec_series): New function.
12030         (aarch64_expand_sve_widened_duplicate): Likewise.
12031         (aarch64_expand_sve_const_vector): Likewise.
12032         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
12033         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
12034         into the register, rather than emitting a SET directly.
12035         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
12036         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
12037         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
12038         (offset_9bit_signed_scaled_p): New functions.
12039         (aarch64_replicate_bitmask_imm): New function.
12040         (aarch64_bitmask_imm): Use it.
12041         (aarch64_cannot_force_const_mem): Reject expressions involving
12042         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
12043         (aarch64_classify_index): Handle SVE indices, by requiring
12044         a plain register index with a scale that matches the element size.
12045         (aarch64_classify_address): Handle SVE addresses.  Assert that
12046         the mode of the address is VOIDmode or an integer mode.
12047         Update call to aarch64_classify_symbol.
12048         (aarch64_classify_symbolic_expression): Update call to
12049         aarch64_classify_symbol.
12050         (aarch64_const_vec_all_in_range_p): New function.
12051         (aarch64_print_vector_float_operand): Likewise.
12052         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
12053         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
12054         and the FP immediates 1.0 and 0.5.
12055         (aarch64_print_address_internal): Handle SVE addresses.
12056         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
12057         (aarch64_regno_regclass): Handle predicate registers.
12058         (aarch64_secondary_reload): Handle big-endian reloads of SVE
12059         data modes.
12060         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
12061         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
12062         (aarch64_convert_sve_vector_bits): New function.
12063         (aarch64_override_options): Use it to handle -msve-vector-bits=.
12064         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12065         rather than an rtx.
12066         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
12067         Handle SVE vector and predicate modes.  Accept VL-based constants
12068         that need only one temporary register, and VL offsets that require
12069         no temporary registers.
12070         (aarch64_conditional_register_usage): Mark the predicate registers
12071         as fixed if SVE isn't available.
12072         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
12073         Return true for SVE vector and predicate modes.
12074         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
12075         rather than an unsigned int.  Handle SVE modes.
12076         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
12077         SVE modes.
12078         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
12079         if SVE is enabled.
12080         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12081         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12082         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12083         (aarch64_sve_float_mul_immediate_p): New functions.
12084         (aarch64_sve_valid_immediate): New function.
12085         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
12086         Explicitly reject structure modes.  Check for INDEX constants.
12087         Handle PTRUE and PFALSE constants.
12088         (aarch64_check_zero_based_sve_index_immediate): New function.
12089         (aarch64_simd_imm_zero_p): Delete.
12090         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
12091         vector modes.  Accept constants in the range of CNT[BHWD].
12092         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
12093         ask for an Advanced SIMD mode.
12094         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
12095         (aarch64_simd_vector_alignment): Handle SVE predicates.
12096         (aarch64_vectorize_preferred_vector_alignment): New function.
12097         (aarch64_simd_vector_alignment_reachable): Use it instead of
12098         the vector size.
12099         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
12100         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
12101         functions.
12102         (MAX_VECT_LEN): Delete.
12103         (expand_vec_perm_d): Add a vec_flags field.
12104         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
12105         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
12106         (aarch64_evpc_ext): Don't apply a big-endian lane correction
12107         for SVE modes.
12108         (aarch64_evpc_rev): Rename to...
12109         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
12110         (aarch64_evpc_rev_global): New function.
12111         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
12112         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
12113         MAX_VECT_LEN.
12114         (aarch64_evpc_sve_tbl): New function.
12115         (aarch64_expand_vec_perm_const_1): Update after rename of
12116         aarch64_evpc_rev.  Handle SVE permutes too, trying
12117         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
12118         than aarch64_evpc_tbl.
12119         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
12120         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
12121         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
12122         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
12123         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
12124         (aarch64_expand_sve_vcond): New functions.
12125         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
12126         of aarch64_vector_mode_p.
12127         (aarch64_dwarf_poly_indeterminate_value): New function.
12128         (aarch64_compute_pressure_classes): Likewise.
12129         (aarch64_can_change_mode_class): Likewise.
12130         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
12131         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
12132         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
12133         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
12134         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
12135         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
12136         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
12137         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
12138         constraints.
12139         (Dn, Dl, Dr): Accept const as well as const_vector.
12140         (Dz): Likewise.  Compare against CONST0_RTX.
12141         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
12142         of "vector" where appropriate.
12143         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
12144         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
12145         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
12146         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
12147         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
12148         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
12149         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
12150         (v_int_equiv): Extend to SVE modes.
12151         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
12152         mode attributes.
12153         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
12154         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
12155         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
12156         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
12157         (SVE_COND_FP_CMP): New int iterators.
12158         (perm_hilo): Handle the new unpack unspecs.
12159         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
12160         attributes.
12161         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
12162         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
12163         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
12164         (aarch64_equality_operator, aarch64_constant_vector_operand)
12165         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
12166         (aarch64_sve_nonimmediate_operand): Likewise.
12167         (aarch64_sve_general_operand): Likewise.
12168         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
12169         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
12170         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
12171         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
12172         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
12173         (aarch64_sve_float_arith_immediate): Likewise.
12174         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
12175         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
12176         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
12177         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
12178         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
12179         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
12180         (aarch64_sve_float_arith_operand): Likewise.
12181         (aarch64_sve_float_arith_with_sub_operand): Likewise.
12182         (aarch64_sve_float_mul_operand): Likewise.
12183         (aarch64_sve_vec_perm_operand): Likewise.
12184         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
12185         (aarch64_mov_operand): Accept const_poly_int and const_vector.
12186         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
12187         as well as const_vector.
12188         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
12189         in file.  Use CONST0_RTX and CONSTM1_RTX.
12190         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
12191         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
12192         Use aarch64_simd_imm_zero.
12193         * config/aarch64/aarch64-sve.md: New file.
12194         * config/aarch64/aarch64.md: Include it.
12195         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
12196         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
12197         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
12198         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
12199         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
12200         (sve): New attribute.
12201         (enabled): Disable instructions with the sve attribute unless
12202         TARGET_SVE.
12203         (movqi, movhi): Pass CONST_POLY_INT operaneds through
12204         aarch64_expand_mov_immediate.
12205         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
12206         CNT[BHSD] immediates.
12207         (movti): Split CONST_POLY_INT moves into two halves.
12208         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
12209         Split additions that need a temporary here if the destination
12210         is the stack pointer.
12211         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
12212         (*add<mode>3_poly_1): New instruction.
12213         (set_clobber_cc): New expander.
12214
12215 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12216
12217         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
12218         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
12219         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
12220         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
12221         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
12222         Change innermode from fixed_mode_size to machine_mode.
12223         (simplify_subreg): Update call accordingly.  Handle a constant-sized
12224         subreg of a variable-length CONST_VECTOR.
12225
12226 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
12227             Alan Hayward  <alan.hayward@arm.com>
12228             David Sherwood  <david.sherwood@arm.com>
12229
12230         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12231         (add_offset_to_base): New function, split out from...
12232         (create_mem_ref): ...here.  When handling a scale other than 1,
12233         check first whether the address is valid without the offset.
12234         Add it into the base if so, leaving the index and scale as-is.
12235
12236 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
12237
12238         PR c++/83778
12239         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12240         fold_for_warn before checking if arg2 is INTEGER_CST.
12241
12242 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
12243
12244         * config/rs6000/predicates.md (load_multiple_operation): Delete.
12245         (store_multiple_operation): Delete.
12246         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12247         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12248         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12249         guarded by TARGET_STRING.
12250         (rs6000_output_load_multiple): Delete.
12251         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12252         OPTION_MASK_STRING / TARGET_STRING handling.
12253         (print_operand) <'N', 'O'>: Add comment that these are unused now.
12254         (const rs6000_opt_masks) <"string">: Change mask to 0.
12255         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12256         (MASK_STRING): Delete.
12257         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12258         parts.  Simplify.
12259         (load_multiple): Delete.
12260         (*ldmsi8): Delete.
12261         (*ldmsi7): Delete.
12262         (*ldmsi6): Delete.
12263         (*ldmsi5): Delete.
12264         (*ldmsi4): Delete.
12265         (*ldmsi3): Delete.
12266         (store_multiple): Delete.
12267         (*stmsi8): Delete.
12268         (*stmsi7): Delete.
12269         (*stmsi6): Delete.
12270         (*stmsi5): Delete.
12271         (*stmsi4): Delete.
12272         (*stmsi3): Delete.
12273         (movmemsi_8reg): Delete.
12274         (corresponding unnamed define_insn): Delete.
12275         (movmemsi_6reg): Delete.
12276         (corresponding unnamed define_insn): Delete.
12277         (movmemsi_4reg): Delete.
12278         (corresponding unnamed define_insn): Delete.
12279         (movmemsi_2reg): Delete.
12280         (corresponding unnamed define_insn): Delete.
12281         (movmemsi_1reg): Delete.
12282         (corresponding unnamed define_insn): Delete.
12283         * config/rs6000/rs6000.opt (mno-string): New.
12284         (mstring): Replace by deprecation warning stub.
12285         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12286
12287 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
12288
12289         * regrename.c (regrename_do_replace): If replacing the same
12290         reg multiple times, try to reuse last created gen_raw_REG.
12291
12292         PR debug/81155
12293         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12294         main to workaround a bug in GDB.
12295
12296 2018-01-12  Tom de Vries  <tom@codesourcery.com>
12297
12298         PR target/83737
12299         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12300
12301 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
12302
12303         PR rtl-optimization/80481
12304         * ira-color.c (get_cap_member): New function.
12305         (allocnos_conflict_by_live_ranges_p): Use it.
12306         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12307         (setup_slot_coalesced_allocno_live_ranges): Ditto.
12308
12309 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
12310
12311         PR target/83628
12312         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12313         (*saddl_se_1): Ditto.
12314         (*ssubsi_1): Ditto.
12315         (*ssubl_se_1): Ditto.
12316
12317 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
12318
12319         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12320         rather than wi::to_widest for DR_INITs.
12321         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12322         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12323         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12324         INTEGER_CSTs.
12325         (vect_analyze_group_access_1): Note that here.
12326
12327 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
12328
12329         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12330         polynomial type sizes.
12331
12332 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
12333
12334         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12335         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12336         (gimple_add_tmp_var): Likewise.
12337
12338 2018-01-12  Martin Liska  <mliska@suse.cz>
12339
12340         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12341         (gimple_alloc_sizes): Likewise.
12342         (dump_gimple_statistics): Use PRIu64 in printf format.
12343         * gimple.h: Change uint64_t to int.
12344
12345 2018-01-12  Martin Liska  <mliska@suse.cz>
12346
12347         * tree-core.h: Use uint64_t instead of int.
12348         * tree.c (tree_node_counts): Likewise.
12349         (tree_node_sizes): Likewise.
12350         (dump_tree_statistics): Use PRIu64 in printf format.
12351
12352 2018-01-12  Martin Liska  <mliska@suse.cz>
12353
12354         * Makefile.in: As qsort_chk is implemented in vec.c, add
12355         vec.o to linkage of gencfn-macros.
12356         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12357         passing the info to record_node_allocation_statistics.
12358         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12359         and pass the info.
12360         * ggc-common.c (struct ggc_usage): Add operator== and use
12361         it in operator< and compare function.
12362         * mem-stats.h (struct mem_usage): Likewise.
12363         * vec.c (struct vec_usage): Remove operator< and compare
12364         function. Can be simply inherited.
12365
12366 2018-01-12  Martin Jambor  <mjambor@suse.cz>
12367
12368         PR target/81616
12369         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12370         * tree-ssa-math-opts.c: Include domwalk.h.
12371         (convert_mult_to_fma_1): New function.
12372         (fma_transformation_info): New type.
12373         (fma_deferring_state): Likewise.
12374         (cancel_fma_deferring): New function.
12375         (result_of_phi): Likewise.
12376         (last_fma_candidate_feeds_initial_phi): Likewise.
12377         (convert_mult_to_fma): Added deferring logic, split actual
12378         transformation to convert_mult_to_fma_1.
12379         (math_opts_dom_walker): New type.
12380         (math_opts_dom_walker::after_dom_children): New method, body moved
12381         here from pass_optimize_widening_mul::execute, added deferring logic
12382         bits.
12383         (pass_optimize_widening_mul::execute): Moved most of code to
12384         math_opts_dom_walker::after_dom_children.
12385         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12386         * config/i386/i386.c (ix86_option_override_internal): Added
12387         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12388
12389 2018-01-12  Richard Biener  <rguenther@suse.de>
12390
12391         PR debug/83157
12392         * dwarf2out.c (gen_variable_die): Do not reset old_die for
12393         inline instance vars.
12394
12395 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
12396
12397         PR target/81819
12398         * config/rx/rx.c (rx_is_restricted_memory_address):
12399         Handle SUBREG case.
12400
12401 2018-01-12  Richard Biener  <rguenther@suse.de>
12402
12403         PR tree-optimization/80846
12404         * target.def (split_reduction): New target hook.
12405         * targhooks.c (default_split_reduction): New function.
12406         * targhooks.h (default_split_reduction): Declare.
12407         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12408         target requests first reduce vectors by combining low and high
12409         parts.
12410         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12411         (get_vectype_for_scalar_type_and_size): Export.
12412         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12413         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12414         * doc/tm.texi: Regenerate.
12415         * config/i386/i386.c (ix86_split_reduction): Implement
12416         TARGET_VECTORIZE_SPLIT_REDUCTION.
12417
12418 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
12419
12420         PR target/83368
12421         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12422         in PIC mode except for TARGET_VXWORKS_RTP.
12423         * config/sparc/sparc.c: Include cfgrtl.h.
12424         (TARGET_INIT_PIC_REG): Define.
12425         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12426         (sparc_pic_register_p): New predicate.
12427         (sparc_legitimate_address_p): Use it.
12428         (sparc_legitimize_pic_address): Likewise.
12429         (sparc_delegitimize_address): Likewise.
12430         (sparc_mode_dependent_address_p): Likewise.
12431         (gen_load_pcrel_sym): Remove 4th parameter.
