gcc/ChangeLog

   1 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
   2
   3         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
   4         for armv5te.
   5
   6 2018-05-28  Mark Wielaard  <mark@klomp.org>
   7
   8         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
   9         if it is an expression containing a minus sign.
  10
  11 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
  12
  13         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
  14
  15 2018-05-27  Paul Koning  <ni1d@arrl.net>
  16
  17         * config/pdp11/pdp11.md (truncsihi2): Remove.
  18
  19 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
  20             Chung-Ju Wu  <jasonwucj@gmail.com>
  21
  22         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
  23         implementation.
  24         (unaligned_store_dw): Ditto.
  25         * config/nds32/nds32-memory-manipulation.c
  26         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
  27         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
  28         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
  29         (emit_setmem_word_loop): Rename to ...
  30         (emit_setmem_doubleword_loop): ... this.
  31         (nds32_gen_dup_4_byte_to_word_value): New function.
  32         (nds32_gen_dup_8_byte_to_double_word_value): New function.
  33         (nds32_expand_setmem_loop): Refine implementation.
  34         (nds32_expand_setmem_loop_v3m): Ditto.
  35         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
  36         pattern.
  37
  38 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
  39
  40         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
  41
  42 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
  43
  44         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
  45         (nds32_init_machine_status): Initialize machine->attr_naked_p and
  46         machine->attr_no_prologue_p.
  47         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
  48         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
  49         (nds32_expand_epilogue): Consider attr_naked_p.
  50         (nds32_expand_epilogue_v3pop): Likewise.
  51         (nds32_can_use_return_insn): Likewise.
  52         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
  53         attr_no_prologue_p fields.
  54         * config/nds32/nds32.opt (mret-in-naked-func): New option.
  55
  56 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
  57
  58         PR target/85918
  59         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
  60         attributes.
  61         * config/i386/sse.md
  62         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
  63         Rename to ...
  64         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
  65         ... this.
  66         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
  67         Rename to ...
  68         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
  69         ... this.
  70         (*<floatsuffix>floatv2div2sf2): Rename to ...
  71         (*float<floatunssuffix>v2div2sf2): ... this.
  72         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
  73         (float<floatunssuffix>v2div2sf2_mask): ... this.
  74         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
  75         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
  76         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
  77         to ...
  78         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
  79         ... this.
  80         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
  81         Rename to ...
  82         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
  83         ... this.
  84         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
  85         Rename to ...
  86         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
  87         ... this.
  88         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
  89         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
  90         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
  91         gen_ufix_truncv8dfv8si2.
  92         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
  93         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
  94         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
  95         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
  96         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
  97         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
  98         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
  99         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
 100
 101 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
 102
 103         PR target/85900
 104         PR target/85345
 105         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
 106
 107 2018-05-25  Jim Wilson  <jimw@sifive.com>
 108
 109         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
 110         * config/riscv/riscv.c (struct machine_function): Add
 111         interrupt_handler_p and attribute_checked_p fields.
 112         (riscv_attribute_table): Add interrupt.
 113         (riscv_interrupt_type_p): New.
 114         (riscv_save_reg_p): Save extra