12432         (load_got_register): Adjust call to above.  Remove obsolete stuff.
12433         (sparc_expand_prologue): Do not call load_got_register here.
12434         (sparc_flat_expand_prologue): Likewise.
12435         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12436         (sparc_use_pseudo_pic_reg): New function.
12437         (sparc_init_pic_reg): Likewise.
12438         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12439         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12440
12441 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
12442
12443         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12444         Add item for branch_cost.
12445
12446 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
12447
12448         PR rtl-optimization/83565
12449         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12450         not extend the result to a larger mode for rotate operations.
12451         (num_sign_bit_copies1): Likewise.
12452
12453 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12454
12455         PR target/40411
12456         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12457         -symbolic.
12458         Use values-Xc.o for -pedantic.
12459         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12460
12461 2018-01-12  Martin Liska  <mliska@suse.cz>
12462
12463         PR ipa/83054
12464         * ipa-devirt.c (final_warning_record::grow_type_warnings):
12465         New function.
12466         (possible_polymorphic_call_targets): Use it.
12467         (ipa_devirt): Likewise.
12468
12469 2018-01-12  Martin Liska  <mliska@suse.cz>
12470
12471         * profile-count.h (enum profile_quality): Use 0 as invalid
12472         enum value of profile_quality.
12473
12474 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
12475
12476         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12477         -mext-string options.
12478
12479 2018-01-12  Richard Biener  <rguenther@suse.de>
12480
12481         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12482         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12483         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12484         Likewise.
12485         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12486
12487 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
12488
12489         * configure.ac (--with-long-double-format): Add support for the
12490         configuration option to change the default long double format on
12491         PowerPC systems.
12492         * config.gcc (powerpc*-linux*-*): Likewise.
12493         * configure: Regenerate.
12494         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12495         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12496         used without modification.
12497
12498 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
12499
12500         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12501         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12502         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12503         MISC_BUILTIN_SPEC_BARRIER.
12504         (rs6000_init_builtins): Likewise.
12505         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12506         enum value.
12507         (speculation_barrier): New define_insn.
12508         * doc/extend.texi: Document __builtin_speculation_barrier.
12509
12510 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
12511
12512         PR target/83203
12513         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12514         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12515         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12516         iterators.
12517         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
12518         integral modes instead of "ss" and "sd".
12519         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12520         vectors with 32-bit and 64-bit elements.
12521         (vecdupssescalarmodesuffix): New mode attribute.
12522         (vec_dup<mode>): Use it.
12523
12524 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
12525
12526         PR target/83330
12527         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12528         frame if argument is passed on stack.
12529
12530 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
12531
12532         PR target/82682
12533         * ree.c (combine_reaching_defs): Optimize also
12534         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12535         reg2=any_extend(exp); reg1=reg2;, formatting fix.
12536
12537 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
12538
12539         PR middle-end/83189
12540         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12541
12542 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
12543
12544         PR middle-end/83718
12545         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12546         after they are computed.
12547
12548 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
12549
12550         PR tree-optimization/83695
12551         * gimple-loop-linterchange.cc
12552         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12553         reset cached scev information after interchange.
12554         (pass_linterchange::execute): Remove call to scev_reset_htab.
12555
12556 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12557
12558         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12559         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12560         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12561         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12562         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12563         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12564         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12565         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12566         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12567         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12568         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12569         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12570         (V_lane_reg): Likewise.
12571         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12572         New define_expand.
12573         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12574         (vfmal_lane_low<mode>_intrinsic,
12575         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12576         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12577         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12578         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12579         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12580         vfmsl_lane_high<mode>_intrinsic): New define_insns.
12581
12582 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12583
12584         * config/arm/arm-cpus.in (fp16fml): New feature.
12585         (ALL_SIMD): Add fp16fml.
12586         (armv8.2-a): Add fp16fml as an option.
12587         (armv8.3-a): Likewise.
12588         (armv8.4-a): Add fp16fml as part of fp16.
12589         * config/arm/arm.h (TARGET_FP16FML): Define.
12590         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12591         when appropriate.
12592         * config/arm/arm-modes.def (V2HF): Define.
12593         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12594         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12595         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12596         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12597         vfmsl_low, vfmsl_high): New set of builtins.
12598         * config/arm/iterators.md (PLUSMINUS): New code iterator.
12599         (vfml_op): New code attribute.
12600         (VFMLHALVES): New int iterator.
12601         (VFML, VFMLSEL): New mode attributes.
12602         (V_reg): Define mapping for V2HF.
12603         (V_hi, V_lo): New mode attributes.
12604         (VF_constraint): Likewise.
12605         (vfml_half, vfml_half_selector): New int attributes.
12606         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12607         define_expand.
12608         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12609         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12610         New define_insn.
12611         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12612         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12613         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12614         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
12615         documentation.
12616         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12617         Document new effective target and option set.
12618
12619 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12620
12621         * config/arm/arm-cpus.in (armv8_4): New feature.
12622         (ARMv8_4a): New fgroup.
12623         (armv8.4-a): New arch.
12624         * config/arm/arm-tables.opt: Regenerate.
12625         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12626         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12627         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12628         Add matching rules for -march=armv8.4-a and extensions.
12629         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12630
12631 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
12632
12633         PR target/81821
12634         * config/rx/rx.md (BW): New mode attribute.
12635         (sync_lock_test_and_setsi): Add mode suffix to insn output.
12636
12637 2018-01-11  Richard Biener  <rguenther@suse.de>
12638
12639         PR tree-optimization/83435
12640         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12641         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12642         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12643
12644 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12645             Alan Hayward  <alan.hayward@arm.com>
12646             David Sherwood  <david.sherwood@arm.com>
12647
12648         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12649         field.
12650         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
12651         (aarch64_print_address_internal): Use it to check for a zero offset.
12652
12653 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12654             Alan Hayward  <alan.hayward@arm.com>
12655             David Sherwood  <david.sherwood@arm.com>
12656
12657         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12658         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12659         Return a poly_int64 rather than a HOST_WIDE_INT.
12660         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12661         rather than a HOST_WIDE_INT.
12662         * config/aarch64/aarch64.h (aarch64_frame): Protect with
12663         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
12664         hard_fp_offset, frame_size, initial_adjust, callee_offset and
12665         final_offset from HOST_WIDE_INT to poly_int64.
12666         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12667         to_constant when getting the number of units in an Advanced SIMD
12668         mode.
12669         (aarch64_builtin_vectorized_function): Check for a constant number
12670         of units.
12671         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12672         GET_MODE_SIZE.
12673         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12674         attribute instead of GET_MODE_NUNITS.
12675         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12676         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12677         GET_MODE_SIZE for fixed-size registers.
12678         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12679         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12680         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12681         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12682         (aarch64_print_operand, aarch64_print_address_internal)
12683         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12684         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12685         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12686         Handle polynomial GET_MODE_SIZE.
12687         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
12688         wider than SImode without modification.
12689         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12690         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12691         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12692         passing and returning SVE modes.
12693         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12694         rather than GEN_INT.
12695         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12696         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12697         (aarch64_allocate_and_probe_stack_space): Likewise.
12698         (aarch64_layout_frame): Cope with polynomial offsets.
12699         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12700         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
12701         polynomial offsets.
12702         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12703         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12704         poly_int64 rather than a HOST_WIDE_INT.
12705         (aarch64_get_separate_components, aarch64_process_components)
12706         (aarch64_expand_prologue, aarch64_expand_epilogue)
12707         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12708         (aarch64_anchor_offset): New function, split out from...
12709         (aarch64_legitimize_address): ...here.
12710         (aarch64_builtin_vectorization_cost): Handle polynomial
12711         TYPE_VECTOR_SUBPARTS.
12712         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12713         GET_MODE_NUNITS.
12714         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12715         number of elements from the PARALLEL rather than the mode.
12716         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12717         rather than GET_MODE_BITSIZE.
12718         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12719         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12720         (aarch64_expand_vec_perm_const_1): Handle polynomial
12721         d->perm.length () and d->perm elements.
12722         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
12723         Apply to_constant to d->perm elements.
12724         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12725         polynomial CONST_VECTOR_NUNITS.
12726         (aarch64_move_pointer): Take amount as a poly_int64 rather
12727         than an int.
12728         (aarch64_progress_pointer): Avoid temporary variable.
12729         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12730         the mode attribute instead of GET_MODE.
12731
12732 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12733             Alan Hayward  <alan.hayward@arm.com>
12734             David Sherwood  <david.sherwood@arm.com>
12735
12736         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12737         x exists before using it.
12738         (aarch64_add_constant_internal): Rename to...
12739         (aarch64_add_offset_1): ...this.  Replace regnum with separate
12740         src and dest rtxes.  Handle the case in which they're different,
12741         including when the offset is zero.  Replace scratchreg with an rtx.
12742         Use 2 additions if there is no spare register into which we can
12743         move a 16-bit constant.
12744         (aarch64_add_constant): Delete.
12745         (aarch64_add_offset): Replace reg with separate src and dest
12746         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
12747         Use aarch64_add_offset_1.
12748         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12749         an rtx rather than an int.  Take the delta as a poly_int64
12750         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
12751         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12752         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12753         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12754         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12755         and aarch64_add_sp.
12756         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12757         aarch64_add_constant.
12758
12759 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12760
12761         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12762         Use scalar_float_mode.
12763
12764 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12765
12766         * config/aarch64/aarch64-simd.md
12767         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12768         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12769         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12770         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12771         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12772         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12773         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12774         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12775         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12776         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12777
12778 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
12779
12780         PR target/83514
12781         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12782         targ_options->x_arm_arch_string is non NULL.
12783
12784 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
12785
12786         * config/aarch64/aarch64.h
12787         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
12788
12789 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
12790
12791         PR target/82096
12792         * expmed.c (emit_store_flag_force): Swap if const op0
12793         and change VOIDmode to mode of op0.
12794
12795 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
12796
12797         PR rtl-optimization/83761
12798         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12799         than bytes to mode_for_size.
12800
12801 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
12802
12803         PR middle-end/83189
12804         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12805         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12806         profile.
12807
12808 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
12809
12810         PR middle-end/83575
12811         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12812         when in layout mode.
12813         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12814         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12815         partition fixup.
12816
12817 2018-01-10  Michael Collison  <michael.collison@arm.com>
12818
12819         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12820         * config/aarch64/aarch64-option-extension.def: Add
12821         AARCH64_OPT_EXTENSION of 'fp16fml'.
12822         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12823         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12824         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12825         * config/aarch64/constraints.md (Ui7): New constraint.
12826         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12827         (VFMLA_SEL_W): Ditto.
12828         (f16quad): Ditto.
12829         (f16mac1): Ditto.
12830         (VFMLA16_LOW): New int iterator.
12831         (VFMLA16_HIGH): Ditto.
12832         (UNSPEC_FMLAL): New unspec.
12833         (UNSPEC_FMLSL): Ditto.
12834         (UNSPEC_FMLAL2): Ditto.
12835         (UNSPEC_FMLSL2): Ditto.
12836         (f16mac): New code attribute.
12837         * config/aarch64/aarch64-simd-builtins.def
12838         (aarch64_fmlal_lowv2sf): Ditto.
12839         (aarch64_fmlsl_lowv2sf): Ditto.
12840         (aarch64_fmlalq_lowv4sf): Ditto.
12841         (aarch64_fmlslq_lowv4sf): Ditto.
12842         (aarch64_fmlal_highv2sf): Ditto.
12843         (aarch64_fmlsl_highv2sf): Ditto.
12844         (aarch64_fmlalq_highv4sf): Ditto.
12845         (aarch64_fmlslq_highv4sf): Ditto.
12846         (aarch64_fmlal_lane_lowv2sf): Ditto.
12847         (aarch64_fmlsl_lane_lowv2sf): Ditto.
12848         (aarch64_fmlal_laneq_lowv2sf): Ditto.
12849         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12850         (aarch64_fmlalq_lane_lowv4sf): Ditto.
12851         (aarch64_fmlsl_lane_lowv4sf): Ditto.
12852         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12853         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12854         (aarch64_fmlal_lane_highv2sf): Ditto.
12855         (aarch64_fmlsl_lane_highv2sf): Ditto.
12856         (aarch64_fmlal_laneq_highv2sf): Ditto.
12857         (aarch64_fmlsl_laneq_highv2sf): Ditto.
12858         (aarch64_fmlalq_lane_highv4sf): Ditto.
12859         (aarch64_fmlsl_lane_highv4sf): Ditto.
12860         (aarch64_fmlalq_laneq_highv4sf): Ditto.
12861         (aarch64_fmlsl_laneq_highv4sf): Ditto.
12862         * config/aarch64/aarch64-simd.md:
12863         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12864         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12865         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12866         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12867         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12868         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12869         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12870         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12871         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12872         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12873         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12874         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12875         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12876         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12877         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12878         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12879         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12880         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12881         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12882         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12883         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12884         (vfmlsl_low_u32): Ditto.
12885         (vfmlalq_low_u32): Ditto.
12886         (vfmlslq_low_u32): Ditto.
12887         (vfmlal_high_u32): Ditto.
12888         (vfmlsl_high_u32): Ditto.
12889         (vfmlalq_high_u32): Ditto.
12890         (vfmlslq_high_u32): Ditto.
12891         (vfmlal_lane_low_u32): Ditto.
12892         (vfmlsl_lane_low_u32): Ditto.
12893         (vfmlal_laneq_low_u32): Ditto.
12894         (vfmlsl_laneq_low_u32): Ditto.
12895         (vfmlalq_lane_low_u32): Ditto.
12896         (vfmlslq_lane_low_u32): Ditto.
12897         (vfmlalq_laneq_low_u32): Ditto.
12898         (vfmlslq_laneq_low_u32): Ditto.
12899         (vfmlal_lane_high_u32): Ditto.
12900         (vfmlsl_lane_high_u32): Ditto.
12901         (vfmlal_laneq_high_u32): Ditto.
12902         (vfmlsl_laneq_high_u32): Ditto.
12903         (vfmlalq_lane_high_u32): Ditto.
12904         (vfmlslq_lane_high_u32): Ditto.
12905         (vfmlalq_laneq_high_u32): Ditto.
12906         (vfmlslq_laneq_high_u32): Ditto.
12907         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12908         (AARCH64_FL_FOR_ARCH8_4): New.
12909         (AARCH64_ISA_F16FML): New ISA flag.
12910         (TARGET_F16FML): New feature flag for fp16fml.
12911         (doc/invoke.texi): Document new fp16fml option.
12912
12913 2018-01-10  Michael Collison  <michael.collison@arm.com>
12914
12915         * config/aarch64/aarch64-builtins.c:
12916         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12917         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12918         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12919         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12920         (AARCH64_ISA_SHA3): New ISA flag.
12921         (TARGET_SHA3): New feature flag for sha3.
12922         * config/aarch64/iterators.md (sha512_op): New int attribute.
12923         (CRYPTO_SHA512): New int iterator.
12924         (UNSPEC_SHA512H): New unspec.
12925         (UNSPEC_SHA512H2): Ditto.
12926         (UNSPEC_SHA512SU0): Ditto.
12927         (UNSPEC_SHA512SU1): Ditto.
12928         * config/aarch64/aarch64-simd-builtins.def
12929         (aarch64_crypto_sha512hqv2di): New builtin.
12930         (aarch64_crypto_sha512h2qv2di): Ditto.
12931         (aarch64_crypto_sha512su0qv2di): Ditto.
12932         (aarch64_crypto_sha512su1qv2di): Ditto.
12933         (aarch64_eor3qv8hi): Ditto.
12934         (aarch64_rax1qv2di): Ditto.
12935         (aarch64_xarqv2di): Ditto.
12936         (aarch64_bcaxqv8hi): Ditto.
12937         * config/aarch64/aarch64-simd.md:
12938         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12939         (aarch64_crypto_sha512su0qv2di): Ditto.
12940         (aarch64_crypto_sha512su1qv2di): Ditto.
12941         (aarch64_eor3qv8hi): Ditto.
12942         (aarch64_rax1qv2di): Ditto.
12943         (aarch64_xarqv2di): Ditto.
12944         (aarch64_bcaxqv8hi): Ditto.
12945         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12946         (vsha512h2q_u64): Ditto.
12947         (vsha512su0q_u64): Ditto.
12948         (vsha512su1q_u64): Ditto.
12949         (veor3q_u16): Ditto.
12950         (vrax1q_u64): Ditto.
12951         (vxarq_u64): Ditto.
12952         (vbcaxq_u16): Ditto.
12953         * config/arm/types.md (crypto_sha512): New type attribute.
12954         (crypto_sha3): Ditto.
12955         (doc/invoke.texi): Document new sha3 option.
12956
12957 2018-01-10  Michael Collison  <michael.collison@arm.com>
12958
12959         * config/aarch64/aarch64-builtins.c:
12960         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12961         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12962         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12963         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12964         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12965         (AARCH64_ISA_SM4): New ISA flag.
12966         (TARGET_SM4): New feature flag for sm4.
12967         * config/aarch64/aarch64-simd-builtins.def
12968         (aarch64_sm3ss1qv4si): Ditto.
12969         (aarch64_sm3tt1aq4si): Ditto.
12970         (aarch64_sm3tt1bq4si): Ditto.
12971         (aarch64_sm3tt2aq4si): Ditto.
12972         (aarch64_sm3tt2bq4si): Ditto.
12973         (aarch64_sm3partw1qv4si): Ditto.
12974         (aarch64_sm3partw2qv4si): Ditto.
12975         (aarch64_sm4eqv4si): Ditto.
12976         (aarch64_sm4ekeyqv4si): Ditto.
12977         * config/aarch64/aarch64-simd.md:
12978         (aarch64_sm3ss1qv4si): Ditto.
12979         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12980         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12981         (aarch64_sm4eqv4si): Ditto.
12982         (aarch64_sm4ekeyqv4si): Ditto.
12983         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12984         (sm3part_op): Ditto.
12985         (CRYPTO_SM3TT): Ditto.
12986         (CRYPTO_SM3PART): Ditto.
12987         (UNSPEC_SM3SS1): New unspec.
12988         (UNSPEC_SM3TT1A): Ditto.
12989         (UNSPEC_SM3TT1B): Ditto.
12990         (UNSPEC_SM3TT2A): Ditto.
12991         (UNSPEC_SM3TT2B): Ditto.
12992         (UNSPEC_SM3PARTW1): Ditto.
12993         (UNSPEC_SM3PARTW2): Ditto.
12994         (UNSPEC_SM4E): Ditto.
12995         (UNSPEC_SM4EKEY): Ditto.
12996         * config/aarch64/constraints.md (Ui2): New constraint.
12997         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12998         * config/arm/types.md (crypto_sm3): New type attribute.
12999         (crypto_sm4): Ditto.
13000         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
13001         (vsm3tt1aq_u32): Ditto.
13002         (vsm3tt1bq_u32): Ditto.
13003         (vsm3tt2aq_u32): Ditto.
13004         (vsm3tt2bq_u32): Ditto.
13005         (vsm3partw1q_u32): Ditto.
13006         (vsm3partw2q_u32): Ditto.
13007         (vsm4eq_u32): Ditto.
13008         (vsm4ekeyq_u32): Ditto.
13009         (doc/invoke.texi): Document new sm4 option.
13010
13011 2018-01-10  Michael Collison  <michael.collison@arm.com>
13012
13013         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
13014         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
13015         (AARCH64_FL_FOR_ARCH8_4): New.
13016         (AARCH64_FL_V8_4): New flag.
13017         (doc/invoke.texi): Document new armv8.4-a option.
13018
13019 2018-01-10  Michael Collison  <michael.collison@arm.com>
13020
13021         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13022         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
13023         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
13024         * config/aarch64/aarch64-option-extension.def: Add
13025         AARCH64_OPT_EXTENSION of 'sha2'.
13026         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
13027         (crypto): Disable sha2 and aes if crypto disabled.
13028         (crypto): Enable aes and sha2 if enabled.
13029         (simd): Disable sha2 and aes if simd disabled.
13030         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
13031         New flags.
13032         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
13033         (TARGET_SHA2): New feature flag for sha2.
13034         (TARGET_AES): New feature flag for aes.
13035         * config/aarch64/aarch64-simd.md:
13036         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
13037         conditional on TARGET_AES.
13038         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
13039         (aarch64_crypto_sha1hsi): Make pattern conditional
13040         on TARGET_SHA2.
13041         (aarch64_crypto_sha1hv4si): Ditto.
13042         (aarch64_be_crypto_sha1hv4si): Ditto.
13043         (aarch64_crypto_sha1su1v4si): Ditto.
13044         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
13045         (aarch64_crypto_sha1su0v4si): Ditto.
13046         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
13047         (aarch64_crypto_sha256su0v4si): Ditto.
13048         (aarch64_crypto_sha256su1v4si): Ditto.
13049         (doc/invoke.texi): Document new aes and sha2 options.
13050
13051 2018-01-10  Martin Sebor  <msebor@redhat.com>
13052
13053         PR tree-optimization/83781
13054         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
13055         as string arrays.
13056
13057 2018-01-11  Martin Sebor  <msebor@gmail.com>
13058             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
13059
13060         PR tree-optimization/83501
13061         PR tree-optimization/81703
13062
13063         * tree-ssa-strlen.c (get_string_cst): Rename...
13064         (get_string_len): ...to this.  Handle global constants.
13065         (handle_char_store): Adjust.
13066
13067 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
13068             Jim Wilson  <jimw@sifive.com>
13069
13070         * config/riscv/riscv-protos.h (riscv_output_return): New.
13071         * config/riscv/riscv.c (struct machine_function): New naked_p field.
13072         (riscv_attribute_table, riscv_output_return),
13073         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
13074         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
13075         (riscv_compute_frame_info): Only compute frame->mask if not a naked
13076         function.
13077         (riscv_expand_prologue): Add early return for naked function.
13078         (riscv_expand_epilogue): Likewise.
13079         (riscv_function_ok_for_sibcall): Return false for naked function.
13080         (riscv_set_current_function): New.
13081         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
13082         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
13083         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
13084         * doc/extend.texi (RISC-V Function Attributes): New.
13085
13086 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
13087
13088         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
13089         check for 128-bit long double before checking TCmode.
13090         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
13091         128-bit long doubles before checking TFmode or TCmode.
13092         (FLOAT128_IBM_P): Likewise.
13093
13094 2018-01-10  Martin Sebor  <msebor@redhat.com>
13095
13096         PR tree-optimization/83671
13097         * builtins.c (c_strlen): Unconditionally return zero for the empty
13098         string.
13099         Use -Warray-bounds for warnings.
13100         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
13101         for non-constant array indices with COMPONENT_REF, arrays of
13102         arrays, and pointers to arrays.
13103         (gimple_fold_builtin_strlen): Determine and set length range for
13104         non-constant character arrays.
13105
13106 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
13107
13108         PR middle-end/81897
13109         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
13110         empty blocks.
13111
13112 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
13113
13114         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
13115
13116 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
13117
13118         PR target/83399
13119         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
13120         VECTOR_MEM_ALTIVEC_OR_VSX_P.
13121         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
13122         indexed_or_indirect_operand predicate.
13123         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
13124         (*vsx_le_perm_load_v8hi): Likewise.
13125         (*vsx_le_perm_load_v16qi): Likewise.
13126         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
13127         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
13128         (*vsx_le_perm_store_v8hi): Likewise.
13129         (*vsx_le_perm_store_v16qi): Likewise.
13130         (eight unnamed splitters): Likewise.
13131
13132 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
13133
13134         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
13135         * config/rs6000/emmintrin.h: Likewise.
13136         * config/rs6000/mmintrin.h: Likewise.
13137         * config/rs6000/xmmintrin.h: Likewise.
13138
13139 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
13140
13141         PR c++/43486
13142         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
13143         "public_flag".
13144         * tree.c (tree_nop_conversion): Return true for location wrapper
13145         nodes.
13146         (maybe_wrap_with_location): New function.
13147         (selftest::check_strip_nops): New function.
13148         (selftest::test_location_wrappers): New function.
13149         (selftest::tree_c_tests): Call it.
13150         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
13151         (maybe_wrap_with_location): New decl.
13152         (EXPR_LOCATION_WRAPPER_P): New macro.
13153         (location_wrapper_p): New inline function.
13154         (tree_strip_any_location_wrapper): New inline function.
13155
13156 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
13157
13158         PR target/83735
13159         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
13160         stack_realign_offset for the largest alignment of stack slot
13161         actually used.
13162         (ix86_find_max_used_stack_alignment): New function.
13163         (ix86_finalize_stack_frame_flags): Use it.  Set
13164         max_used_stack_alignment if we don't realign stack.
13165         * config/i386/i386.h (machine_function): Add
13166         max_used_stack_alignment.
13167
13168 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
13169
13170         * config/arm/arm.opt (-mbranch-cost): New option.
13171         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
13172         account.
13173
13174 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
13175
13176         PR target/83629
13177         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
13178         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
13179
13180 2018-01-10  Richard Biener  <rguenther@suse.de>
13181
13182         PR debug/83765
13183         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
13184         early out so it also covers the case where we have a non-NULL
13185         origin.
13186
13187 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
13188
13189         PR tree-optimization/83753
13190         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
13191         for non-strided grouped accesses if the number of elements is 1.
13192
13193 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
13194
13195         PR target/81616
13196         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
13197         * i386.h (TARGET_USE_GATHER): Define.
13198         * x86-tune.def (X86_TUNE_USE_GATHER): New.
13199
13200 2018-01-10  Martin Liska  <mliska@suse.cz>
13201
13202         PR bootstrap/82831
13203         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
13204         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
13205         partitioning.
13206         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
13207         CLEANUP_NO_PARTITIONING is not set.
13208
13209 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
13210
13211         * doc/rtl.texi: Remove documentation of (const ...) wrappers
13212         for vectors, as a partial revert of r254296.
13213         * rtl.h (const_vec_p): Delete.
13214         (const_vec_duplicate_p): Don't test for vector CONSTs.
13215         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
13216         * expmed.c (make_tree): Likewise.
13217
13218         Revert:
13219         * common.md (E, F): Use CONSTANT_P instead of checking for
13220         CONST_VECTOR.
13221         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
13222         checking for CONST_VECTOR.
13223
13224 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
13225
13226         PR middle-end/83575
13227         * predict.c (force_edge_cold): Handle in more sane way edges
13228         with no prediction.
13229
13230 2018-01-09  Carl Love  <cel@us.ibm.com>
13231
13232         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13233         V4SI, V4SF types.
13234         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13235         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13236         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13237         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
13238         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13239         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
13240         * config/rs6000/rs6000-protos.h: Add extern defition for
13241         rs6000_generate_float2_double_code.
13242         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13243         function.
13244         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13245         (float2_v2df): Add define_expand.
13246
13247 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
13248
13249         PR target/83628
13250         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13251         op_mode in the force_to_mode call.
13252
13253 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
13254
13255         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13256         instead of checking each element individually.
13257         (aarch64_evpc_uzp): Likewise.
13258         (aarch64_evpc_zip): Likewise.
13259         (aarch64_evpc_ext): Likewise.
13260         (aarch64_evpc_rev): Likewise.
13261         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13262         instead of checking each element individually.  Return true without
13263         generating rtl if
13264         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13265         whether all selected elements come from the same input, instead of
13266         checking each element individually.  Remove calls to gen_rtx_REG,
13267         start_sequence and end_sequence and instead assert that no rtl is
13268         generated.
13269
13270 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
13271
13272         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13273         order of HIGH and CONST checks.
13274
13275 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
13276
13277         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13278         if the destination isn't an SSA_NAME.
13279
13280 2018-01-09  Richard Biener  <rguenther@suse.de>
13281
13282         PR tree-optimization/83668
13283         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13284         move prologue...
13285         (canonicalize_loop_form): ... here, renamed from ...
13286         (canonicalize_loop_closed_ssa_form): ... this and amended to
13287         swap successor edges for loop exit blocks to make us use
13288         the RPO order we need for initial schedule generation.
13289
13290 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
13291
13292         PR tree-optimization/64811
13293         * match.pd: When optimizing comparisons with Inf, avoid
13294         introducing or losing exceptions from comparisons with NaN.
13295
13296 2018-01-09  Martin Liska  <mliska@suse.cz>
13297
13298         PR sanitizer/82517
13299         * asan.c (shadow_mem_size): Add gcc_assert.
13300
13301 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
13302
13303         Don't save registers in main().
13304
13305         PR target/83738
13306         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13307         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13308         * config/avr/avr.c (avr_set_current_function): Don't error if
13309         naked, OS_task or OS_main are specified at the same time.
13310         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13311         OS_main.
13312         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13313         attribute.
13314         * common/config/avr/avr-common.c (avr_option_optimization_table):
13315         Switch on -mmain-is-OS_task for optimizing compilations.
13316
13317 2018-01-09  Richard Biener  <rguenther@suse.de>
13318
13319         PR tree-optimization/83572
13320         * graphite.c: Include cfganal.h.
13321         (graphite_transform_loops): Connect infinite loops to exit
13322         and remove fake edges at the end.
13323
13324 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
13325
13326         * ipa-inline.c (edge_badness): Revert accidental checkin.
13327
13328 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
13329
13330         PR ipa/80763
13331         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13332         symbols; not inline clones.
13333
13334 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
13335
13336         PR target/83507
13337         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13338         hard registers.  Formatting fixes.
13339
13340         PR preprocessor/83722
13341         * gcc.c (try_generate_repro): Pass
13342         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13343         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13344         do_report_bug.
13345
13346 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
13347             Kito Cheng  <kito.cheng@gmail.com>
13348
13349         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13350         (riscv_leaf_function_p): Delete.
13351         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13352
13353 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
13354
13355         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13356         function.
13357         (do_ifelse): New function.
13358         (do_isel): New function.
13359         (do_sub3): New function.
13360         (do_add3): New function.
13361         (do_load_mask_compare): New function.
13362         (do_overlap_load_compare): New function.
13363         (expand_compare_loop): New function.
13364         (expand_block_compare): Call expand_compare_loop() when appropriate.
13365         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13366         option description.
13367         (-mblock-compare-inline-loop-limit): New option.
13368
13369 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13370
13371         PR target/83677
13372         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13373         Reverse order of second and third operands in first alternative.
13374         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13375         of first and second elements in UNSPEC_VPERMR vector.
13376         (altivec_expand_vec_perm_le): Likewise.
13377
13378 2018-01-08  Jeff Law  <law@redhat.com>
13379
13380         PR rtl-optimizatin/81308
13381         * tree-switch-conversion.c (cfg_altered): New file scoped static.
13382         (process_switch): If group_case_labels makes a change, then set
13383         cfg_altered.
13384         (pass_convert_switch::execute): If a switch is converted, then
13385         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
13386
13387         PR rtl-optimization/81308
13388         * recog.c (split_all_insns): Conditionally cleanup the CFG after
13389         splitting insns.
13390
13391 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
13392
13393         PR target/83663 - Revert r255946
13394         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13395         generation for cases where splatting a value is not useful.
13396         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13397         across a vec_duplicate and a paradoxical subreg forming a vector
13398         mode to a vec_concat.
13399
13400 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13401
13402         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13403         -march=armv8.3-a variants.
13404         * config/arm/t-multilib: Likewise.
13405         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
13406
13407 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
13408
13409         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13410         to generate rtl.
13411         (cceq_ior_compare_complement): Give it a name so I can use it, and
13412         change boolean_or_operator predicate to boolean_operator so it can
13413         be used to generate a crand.
13414         (eqne): New code iterator.
13415         (bd/bd_neg): New code_attrs.
13416         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13417         a single define_insn.
13418         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13419         decrement (bdnzt/bdnzf/bdzt/bdzf).
13420         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13421         with the new names of the branch decrement patterns, and added the
13422         names of the branch decrement conditional patterns.
13423
13424 2018-01-08  Richard Biener  <rguenther@suse.de>
13425
13426         PR tree-optimization/83563
13427         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13428         cache.
13429
13430 2018-01-08  Richard Biener  <rguenther@suse.de>
13431
13432         PR middle-end/83713
13433         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13434
13435 2018-01-08  Richard Biener  <rguenther@suse.de>
13436
13437         PR tree-optimization/83685
13438         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13439         references to abnormals.
13440
13441 2018-01-08  Richard Biener  <rguenther@suse.de>
13442
13443         PR lto/83719
13444         * dwarf2out.c (output_indirect_strings): Handle empty
13445         skeleton_debug_str_hash.
13446         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13447
13448 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
13449
13450         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13451         (emit_store_direct): Likewise.
13452         (arc_trampoline_adjust_address): Likewise.
13453         (arc_asm_trampoline_template): New function.
13454         (arc_initialize_trampoline): Use asm_trampoline_template.
13455         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13456         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13457         * config/arc/arc.md (flush_icache): Delete pattern.
13458
13459 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
13460
13461         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13462         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13463         munaligned-access.
13464
13465 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
13466
13467         PR target/83681
13468         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13469         by not USED_FOR_TARGET.
13470         (make_pass_resolve_sw_modes): Likewise.
13471
13472 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
13473
13474         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13475         USED_FOR_TARGET.
13476
13477 2018-01-08  Richard Biener  <rguenther@suse.de>
13478
13479         PR middle-end/83580
13480         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13481
13482 2018-01-08  Richard Biener  <rguenther@suse.de>
13483
13484         PR middle-end/83517
13485         * match.pd ((t * 2) / 2) -> t): Add missing :c.
13486
13487 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
13488
13489         PR middle-end/81897
13490         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13491         basic blocks with a small number of successors.
13492         (convert_control_dep_chain_into_preds): Improve handling of
13493         forwarder blocks.
13494         (dump_predicates): Split apart into...
13495         (dump_pred_chain): ...here...
13496         (dump_pred_info): ...and here.
13497         (can_one_predicate_be_invalidated_p): Add debugging printfs.
13498         (can_chain_union_be_invalidated_p): Improve check for invalidation
13499         of paths.
13500         (uninit_uses_cannot_happen): Avoid unnecessary if
13501         convert_control_dep_chain_into_preds yielded nothing.
13502
13503 2018-01-06  Martin Sebor  <msebor@redhat.com>
13504
13505         PR tree-optimization/83640
13506         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13507         subtracting negative offset from size.
13508         (builtin_access::overlap): Adjust offset bounds of the access to fall
13509         within the size of the object if possible.
13510
13511 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
13512
13513         PR rtl-optimization/83699
13514         * expmed.c (extract_bit_field_1): Restrict the vector usage of
13515         extract_bit_field_as_subreg to cases in which the extracted
13516         value is also a vector.
13517
13518         * lra-constraints.c (process_alt_operands): Test for the equivalence
13519         substitutions when detecting a possible reload cycle.
13520
13521 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
13522
13523         PR debug/83480
13524         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13525         by default if flag_selective_schedling{,2}.  Formatting fixes.
13526
13527         PR rtl-optimization/83682
13528         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13529         if it has non-VECTOR_MODE element mode.
13530         (vec_duplicate_p): Likewise.
13531
13532         PR middle-end/83694
13533         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13534         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13535
13536 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
13537
13538         PR target/83604
13539         * config/i386/i386-builtin.def
13540         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13541         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13542         Require also OPTION_MASK_ISA_AVX512F in addition to
13543         OPTION_MASK_ISA_GFNI.
13544         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13545         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13546         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13547         to OPTION_MASK_ISA_GFNI.
13548         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13549         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13550         OPTION_MASK_ISA_AVX512BW.
13551         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13552         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13553         addition to OPTION_MASK_ISA_GFNI.
13554         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13555         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13556         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13557         to OPTION_MASK_ISA_GFNI.
13558         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13559         a requirement for all ISAs rather than any of them with a few
13560         exceptions.
13561         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13562         processing.
13563         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13564         bitmasks to be enabled with 3 exceptions, instead of requiring any
13565         enabled ISA with lots of exceptions.
13566         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13567         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13568         Change avx512bw in isa attribute to avx512f.
13569         * config/i386/sgxintrin.h: Add license boilerplate.
13570         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
13571         to __AVX512F__ and __AVX512VL to __AVX512VL__.
13572         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13573         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13574         defined.
13575         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13576         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13577         temporarily sse2 rather than sse if not enabled already.
13578
13579         PR target/83604
13580         * config/i386/sse.md (VI248_VLBW): Rename to ...
13581         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
13582         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13583         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13584         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13585         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13586         mode iterator instead of VI248_VLBW.
13587
13588 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
13589
13590         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13591         (record_modified): Skip clobbers; add debug output.
13592         (param_change_prob): Use sreal frequencies.
13593
13594 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
13595
13596         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13597         punt for user-aligned variables.
13598
13599 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
13600
13601         * tree-chrec.c (chrec_contains_symbols): Return true for
13602         POLY_INT_CST.
13603
13604 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
13605
13606         PR target/82439
13607         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13608         of (x|y) == x for BICS pattern.
13609
13610 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
13611
13612         PR tree-optimization/83605
13613         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13614         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13615         can throw.
13616
13617 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
13618
13619         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13620         * config/epiphany/rtems.h: New file.
13621
13622 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
13623             Uros Bizjak  <ubizjak@gmail.com>
13624
13625         PR target/83554
13626         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13627         QIreg_operand instead of register_operand predicate.
13628         * config/i386/i386.c (ix86_rop_should_change_byte_p,
13629         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13630         comments instead of -fmitigate[-_]rop.
13631
13632 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13633
13634         PR bootstrap/81926
13635         * cgraphunit.c (symbol_table::compile): Switch to text_section
13636         before calling assembly_start debug hook.
13637         * run-rtl-passes.c (run_rtl_passes): Likewise.
13638         Include output.h.
13639
13640 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13641
13642         * tree-vrp.c (extract_range_from_binary_expr_1): Check
13643         range_int_cst_p rather than !symbolic_range_p before calling
13644         extract_range_from_multiplicative_op_1.
13645
13646 2018-01-04  Jeff Law  <law@redhat.com>
13647
13648         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13649         redundant test in assertion.
13650
13651 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13652
13653         * doc/rtl.texi: Document machine_mode wrapper classes.
13654
13655 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13656
13657         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13658         using tree_to_uhwi.
13659
13660 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13661
13662         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13663         the VEC_PERM_EXPR fold to fail.
13664
13665 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
13666
13667         PR debug/83585
13668         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13669         to switched_sections.
13670
13671 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13672
13673         PR target/83680
13674         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13675         test for d.testing.
13676
13677 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
13678
13679         PR target/83387
13680         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13681         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13682
13683 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
13684
13685         PR debug/83666
13686         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13687         is BLKmode and bitpos not zero or mode change is needed.
13688
13689 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
13690
13691         PR target/83675
13692         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13693         TARGET_VIS2.
13694
13695 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
13696
13697         PR target/83628
13698         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13699         instead of MULT rtx.  Update all corresponding splitters.
13700         (*saddl_se): Ditto.
13701         (*ssub<modesuffix>): Ditto.
13702         (*ssubl_se): Ditto.
13703         (*cmp_sadd_di): Update split patterns.
13704         (*cmp_sadd_si): Ditto.
13705         (*cmp_sadd_sidi): Ditto.
13706         (*cmp_ssub_di): Ditto.
13707         (*cmp_ssub_si): Ditto.
13708         (*cmp_ssub_sidi): Ditto.
13709         * config/alpha/predicates.md (const23_operand): New predicate.
13710         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13711         Look for ASHIFT, not MULT inner operand.
13712         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13713
13714 2018-01-04  Martin Liska  <mliska@suse.cz>
13715
13716         PR gcov-profile/83669
13717         * gcov.c (output_intermediate_file): Add version to intermediate
13718         gcov file.
13719         * doc/gcov.texi: Document new field 'version' in intermediate
13720         file format. Fix location of '-k' option of gcov command.
13721
13722 2018-01-04  Martin Liska  <mliska@suse.cz>
13723
13724         PR ipa/82352
13725         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13726
13727 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
13728
13729         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13730
13731 2018-01-03  Martin Sebor  <msebor@redhat.com>
13732
13733         PR tree-optimization/83655
13734         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13735         checking calls with invalid arguments.
13736
13737 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13738
13739         * tree-vect-stmts.c (vect_get_store_rhs): New function.
13740         (vectorizable_mask_load_store): Delete.
13741         (vectorizable_call): Return false for masked loads and stores.
13742         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
13743         instead of gimple_assign_rhs1.
13744         (vectorizable_load): Handle IFN_MASK_LOAD.
13745         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13746
13747 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13748
13749         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13750         split out from..,
13751         (vectorizable_mask_load_store): ...here.
13752         (vectorizable_load): ...and here.
13753
13754 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13755
13756         * tree-vect-stmts.c (vect_build_all_ones_mask)
13757         (vect_build_zero_merge_argument): New functions, split out from...
13758         (vectorizable_load): ...here.
13759
13760 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13761
13762         * tree-vect-stmts.c (vect_check_store_rhs): New function,
13763         split out from...
13764         (vectorizable_mask_load_store): ...here.
13765         (vectorizable_store): ...and here.
13766
13767 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13768
13769         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13770         split out from...
13771         (vectorizable_mask_load_store): ...here.
13772
13773 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13774
13775         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13776         (vect_model_store_cost): Take a vec_load_store_type instead of a
13777         vect_def_type.
13778         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13779         (vect_model_store_cost): Take a vec_load_store_type instead of a
13780         vect_def_type.
13781         (vectorizable_mask_load_store): Update accordingly.
13782         (vectorizable_store): Likewise.
13783         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13784
13785 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13786
13787         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13788         IFN_MASK_LOAD calls here rather than...
13789         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13790
13791 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13792             Alan Hayward  <alan.hayward@arm.com>
13793             David Sherwood  <david.sherwood@arm.com>
13794
13795         * expmed.c (extract_bit_field_1): For vector extracts,
13796         fall back to extract_bit_field_as_subreg if vec_extract
13797         isn't available.
13798
13799 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13800             Alan Hayward  <alan.hayward@arm.com>
13801             David Sherwood  <david.sherwood@arm.com>
13802
13803         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13804         they are variable or constant sized.
13805         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13806         slots for constant-sized data.
13807
13808 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13809             Alan Hayward  <alan.hayward@arm.com>
13810             David Sherwood  <david.sherwood@arm.com>
13811
13812         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13813         handling COND_EXPRs with boolean comparisons, try to find a better
13814         basis for the mask type than the boolean itself.
13815
13816 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13817
13818         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13819         is calculated and how it can be overridden.
13820         * genmodes.c (max_bitsize_mode_any_mode): New variable.
13821         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13822         if defined.
13823         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13824         if nonzero.
13825
13826 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13827             Alan Hayward  <alan.hayward@arm.com>
13828             David Sherwood  <david.sherwood@arm.com>
13829
13830         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13831         Remove the mode argument.
13832         (aarch64_simd_valid_immediate): Remove the mode and inverse
13833         arguments.
13834         * config/aarch64/iterators.md (bitsize): New iterator.
13835         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13836         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13837         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13838         aarch64_simd_valid_immediate.
13839         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13840         (aarch64_reg_or_bic_imm): Likewise.
13841         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13842         with an insn_type enum and msl with a modifier_type enum.
13843         Replace element_width with a scalar_mode.  Change the shift
13844         to unsigned int.  Add constructors for scalar_float_mode and
13845         scalar_int_mode elements.
13846         (aarch64_vect_float_const_representable_p): Delete.
13847         (aarch64_can_const_movi_rtx_p)
13848         (aarch64_simd_scalar_immediate_valid_for_move)
13849         (aarch64_simd_make_constant): Update call to
13850         aarch64_simd_valid_immediate.
13851         (aarch64_advsimd_valid_immediate_hs): New function.
13852         (aarch64_advsimd_valid_immediate): Likewise.
13853         (aarch64_simd_valid_immediate): Remove mode and inverse
13854         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
13855         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13856         and aarch64_float_const_representable_p on the result.
13857         (aarch64_output_simd_mov_immediate): Remove mode argument.
13858         Update call to aarch64_simd_valid_immediate and use of
13859         simd_immediate_info.
13860         (aarch64_output_scalar_simd_mov_immediate): Update call
13861         accordingly.
13862
13863 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13864             Alan Hayward  <alan.hayward@arm.com>
13865             David Sherwood  <david.sherwood@arm.com>
13866
13867         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13868         (mode_nunits): Likewise CONST_MODE_NUNITS.
13869         * machmode.def (ADJUST_NUNITS): Document.
13870         * genmodes.c (mode_data::need_nunits_adj): New field.
13871         (blank_mode): Update accordingly.
13872         (adj_nunits): New variable.
13873         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13874         parameter.
13875         (emit_mode_size_inline): Set need_bytesize_adj for all modes
13876         listed in adj_nunits.
13877         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13878         listed in adj_nunits.  Don't emit case statements for such modes.
13879         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13880         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
13881         nothing if adj_nunits is nonnull.
13882         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13883         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13884         (emit_mode_fbit): Update use of print_maybe_const_decl.
13885         (emit_move_size): Likewise.  Treat the array as non-const
13886         if adj_nunits.
13887         (emit_mode_adjustments): Handle adj_nunits.
13888
13889 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13890
13891         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13892         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13893         (VECTOR_MODES): Use it.
13894         (make_vector_modes): Take the prefix as an argument.
13895
13896 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13897             Alan Hayward  <alan.hayward@arm.com>
13898             David Sherwood  <david.sherwood@arm.com>
13899
13900         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13901         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13902         for MODE_VECTOR_BOOL.
13903         * machmode.def (VECTOR_BOOL_MODE): Document.
13904         * genmodes.c (VECTOR_BOOL_MODE): New macro.
13905         (make_vector_bool_mode): New function.
13906         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13907         MODE_VECTOR_BOOL.
13908         * lto-streamer-in.c (lto_input_mode_table): Likewise.
13909         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13910         Likewise.
13911         * stor-layout.c (int_mode_for_mode): Likewise.
13912         * tree.c (build_vector_type_for_mode): Likewise.
13913         * varasm.c (output_constant_pool_2): Likewise.
13914         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13915         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
13916         for MODE_VECTOR_BOOL.
13917         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13918         of mode class checks.
13919         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13920         instead of a list of mode class checks.
13921         (expand_vector_scalar_condition): Likewise.
13922         (type_for_widest_vector_mode): Handle BImode as an inner mode.
13923
13924 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13925             Alan Hayward  <alan.hayward@arm.com>
13926             David Sherwood  <david.sherwood@arm.com>
13927
13928         * machmode.h (mode_size): Change from unsigned short to
13929         poly_uint16_pod.
13930         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13931         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13932         or if measurement_type is not polynomial.
13933         (fixed_size_mode::includes_p): Check for constant-sized modes.
13934         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13935         return a poly_uint16 rather than an unsigned short.
13936         (emit_mode_size): Change the type of mode_size from unsigned short
13937         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
13938         (emit_mode_adjustments): Cope with polynomial vector sizes.
13939         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13940         for GET_MODE_SIZE.
13941         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13942         for GET_MODE_SIZE.
13943         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13944         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13945         * caller-save.c (setup_save_areas): Likewise.
13946         (replace_reg_with_saved_mem): Likewise.
13947         * calls.c (emit_library_call_value_1): Likewise.
13948         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13949         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13950         (gen_lowpart_for_combine): Likewise.
13951         * convert.c (convert_to_integer_1): Likewise.
13952         * cse.c (equiv_constant, cse_insn): Likewise.
13953         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13954         (cselib_subst_to_values): Likewise.
13955         * dce.c (word_dce_process_block): Likewise.
13956         * df-problems.c (df_word_lr_mark_ref): Likewise.
13957         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13958         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13959         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13960         (rtl_for_decl_location): Likewise.
13961         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13962         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13963         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13964         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13965         (expand_expr_real_1): Likewise.
13966         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13967         (pad_below): Likewise.
13968         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13969         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13970         * ira.c (get_subreg_tracking_sizes): Likewise.
13971         * ira-build.c (ira_create_allocno_objects): Likewise.
13972         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13973         (ira_sort_regnos_for_alter_reg): Likewise.
13974         * ira-costs.c (record_operand_costs): Likewise.
13975         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13976         (resolve_simple_move): Likewise.
13977         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13978         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13979         (lra_constraints): Likewise.
13980         (CONST_POOL_OK_P): Reject variable-sized modes.
13981         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13982         (add_pseudo_to_slot, lra_spill): Likewise.
13983         * omp-low.c (omp_clause_aligned_alignment): Likewise.
13984         * optabs-query.c (get_best_extraction_insn): Likewise.
13985         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13986         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13987         (expand_mult_highpart, valid_multiword_target_p): Likewise.
13988         * recog.c (offsettable_address_addr_space_p): Likewise.
13989         * regcprop.c (maybe_mode_change): Likewise.
13990         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13991         * regrename.c (build_def_use): Likewise.
13992         * regstat.c (dump_reg_info): Likewise.
13993         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13994         (find_reloads, find_reloads_subreg_address): Likewise.
13995         * reload1.c (eliminate_regs_1): Likewise.
13996         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13997         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13998         (simplify_binary_operation_1, simplify_subreg): Likewise.
13999         * targhooks.c (default_function_arg_padding): Likewise.
14000         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
14001         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
14002         (verify_gimple_assign_ternary): Likewise.
14003         * tree-inline.c (estimate_move_cost): Likewise.
14004         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14005         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
14006         (get_address_cost_ainc): Likewise.
14007         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
14008         (vect_supportable_dr_alignment): Likewise.
14009         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14010         (vectorizable_reduction): Likewise.
14011         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
14012         (vectorizable_operation, vectorizable_load): Likewise.
14013         * tree.c (build_same_sized_truth_vector_type): Likewise.
14014         * valtrack.c (cleanup_auto_inc_dec): Likewise.
14015         * var-tracking.c (emit_note_insn_var_location): Likewise.
14016         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
14017         (ADDR_VEC_ALIGN): Likewise.
14018
14019 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14020             Alan Hayward  <alan.hayward@arm.com>
14021             David Sherwood  <david.sherwood@arm.com>
14022
14023         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
14024         unsigned short.
14025         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14026         or if measurement_type is polynomial.
14027         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
14028         * combine.c (make_extraction): Likewise.
14029         * dse.c (find_shift_sequence): Likewise.
14030         * dwarf2out.c (mem_loc_descriptor): Likewise.
14031         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
14032         (extract_bit_field, extract_low_bits): Likewise.
14033         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
14034         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
14035         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
14036         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
14037         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14038         * reload.c (find_reloads): Likewise.
14039         * reload1.c (alter_reg): Likewise.
14040         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
14041         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
14042         * tree-if-conv.c (predicate_mem_writes): Likewise.
14043         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
14044         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
14045         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
14046         * valtrack.c (dead_debug_insert_temp): Likewise.
14047         * varasm.c (mergeable_constant_section): Likewise.
14048         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
14049
14050 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14051             Alan Hayward  <alan.hayward@arm.com>
14052             David Sherwood  <david.sherwood@arm.com>
14053
14054         * expr.c (expand_assignment): Cope with polynomial mode sizes
14055         when assigning to a CONCAT.
14056
14057 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14058             Alan Hayward  <alan.hayward@arm.com>
14059             David Sherwood  <david.sherwood@arm.com>
14060
14061         * machmode.h (mode_precision): Change from unsigned short to
14062         poly_uint16_pod.
14063         (mode_to_precision): Return a poly_uint16 rather than an unsigned
14064         short.
14065         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
14066         or if measurement_type is not polynomial.
14067         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
14068         in which the mode is already known to be a scalar_int_mode.
14069         * genmodes.c (emit_mode_precision): Change the type of mode_precision
14070         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
14071         initializer.
14072         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14073         for GET_MODE_PRECISION.
14074         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14075         for GET_MODE_PRECISION.
14076         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
14077         as polynomial.
14078         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
14079         (expand_field_assignment, make_extraction): Likewise.
14080         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
14081         (get_last_value): Likewise.
14082         * convert.c (convert_to_integer_1): Likewise.
14083         * cse.c (cse_insn): Likewise.
14084         * expr.c (expand_expr_real_1): Likewise.
14085         * lra-constraints.c (simplify_operand_subreg): Likewise.
14086         * optabs-query.c (can_atomic_load_p): Likewise.
14087         * optabs.c (expand_atomic_load): Likewise.
14088         (expand_atomic_store): Likewise.
14089         * ree.c (combine_reaching_defs): Likewise.
14090         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
14091         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
14092         * tree.h (type_has_mode_precision_p): Likewise.
14093         * ubsan.c (instrument_si_overflow): Likewise.
14094
14095 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14096             Alan Hayward  <alan.hayward@arm.com>
14097             David Sherwood  <david.sherwood@arm.com>
14098
14099         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
14100         polynomial numbers of units.
14101         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
14102         (valid_vector_subparts_p): New function.
14103         (build_vector_type): Remove temporary shim and take the number
14104         of units as a poly_uint64 rather than an int.
14105         (build_opaque_vector_type): Take the number of units as a
14106         poly_uint64 rather than an int.
14107         * tree.c (build_vector_from_ctor): Handle polynomial
14108         TYPE_VECTOR_SUBPARTS.
14109         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
14110         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
14111         (build_vector_from_val): If the number of units is variable,
14112         use build_vec_duplicate_cst for constant operands and
14113         VEC_DUPLICATE_EXPR otherwise.
14114         (make_vector_type): Remove temporary is_constant ().
14115         (build_vector_type, build_opaque_vector_type): Take the number of
14116         units as a poly_uint64 rather than an int.
14117         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
14118         VECTOR_CST_NELTS.
14119         * cfgexpand.c (expand_debug_expr): Likewise.
14120         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
14121         (store_constructor, expand_expr_real_1): Likewise.
14122         (const_scalar_mask_from_tree): Likewise.
14123         * fold-const-call.c (fold_const_reduction): Likewise.
14124         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
14125         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
14126         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
14127         (fold_relational_const): Likewise.
14128         (native_interpret_vector): Likewise.  Change the size from an
14129         int to an unsigned int.
14130         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
14131         TYPE_VECTOR_SUBPARTS.
14132         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
14133         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
14134         duplicating a non-constant operand into a variable-length vector.
14135         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
14136         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
14137         * ipa-icf.c (sem_variable::equals): Likewise.
14138         * match.pd: Likewise.
14139         * omp-simd-clone.c (simd_clone_subparts): Likewise.
14140         * print-tree.c (print_node): Likewise.
14141         * stor-layout.c (layout_type): Likewise.
14142         * targhooks.c (default_builtin_vectorization_cost): Likewise.
14143         * tree-cfg.c (verify_gimple_comparison): Likewise.
14144         (verify_gimple_assign_binary): Likewise.
14145         (verify_gimple_assign_ternary): Likewise.
14146         (verify_gimple_assign_single): Likewise.
14147         * tree-pretty-print.c (dump_generic_node): Likewise.
14148         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14149         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
14150         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
14151         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
14152         (vect_shift_permute_load_chain): Likewise.
14153         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
14154         (expand_vector_condition, optimize_vector_constructor): Likewise.
14155         (lower_vec_perm, get_compute_type): Likewise.
14156         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14157         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
14158         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
14159         (vect_recog_mask_conversion_pattern): Likewise.
14160         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
14161         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
14162         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14163         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
14164         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
14165         (vectorizable_shift, vectorizable_operation, vectorizable_store)
14166         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
14167         (supportable_widening_operation): Likewise.
14168         (supportable_narrowing_operation): Likewise.
14169         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
14170         Likewise.
14171         * varasm.c (output_constant): Likewise.
14172
14173 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14174             Alan Hayward  <alan.hayward@arm.com>
14175             David Sherwood  <david.sherwood@arm.com>
14176
14177         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
14178         so that both the length == 3 and length != 3 cases set up their
14179         own permute vectors.  Add comments explaining why we know the
14180         number of elements is constant.
14181         (vect_permute_load_chain): Likewise.
14182
14183 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14184             Alan Hayward  <alan.hayward@arm.com>
14185             David Sherwood  <david.sherwood@arm.com>
14186
14187         * machmode.h (mode_nunits): Change from unsigned char to
14188         poly_uint16_pod.
14189         (ONLY_FIXED_SIZE_MODES): New macro.
14190         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
14191         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
14192         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
14193         New typedefs.
14194         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
14195         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
14196         or if measurement_type is not polynomial.
14197         * genmodes.c (ZERO_COEFFS): New macro.
14198         (emit_mode_nunits_inline): Make mode_nunits_inline return a
14199         poly_uint16.
14200         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
14201         Use ZERO_COEFFS when emitting initializers.
14202         * data-streamer.h (bp_pack_poly_value): New function.
14203         (bp_unpack_poly_value): Likewise.
14204         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14205         for GET_MODE_NUNITS.
14206         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14207         for GET_MODE_NUNITS.
14208         * tree.c (make_vector_type): Remove temporary shim and make
14209         the real function take the number of units as a poly_uint64
14210         rather than an int.
14211         (build_vector_type_for_mode): Handle polynomial nunits.
14212         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
14213         * emit-rtl.c (const_vec_series_p_1): Likewise.
14214         (gen_rtx_CONST_VECTOR): Likewise.
14215         * fold-const.c (test_vec_duplicate_folding): Likewise.
14216         * genrecog.c (validate_pattern): Likewise.
14217         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
14218         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14219         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
14220         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
14221         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
14222         * rtlanal.c (subreg_get_info): Likewise.
14223         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14224         (vect_grouped_load_supported): Likewise.
14225         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14226         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14227         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14228         (simplify_const_unary_operation, simplify_binary_operation_1)
14229         (simplify_const_binary_operation, simplify_ternary_operation)
14230         (test_vector_ops_duplicate, test_vector_ops): Likewise.
14231         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14232         instead of CONST_VECTOR_NUNITS.
14233         * varasm.c (output_constant_pool_2): Likewise.
14234         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14235         explicit-encoded elements in the XVEC for variable-length vectors.
14236
14237 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14238
14239         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14240
14241 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14242             Alan Hayward  <alan.hayward@arm.com>
14243             David Sherwood  <david.sherwood@arm.com>
14244
14245         * coretypes.h (fixed_size_mode): Declare.
14246         (fixed_size_mode_pod): New typedef.
14247         * builtins.h (target_builtins::x_apply_args_mode)
14248         (target_builtins::x_apply_result_mode): Change type to
14249         fixed_size_mode_pod.
14250         * builtins.c (apply_args_size, apply_result_size, result_vector)
14251         (expand_builtin_apply_args_1, expand_builtin_apply)
14252         (expand_builtin_return): Update accordingly.
14253
14254 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14255
14256         * cse.c (hash_rtx_cb): Hash only the encoded elements.
14257         * cselib.c (cselib_hash_rtx): Likewise.
14258         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14259         CONST_VECTOR encoding.
14260
14261 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
14262             Jeff Law  <law@redhat.com>
14263
14264         PR target/83641
14265         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14266         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14267         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14268         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14269
14270         PR target/83641
14271         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14272         explicitly probe *sp in a noreturn function if there were any callee
14273         register saves or frame pointer is needed.
14274
14275 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
14276
14277         PR debug/83621
14278         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14279         BLKmode for ternary, binary or unary expressions.
14280
14281         PR debug/83645
14282         * var-tracking.c (delete_vta_debug_insn): New inline function.
14283         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14284         insns from get_insns () to NULL instead of each bb separately.
14285         Use delete_vta_debug_insn.  No longer static.
14286         (vt_debug_insns_local, variable_tracking_main_1): Adjust
14287         delete_vta_debug_insns callers.
14288         * rtl.h (delete_vta_debug_insns): Declare.
14289         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14290         instead of variable_tracking_main.
14291
14292 2018-01-03  Martin Sebor  <msebor@redhat.com>
14293
14294         PR tree-optimization/83603
14295         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14296         arguments past the endof the argument list in functions declared
14297         without a prototype.
14298         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14299         Avoid checking when arguments are null.
14300
14301 2018-01-03  Martin Sebor  <msebor@redhat.com>
14302
14303         PR c/83559
14304         * doc/extend.texi (attribute const): Fix a typo.
14305         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14306         issuing -Wsuggest-attribute for void functions.
14307
14308 2018-01-03  Martin Sebor  <msebor@redhat.com>
14309
14310         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14311         offset_int::from instead of wide_int::to_shwi.
14312         (maybe_diag_overlap): Remove assertion.
14313         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14314         * gimple-ssa-sprintf.c (format_directive): Same.
14315         (parse_directive): Same.
14316         (sprintf_dom_walker::compute_format_length): Same.
14317         (try_substitute_return_value): Same.
14318
14319 2018-01-03  Jeff Law  <law@redhat.com>
14320
14321         PR middle-end/83654
14322         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14323         non-constant residual for zero at runtime and avoid probing in
14324         that case.  Reorganize code for trailing problem to mirror handling
14325         of the residual.
14326
14327 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14328
14329         PR tree-optimization/83501
14330         * tree-ssa-strlen.c (get_string_cst): New.
14331         (handle_char_store): Call get_string_cst.
14332
14333 2018-01-03  Martin Liska  <mliska@suse.cz>
14334
14335         PR tree-optimization/83593
14336         * tree-ssa-strlen.c: Include tree-cfg.h.
14337         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14338         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14339         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14340         to false.
14341         (strlen_dom_walker::before_dom_children): Call
14342         gimple_purge_dead_eh_edges. Dump tranformation with details
14343         dump flags.
14344         (strlen_dom_walker::before_dom_children): Update call by adding
14345         new argument cleanup_eh.
14346         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14347
14348 2018-01-03  Martin Liska  <mliska@suse.cz>
14349
14350         PR ipa/83549
14351         * cif-code.def (VARIADIC_THUNK): New enum value.
14352         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14353         thunks.
14354
14355 2018-01-03  Jan Beulich  <jbeulich@suse.com>
14356
14357         * sse.md (mov<mode>_internal): Tighten condition for when to use
14358         vmovdqu<ssescalarsize> for TI and OI modes.
14359
14360 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
14361
14362         Update copyright years.
14363
14364 2018-01-03  Martin Liska  <mliska@suse.cz>
14365
14366         PR ipa/83594
14367         * ipa-visibility.c (function_and_variable_visibility): Skip
14368         functions with noipa attribure.
14369
14370 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
14371
14372         * gcc.c (process_command): Update copyright notice dates.
14373         * gcov-dump.c (print_version): Ditto.
14374         * gcov.c (print_version): Ditto.
14375         * gcov-tool.c (print_version): Ditto.
14376         * gengtype.c (create_file): Ditto.
14377         * doc/cpp.texi: Bump @copying's copyright year.
14378         * doc/cppinternals.texi: Ditto.
14379         * doc/gcc.texi: Ditto.
14380         * doc/gccint.texi: Ditto.
14381         * doc/gcov.texi: Ditto.
14382         * doc/install.texi: Ditto.
14383         * doc/invoke.texi: Ditto.
14384
14385 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14386
14387         * vector-builder.h (vector_builder::m_full_nelts): Change from
14388         unsigned int to poly_uint64.
14389         (vector_builder::full_nelts): Update prototype accordingly.
14390         (vector_builder::new_vector): Likewise.
14391         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14392         (vector_builder::operator ==): Likewise.
14393         (vector_builder::finalize): Likewise.
14394         * int-vector-builder.h (int_vector_builder::int_vector_builder):
14395         Take the number of elements as a poly_uint64 rather than an
14396         unsigned int.
14397         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14398         from unsigned int to poly_uint64.
14399         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14400         (vec_perm_indices::new_vector): Likewise.
14401         (vec_perm_indices::length): Likewise.
14402         (vec_perm_indices::nelts_per_input): Likewise.
14403         (vec_perm_indices::input_nelts): Likewise.
14404         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14405         number of elements per input as a poly_uint64 rather than an
14406         unsigned int.  Use the original encoding for variable-length
14407         vectors, rather than clamping each individual element.
14408         For the second and subsequent elements in each pattern,
14409         clamp the step and base before clamping their sum.
14410         (vec_perm_indices::series_p): Handle polynomial element counts.
14411         (vec_perm_indices::all_in_range_p): Likewise.
14412         (vec_perm_indices_to_tree): Likewise.
14413         (vec_perm_indices_to_rtx): Likewise.
14414         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14415         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14416         (tree_vector_builder::new_binary_operation): Handle polynomial
14417         element counts.  Return false if we need to know the number
14418         of elements at compile time.
14419         * fold-const.c (fold_vec_perm): Punt if the number of elements
14420         isn't known at compile time.
14421
14422 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14423
14424         * vec-perm-indices.h (vec_perm_builder): Change element type
14425         from HOST_WIDE_INT to poly_int64.
14426         (vec_perm_indices::element_type): Update accordingly.
14427         (vec_perm_indices::clamp): Handle polynomial element_types.
14428         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14429         (vec_perm_indices::all_in_range_p): Likewise.
14430         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14431         than shwi trees.
14432         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14433         polynomial vec_perm_indices element types.
14434         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14435         * fold-const.c (fold_vec_perm): Likewise.
14436         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14437         * tree-vect-generic.c (lower_vec_perm): Likewise.
14438         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14439         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14440         element type to HOST_WIDE_INT.
14441
14442 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14443             Alan Hayward  <alan.hayward@arm.com>
14444             David Sherwood  <david.sherwood@arm.com>
14445
14446         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14447         rather than an int.  Use plus_constant.
14448         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14449         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14450
14451 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14452             Alan Hayward  <alan.hayward@arm.com>
14453             David Sherwood  <david.sherwood@arm.com>
14454
14455         * calls.c (emit_call_1, expand_call): Change struct_value_size from
14456         a HOST_WIDE_INT to a poly_int64.
14457
14458 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14459             Alan Hayward  <alan.hayward@arm.com>
14460             David Sherwood  <david.sherwood@arm.com>
14461
14462         * calls.c (load_register_parameters): Cope with polynomial
14463         mode sizes.  Require a constant size for BLKmode parameters
14464         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
14465         forces a parameter to be padded at the lsb end in order to
14466         fill a complete number of words, require the parameter size
14467         to be ordered wrt UNITS_PER_WORD.
14468
14469 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14470             Alan Hayward  <alan.hayward@arm.com>
14471             David Sherwood  <david.sherwood@arm.com>
14472
14473         * reload1.c (spill_stack_slot_width): Change element type
14474         from unsigned int to poly_uint64_pod.
14475         (alter_reg): Treat mode sizes as polynomial.
14476
14477 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14478             Alan Hayward  <alan.hayward@arm.com>
14479             David Sherwood  <david.sherwood@arm.com>
14480
14481         * reload.c (complex_word_subreg_p): New function.
14482         (reload_inner_reg_of_subreg, push_reload): Use it.
14483
14484 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14485             Alan Hayward  <alan.hayward@arm.com>
14486             David Sherwood  <david.sherwood@arm.com>
14487
14488         * lra-constraints.c (process_alt_operands): Reject matched
14489         operands whose sizes aren't ordered.
14490         (match_reload): Refer to this check here.
14491
14492 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14493             Alan Hayward  <alan.hayward@arm.com>
14494             David Sherwood  <david.sherwood@arm.com>
14495
14496         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14497         that the mode size is in the set {1, 2, 4, 8, 16}.
14498
14499 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14500             Alan Hayward  <alan.hayward@arm.com>
14501             David Sherwood  <david.sherwood@arm.com>
14502
14503         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14504         Use plus_constant instead of gen_rtx_PLUS.
14505
14506 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14507             Alan Hayward  <alan.hayward@arm.com>
14508             David Sherwood  <david.sherwood@arm.com>
14509
14510         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14511         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14512         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14513         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14514         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14515         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14516         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14517         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14518         * config/i386/i386.c (ix86_push_rounding): ...this new function.
14519         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14520         a poly_int64.
14521         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14522         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14523         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14524         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14525         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14526         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14527         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14528         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14529         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14530         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14531         function.
14532         * expr.c (emit_move_resolve_push): Treat the input and result
14533         of PUSH_ROUNDING as a poly_int64.
14534         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14535         (emit_push_insn): Likewise.
14536         * lra-eliminations.c (mark_not_eliminable): Likewise.
14537         * recog.c (push_operand): Likewise.
14538         * reload1.c (elimination_effects): Likewise.
14539         * rtlanal.c (nonzero_bits1): Likewise.
14540         * calls.c (store_one_arg): Likewise.  Require the padding to be
14541         known at compile time.
14542
14543 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14544             Alan Hayward  <alan.hayward@arm.com>
14545             David Sherwood  <david.sherwood@arm.com>
14546
14547         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14548         Use plus_constant instead of gen_rtx_PLUS.
14549
14550 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14551             Alan Hayward  <alan.hayward@arm.com>
14552             David Sherwood  <david.sherwood@arm.com>
14553
14554         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14555         rather than an int.
14556
14557 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14558             Alan Hayward  <alan.hayward@arm.com>
14559             David Sherwood  <david.sherwood@arm.com>
14560
14561         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14562         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14563         via stack temporaries.  Treat the mode size as polynomial too.
14564
14565 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14566             Alan Hayward  <alan.hayward@arm.com>
14567             David Sherwood  <david.sherwood@arm.com>
14568
14569         * expr.c (expand_expr_real_2): When handling conversions involving
14570         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14571         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
14572         as a poly_uint64 too.
14573
14574 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14575             Alan Hayward  <alan.hayward@arm.com>
14576             David Sherwood  <david.sherwood@arm.com>
14577
14578         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14579
14580 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14581             Alan Hayward  <alan.hayward@arm.com>
14582             David Sherwood  <david.sherwood@arm.com>
14583
14584         * combine.c (can_change_dest_mode): Handle polynomial
14585         REGMODE_NATURAL_SIZE.
14586         * expmed.c (store_bit_field_1): Likewise.
14587         * expr.c (store_constructor): Likewise.
14588         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14589         and polynomial REGMODE_NATURAL_SIZE.
14590         (gen_lowpart_common): Likewise.
14591         * reginfo.c (record_subregs_of_mode): Likewise.
14592         * rtlanal.c (read_modify_subreg_p): Likewise.
14593
14594 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14595             Alan Hayward  <alan.hayward@arm.com>
14596             David Sherwood  <david.sherwood@arm.com>
14597
14598         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14599         numbers of elements.
14600
14601 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14602             Alan Hayward  <alan.hayward@arm.com>
14603             David Sherwood  <david.sherwood@arm.com>
14604
14605         * match.pd: Cope with polynomial numbers of vector elements.
14606
14607 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14608             Alan Hayward  <alan.hayward@arm.com>
14609             David Sherwood  <david.sherwood@arm.com>
14610
14611         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14612         in a POINTER_PLUS_EXPR.
14613
14614 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14615             Alan Hayward  <alan.hayward@arm.com>
14616             David Sherwood  <david.sherwood@arm.com>
14617
14618         * omp-simd-clone.c (simd_clone_subparts): New function.
14619         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14620         (ipa_simd_modify_function_body): Likewise.
14621
14622 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14623             Alan Hayward  <alan.hayward@arm.com>
14624             David Sherwood  <david.sherwood@arm.com>
14625
14626         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14627         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14628         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14629         (expand_vector_condition, vector_element): Likewise.
14630         (subparts_gt): New function.
14631         (get_compute_type): Use subparts_gt.
14632         (count_type_subparts): Delete.
14633         (expand_vector_operations_1): Use subparts_gt instead of
14634         count_type_subparts.
14635
14636 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14637             Alan Hayward  <alan.hayward@arm.com>
14638             David Sherwood  <david.sherwood@arm.com>
14639
14640         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14641         (vect_compile_time_alias): ...this new function.  Do the calculation
14642         on poly_ints rather than trees.
14643         (vect_prune_runtime_alias_test_list): Update call accordingly.
14644
14645 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14646             Alan Hayward  <alan.hayward@arm.com>
14647             David Sherwood  <david.sherwood@arm.com>
14648
14649         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14650         numbers of units.
14651         (vect_schedule_slp_instance): Likewise.
14652
14653 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14654             Alan Hayward  <alan.hayward@arm.com>
14655             David Sherwood  <david.sherwood@arm.com>
14656
14657         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14658         constant and extern definitions for variable-length vectors.
14659         (vect_get_constant_vectors): Note that the number of units
14660         is known to be constant.
14661
14662 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14663             Alan Hayward  <alan.hayward@arm.com>
14664             David Sherwood  <david.sherwood@arm.com>
14665
14666         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14667         of units as polynomial.  Choose between WIDE and NARROW based
14668         on multiple_p.
14669
14670 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14671             Alan Hayward  <alan.hayward@arm.com>
14672             David Sherwood  <david.sherwood@arm.com>
14673
14674         * tree-vect-stmts.c (simd_clone_subparts): New function.
14675         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14676
14677 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14678             Alan Hayward  <alan.hayward@arm.com>
14679             David Sherwood  <david.sherwood@arm.com>
14680
14681         * tree-vect-stmts.c (vectorizable_call): Treat the number of
14682         vectors as polynomial.  Use build_index_vector for
14683         IFN_GOMP_SIMD_LANE.
14684
14685 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14686             Alan Hayward  <alan.hayward@arm.com>
14687             David Sherwood  <david.sherwood@arm.com>
14688
14689         * tree-vect-stmts.c (get_load_store_type): Treat the number of
14690         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14691         for variable-length vectors.
14692         (vectorizable_mask_load_store): Treat the number of units as
14693         polynomial, asserting that it is constant if the condition has
14694         already been enforced.
14695         (vectorizable_store, vectorizable_load): Likewise.
14696
14697 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14698             Alan Hayward  <alan.hayward@arm.com>
14699             David Sherwood  <david.sherwood@arm.com>
14700
14701         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14702         of units as polynomial.  Punt if we can't tell at compile time
14703         which vector contains the final result.
14704
14705 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14706             Alan Hayward  <alan.hayward@arm.com>
14707             David Sherwood  <david.sherwood@arm.com>
14708
14709         * tree-vect-loop.c (vectorizable_induction): Treat the number
14710         of units as polynomial.  Punt on SLP inductions.  Use an integer
14711         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
14712         cast of such a series for variable-length floating-point
14713         reductions.
14714
14715 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14716             Alan Hayward  <alan.hayward@arm.com>
14717             David Sherwood  <david.sherwood@arm.com>
14718
14719         * tree.h (build_index_vector): Declare.
14720         * tree.c (build_index_vector): New function.
14721         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14722         of units as polynomial, forcibly converting it to a constant if
14723         vectorizable_reduction has already enforced the condition.
14724         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
14725         to create a {1,2,3,...} vector.
14726         (vectorizable_reduction): Treat the number of units as polynomial.
14727         Choose vectype_in based on the largest scalar element size rather
14728         than the smallest number of units.  Enforce the restrictions
14729         relied on above.
14730
14731 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14732             Alan Hayward  <alan.hayward@arm.com>
14733             David Sherwood  <david.sherwood@arm.com>
14734
14735         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14736         number of units as polynomial.
14737
14738 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14739             Alan Hayward  <alan.hayward@arm.com>
14740             David Sherwood  <david.sherwood@arm.com>
14741
14742         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14743         * target.def (autovectorize_vector_sizes): Return the vector sizes
14744         by pointer, using vector_sizes rather than a bitmask.
14745         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14746         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14747         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14748         Likewise.
14749         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14750         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14751         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14752         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14753         * omp-general.c (omp_max_vf): Likewise.
14754         * omp-low.c (omp_clause_aligned_alignment): Likewise.
14755         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14756         * tree-vect-loop.c (vect_analyze_loop): Likewise.
14757         * tree-vect-slp.c (vect_slp_bb): Likewise.
14758         * doc/tm.texi: Regenerate.
14759         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14760         to a poly_uint64.
14761         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14762         the vector size as a poly_uint64 rather than an unsigned int.
14763         (current_vector_size): Change from an unsigned int to a poly_uint64.
14764         (get_vectype_for_scalar_type): Update accordingly.
14765         * tree.h (build_truth_vector_type): Take the size and number of
14766         units as a poly_uint64 rather than an unsigned int.
14767         (build_vector_type): Add a temporary overload that takes
14768         the number of units as a poly_uint64 rather than an unsigned int.
14769         * tree.c (make_vector_type): Likewise.
14770         (build_truth_vector_type): Take the number of units as a poly_uint64
14771         rather than an unsigned int.
14772
14773 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14774             Alan Hayward  <alan.hayward@arm.com>
14775             David Sherwood  <david.sherwood@arm.com>
14776
14777         * target.def (get_mask_mode): Take the number of units and length
14778         as poly_uint64s rather than unsigned ints.
14779         * targhooks.h (default_get_mask_mode): Update accordingly.
14780         * targhooks.c (default_get_mask_mode): Likewise.
14781         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14782         * doc/tm.texi: Regenerate.
14783
14784 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14785             Alan Hayward  <alan.hayward@arm.com>
14786             David Sherwood  <david.sherwood@arm.com>
14787
14788         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14789         * omp-general.c (omp_max_vf): Likewise.
14790         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14791         (expand_omp_simd): Handle polynomial safelen.
14792         * omp-low.c (omplow_simd_context): Add a default constructor.
14793         (omplow_simd_context::max_vf): Change from int to poly_uint64.
14794         (lower_rec_simd_input_clauses): Update accordingly.
14795         (lower_rec_input_clauses): Likewise.
14796
14797 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14798             Alan Hayward  <alan.hayward@arm.com>
14799             David Sherwood  <david.sherwood@arm.com>
14800
14801         * tree-vectorizer.h (vect_nunits_for_cost): New function.
14802         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14803         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14804         (vect_analyze_slp_cost): Likewise.
14805         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14806         (vect_model_load_cost): Likewise.
14807
14808 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14809             Alan Hayward  <alan.hayward@arm.com>
14810             David Sherwood  <david.sherwood@arm.com>
14811
14812         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14813         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14814         from an unsigned int * to a poly_uint64_pod *.
14815         (calculate_unrolling_factor): New function.
14816         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
14817
14818 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14819             Alan Hayward  <alan.hayward@arm.com>
14820             David Sherwood  <david.sherwood@arm.com>
14821
14822         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14823         from an unsigned int to a poly_uint64.
14824         (_loop_vec_info::slp_unrolling_factor): Likewise.
14825         (_loop_vec_info::vectorization_factor): Change from an int
14826         to a poly_uint64.
14827         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14828         (vect_get_num_vectors): New function.
14829         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14830         (vect_get_num_copies): Use vect_get_num_vectors.
14831         (vect_analyze_data_ref_dependences): Change max_vf from an int *
14832         to an unsigned int *.
14833         (vect_analyze_data_refs): Change min_vf from an int * to a
14834         poly_uint64 *.
14835         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14836         than an unsigned HOST_WIDE_INT.
14837         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14838         (vect_analyze_data_ref_dependence): Change max_vf from an int *
14839         to an unsigned int *.
14840         (vect_analyze_data_ref_dependences): Likewise.
14841         (vect_compute_data_ref_alignment): Handle polynomial vf.
14842         (vect_enhance_data_refs_alignment): Likewise.
14843         (vect_prune_runtime_alias_test_list): Likewise.
14844         (vect_shift_permute_load_chain): Likewise.
14845         (vect_supportable_dr_alignment): Likewise.
14846         (dependence_distance_ge_vf): Take the vectorization factor as a
14847         poly_uint64 rather than an unsigned HOST_WIDE_INT.
14848         (vect_analyze_data_refs): Change min_vf from an int * to a
14849         poly_uint64 *.
14850         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14851         vfm1 as a poly_uint64 rather than an int.  Make the same change
14852         for the returned bound_scalar.
14853         (vect_gen_vector_loop_niters): Handle polynomial vf.
14854         (vect_do_peeling): Likewise.  Update call to
14855         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14856         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14857         be constant.
14858         * tree-vect-loop.c (vect_determine_vectorization_factor)
14859         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14860         (vect_get_known_peeling_cost): Likewise.
14861         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14862         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14863         (vect_transform_loop): Likewise.  Use the lowest possible VF when
14864         updating the upper bounds of the loop.
14865         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
14866         rather than an int.
14867         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14868         polynomial unroll factors.
14869         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14870         (vect_make_slp_decision): Likewise.
14871         (vect_supported_load_permutation_p): Likewise, and polynomial
14872         vf too.
14873         (vect_analyze_slp_cost): Handle polynomial vf.
14874         (vect_slp_analyze_node_operations): Likewise.
14875         (vect_slp_analyze_bb_1): Likewise.
14876         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14877         than an unsigned HOST_WIDE_INT.
14878         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14879         (vectorizable_load): Handle polynomial vf.
14880         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14881         a poly_uint64.
14882         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14883
14884 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14885             Alan Hayward  <alan.hayward@arm.com>
14886             David Sherwood  <david.sherwood@arm.com>
14887
14888         * match.pd: Handle bit operations involving three constants
14889         and try to fold one pair.
14890
14891 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
14892
14893         * tree-vect-loop-manip.c: Include gimple-fold.h.
14894         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14895         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
14896         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14897         Add a path that uses a step of VF instead of 1, but disable it
14898         for now.
14899         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14900         and niters_no_overflow parameters.  Update calls to
14901         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14902         Create a new SSA name if the latter choses to use a ste other
14903         than zero, and return it via niters_vector_mult_vf_var.
14904         * tree-vect-loop.c (vect_transform_loop): Update calls to
14905         vect_do_peeling, vect_gen_vector_loop_niters and
14906         slpeel_make_loop_iterate_ntimes.
14907         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14908         (vect_gen_vector_loop_niters): Update declarations after above changes.
14909
14910 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
14911
14912         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14913         128-bit round to integer instructions.
14914         (ceil<mode>2): Likewise.
14915         (btrunc<mode>2): Likewise.
14916         (round<mode>2): Likewise.
14917
14918 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
14919
14920         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14921         unaligned VSX load/store on P8/P9.
14922         (expand_block_clear): Allow the use of unaligned VSX
14923         load/store on P8/P9.
14924
14925 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
14926
14927         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14928         New function.
14929         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14930         swap associated with both a load and a store.
14931
14932 2018-01-02  Andrew Waterman  <andrew@sifive.com>
14933
14934         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14935         * config/riscv/riscv.md (clear_cache): Use it.
14936
14937 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
14938
14939         * web.c: Remove out-of-date comment.
14940
14941 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14942
14943         * expr.c (fixup_args_size_notes): Check that any existing
14944         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14945         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14946         (emit_single_push_insn): ...here.
14947
14948 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14949
14950         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14951         (const_vector_encoded_nelts): New function.
14952         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14953         (const_vector_int_elt, const_vector_elt): Declare.
14954         * emit-rtl.c (const_vector_int_elt_1): New function.
14955         (const_vector_elt): Likewise.
14956         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14957         of CONST_VECTOR_ELT.
14958
14959 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14960
14961         * expr.c: Include rtx-vector-builder.h.
14962         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14963         directly on the tree encoding.
14964         (const_vector_from_tree): Likewise.
14965         * optabs.c: Include rtx-vector-builder.h.
14966         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14967         sequence of "u" values.
14968         * vec-perm-indices.c: Include rtx-vector-builder.h.
14969         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14970         directly on the vec_perm_indices encoding.
14971
14972 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14973
14974         * doc/rtl.texi (const_vector): Describe new encoding scheme.
14975         * Makefile.in (OBJS): Add rtx-vector-builder.o.
14976         * rtx-vector-builder.h: New file.
14977         * rtx-vector-builder.c: Likewise.
14978         * rtl.h (rtx_def::u2): Add a const_vector field.
14979         (CONST_VECTOR_NPATTERNS): New macro.
14980         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14981         (CONST_VECTOR_DUPLICATE_P): Likewise.
14982         (CONST_VECTOR_STEPPED_P): Likewise.
14983         (CONST_VECTOR_ENCODED_ELT): Likewise.
14984         (const_vec_duplicate_p): Check for a duplicated vector encoding.
14985         (unwrap_const_vec_duplicate): Likewise.
14986         (const_vec_series_p): Check for a non-duplicated vector encoding.
14987         Say that the function only returns true for integer vectors.
14988         * emit-rtl.c: Include rtx-vector-builder.h.
14989         (gen_const_vec_duplicate_1): Delete.
14990         (gen_const_vector): Call gen_const_vec_duplicate instead of
14991         gen_const_vec_duplicate_1.
14992         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14993         (gen_const_vec_duplicate): Use rtx_vector_builder.
14994         (gen_const_vec_series): Likewise.
14995         (gen_rtx_CONST_VECTOR): Likewise.
14996         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14997         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14998         Build a new vector rather than modifying a CONST_VECTOR in-place.
14999         (handle_special_swappables): Update call accordingly.
15000         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
15001         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15002         Build a new vector rather than modifying a CONST_VECTOR in-place.
15003         (handle_special_swappables): Update call accordingly.
15004
15005 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15006
15007         * simplify-rtx.c (simplify_const_binary_operation): Use
15008         CONST_VECTOR_ELT instead of XVECEXP.
15009
15010 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15011
15012         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
15013         the selector elements to be different from the data elements
15014         if the selector is a VECTOR_CST.
15015         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
15016         ssizetype for the selector.
15017
15018 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15019
15020         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
15021         before testing each element individually.
15022         * tree-vect-generic.c (lower_vec_perm): Likewise.
15023
15024 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15025
15026         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
15027         * selftest-run-tests.c (selftest::run_tests): Call it.
15028         * vector-builder.h (vector_builder::operator ==): New function.
15029         (vector_builder::operator !=): Likewise.
15030         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
15031         (vec_perm_indices::all_from_input_p): New function.
15032         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15033         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
15034         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
15035         instead of reading the VECTOR_CST directly.  Detect whether both
15036         vector inputs are the same before constructing the vec_perm_indices,
15037         and update the number of inputs argument accordingly.  Use the
15038         utility functions added above.  Only construct sel2 if we need to.
15039
15040 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15041
15042         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
15043         the broadcast of the low byte.
15044         (expand_mult_highpart): Use an explicit encoding for the permutes.
15045         * optabs-query.c (can_mult_highpart_p): Likewise.
15046         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
15047         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15048         (vectorizable_bswap): Likewise.
15049         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
15050         explicit encoding for the power-of-2 permutes.
15051         (vect_permute_store_chain): Likewise.
15052         (vect_grouped_load_supported): Likewise.
15053         (vect_permute_load_chain): Likewise.
15054
15055 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15056
15057         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
15058         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
15059         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
15060         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15061         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15062         (vect_gen_perm_mask_any): Likewise.
15063
15064 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15065
15066         * int-vector-builder.h: New file.
15067         * vec-perm-indices.h: Include int-vector-builder.h.
15068         (vec_perm_indices): Redefine as an int_vector_builder.
15069         (auto_vec_perm_indices): Delete.
15070         (vec_perm_builder): Redefine as a stand-alone class.
15071         (vec_perm_indices::vec_perm_indices): New function.
15072         (vec_perm_indices::clamp): Likewise.
15073         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
15074         (vec_perm_indices::new_vector): New function.
15075         (vec_perm_indices::new_expanded_vector): Update for new
15076         vec_perm_indices class.
15077         (vec_perm_indices::rotate_inputs): New function.
15078         (vec_perm_indices::all_in_range_p): Operate directly on the
15079         encoded form, without computing elided elements.
15080         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
15081         encoding.  Update for new vec_perm_indices class.
15082         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
15083         the given vec_perm_builder.
15084         (expand_vec_perm_var): Update vec_perm_builder constructor.
15085         (expand_mult_highpart): Use vec_perm_builder instead of
15086         auto_vec_perm_indices.
15087         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
15088         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
15089         or double series encoding as appropriate.
15090         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
15091         vec_perm_indices instead of auto_vec_perm_indices.
15092         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15093         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15094         (vect_permute_store_chain): Likewise.
15095         (vect_grouped_load_supported): Likewise.
15096         (vect_permute_load_chain): Likewise.
15097         (vect_shift_permute_load_chain): Likewise.
15098         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15099         (vect_transform_slp_perm_load): Likewise.
15100         (vect_schedule_slp_instance): Likewise.
15101         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15102         (vectorizable_mask_load_store): Likewise.
15103         (vectorizable_bswap): Likewise.
15104         (vectorizable_store): Likewise.
15105         (vectorizable_load): Likewise.
15106         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
15107         vec_perm_indices instead of auto_vec_perm_indices.  Use
15108         tree_to_vec_perm_builder to read the vector from a tree.
15109         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
15110         vec_perm_builder instead of a vec_perm_indices.
15111         (have_whole_vector_shift): Use vec_perm_builder and
15112         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
15113         truncation to calc_vec_perm_mask_for_shift.
15114         (vect_create_epilog_for_reduction): Likewise.
15115         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
15116         from auto_vec_perm_indices to vec_perm_indices.
15117         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15118         instead of changing individual elements.
15119         (aarch64_vectorize_vec_perm_const): Use new_vector to install
15120         the vector in d.perm.
15121         * config/arm/arm.c (expand_vec_perm_d::perm): Change
15122         from auto_vec_perm_indices to vec_perm_indices.
15123         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15124         instead of changing individual elements.
15125         (arm_vectorize_vec_perm_const): Use new_vector to install
15126         the vector in d.perm.
15127         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
15128         Update vec_perm_builder constructor.
15129         (rs6000_expand_interleave): Likewise.
15130         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
15131         (rs6000_expand_interleave): Likewise.
15132
15133 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15134
15135         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
15136         to qimode could truncate the indices.
15137         * optabs.c (expand_vec_perm_var): Likewise.
15138
15139 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15140
15141         * Makefile.in (OBJS): Add vec-perm-indices.o.
15142         * vec-perm-indices.h: New file.
15143         * vec-perm-indices.c: Likewise.
15144         * target.h (vec_perm_indices): Replace with a forward class
15145         declaration.
15146         (auto_vec_perm_indices): Move to vec-perm-indices.h.
15147         * optabs.h: Include vec-perm-indices.h.
15148         (expand_vec_perm): Delete.
15149         (selector_fits_mode_p, expand_vec_perm_var): Declare.
15150         (expand_vec_perm_const): Declare.
15151         * target.def (vec_perm_const_ok): Replace with...
15152         (vec_perm_const): ...this new hook.
15153         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
15154         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
15155         * doc/tm.texi: Regenerate.
15156         * optabs.def (vec_perm_const): Delete.
15157         * doc/md.texi (vec_perm_const): Likewise.
15158         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
15159         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
15160         expand_vec_perm for constant permutation vectors.  Assert that
15161         the mode of variable permutation vectors is the integer equivalent
15162         of the mode that is being permuted.
15163         * optabs-query.h (selector_fits_mode_p): Declare.
15164         * optabs-query.c: Include vec-perm-indices.h.
15165         (selector_fits_mode_p): New function.
15166         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
15167         is defined, instead of checking whether the vec_perm_const_optab
15168         exists.  Use targetm.vectorize.vec_perm_const instead of
15169         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
15170         fit in the vector mode before using a variable permute.
15171         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
15172         vec_perm_indices instead of an rtx.
15173         (expand_vec_perm): Replace with...
15174         (expand_vec_perm_const): ...this new function.  Take the selector
15175         as a vec_perm_indices rather than an rtx.  Also take the mode of
15176         the selector.  Update call to shift_amt_for_vec_perm_mask.
15177         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
15178         Use vec_perm_indices::new_expanded_vector to expand the original
15179         selector into bytes.  Check whether the indices fit in the vector
15180         mode before using a variable permute.
15181         (expand_vec_perm_var): Make global.
15182         (expand_mult_highpart): Use expand_vec_perm_const.
15183         * fold-const.c: Includes vec-perm-indices.h.
15184         * tree-ssa-forwprop.c: Likewise.
15185         * tree-vect-data-refs.c: Likewise.
15186         * tree-vect-generic.c: Likewise.
15187         * tree-vect-loop.c: Likewise.
15188         * tree-vect-slp.c: Likewise.
15189         * tree-vect-stmts.c: Likewise.
15190         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
15191         Delete.
15192         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
15193         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
15194         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
15195         (aarch64_vectorize_vec_perm_const): ...this new function.
15196         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15197         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15198         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
15199         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
15200         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15201         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15202         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
15203         into...
15204         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
15205         check for NEON modes.
15206         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
15207         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
15208         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
15209         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
15210         into...
15211         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
15212         the old VEC_PERM_CONST conditions.
15213         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
15214         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
15215         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
15216         (ia64_vectorize_vec_perm_const_ok): Merge into...
15217         (ia64_vectorize_vec_perm_const): ...this new function.
15218         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
15219         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
15220         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
15221         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
15222         * config/mips/mips.c (mips_expand_vec_perm_const)
15223         (mips_vectorize_vec_perm_const_ok): Merge into...
15224         (mips_vectorize_vec_perm_const): ...this new function.
15225         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15226         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15227         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15228         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15229         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15230         (rs6000_expand_vec_perm_const): Delete.
15231         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
15232         Delete.
15233         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15234         (altivec_expand_vec_perm_const_le): Take each operand individually.
15235         Operate on constant selectors rather than rtxes.
15236         (altivec_expand_vec_perm_const): Likewise.  Update call to
15237         altivec_expand_vec_perm_const_le.
15238         (rs6000_expand_vec_perm_const): Delete.
15239         (rs6000_vectorize_vec_perm_const_ok): Delete.
15240         (rs6000_vectorize_vec_perm_const): New function.
15241         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15242         an element count and rtx array.
15243         (rs6000_expand_extract_even): Update call accordingly.
15244         (rs6000_expand_interleave): Likewise.
15245         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15246         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15247         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15248         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15249         (rs6000_expand_vec_perm_const): Delete.
15250         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15251         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15252         (altivec_expand_vec_perm_const_le): Take each operand individually.
15253         Operate on constant selectors rather than rtxes.
15254         (altivec_expand_vec_perm_const): Likewise.  Update call to
15255         altivec_expand_vec_perm_const_le.
15256         (rs6000_expand_vec_perm_const): Delete.
15257         (rs6000_vectorize_vec_perm_const_ok): Delete.
15258         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
15259         reference to the SPE evmerge intructions.
15260         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15261         an element count and rtx array.
15262         (rs6000_expand_extract_even): Update call accordingly.
15263         (rs6000_expand_interleave): Likewise.
15264         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15265         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15266         new function.
15267         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15268
15269 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15270
15271         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15272         vector mode and that that mode matches the mode of the data
15273         being permuted.
15274         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15275         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
15276         directly using expand_vec_perm_1 when forcing selectors into
15277         registers.
15278         (expand_vec_perm_var): New function, split out from expand_vec_perm.
15279
15280 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15281
15282         * optabs-query.h (can_vec_perm_p): Delete.
15283         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15284         * optabs-query.c (can_vec_perm_p): Split into...
15285         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15286         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15287         particular selector is valid.
15288         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15289         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15290         (vect_grouped_load_supported): Likewise.
15291         (vect_shift_permute_load_chain): Likewise.
15292         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15293         (vect_transform_slp_perm_load): Likewise.
15294         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15295         (vectorizable_bswap): Likewise.
15296         (vect_gen_perm_mask_checked): Likewise.
15297         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
15298         implementations of variable permutation vectors into account
15299         when deciding which selector to use.
15300         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15301         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15302         with a false third argument.
15303         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15304         to test whether the constant selector is valid and can_vec_perm_var_p
15305         to test whether a variable selector is valid.
15306
15307 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15308
15309         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15310         * optabs-query.c (can_vec_perm_p): Likewise.
15311         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15312         instead of vec_perm_indices.
15313         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15314         (vect_gen_perm_mask_checked): Likewise,
15315         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15316         (vect_gen_perm_mask_checked): Likewise,
15317
15318 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
15319
15320         * optabs-query.h (qimode_for_vec_perm): Declare.
15321         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15322         (qimode_for_vec_perm): ...this new function.
15323         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15324
15325 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
15326
15327         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15328         does not have a conditional at the top.
15329
15330 2018-01-02  Richard Biener  <rguenther@suse.de>
15331
15332         * ipa-inline.c (big_speedup_p): Fix expression.
15333
15334 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
15335
15336         PR target/81616
15337         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15338         for generic 4->6.
15339
15340 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
15341
15342         PR target/81616
15343         Generic tuning.
15344         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15345         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15346         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15347         cond_taken_branch_cost 3->4.
15348
15349 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
15350
15351         PR tree-optimization/83581
15352         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15353         TODO_cleanup_cfg if any changes have been made.
15354
15355         PR middle-end/83608
15356         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15357         convert_modes if target mode has the right side, but different mode
15358         class.
15359
15360         PR middle-end/83609
15361         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15362         last argument when extracting from CONCAT.  If either from_real or
15363         from_imag is NULL, use expansion through memory.  If result is not
15364         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15365         the parts directly to inner mode, if even that fails, use expansion
15366         through memory.
15367
15368         PR middle-end/83623
15369         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15370         check for bswap in mode rather than HImode and use that in expand_unop
15371         too.
15372 \f
15373 Copyright (C) 2018 Free Software Foundation, Inc.
15374
15375 Copying and distribution of this file, with or without modification,
15376 are permitted in any medium without royalty provided the copyright
15377 notice and this notice are preserved.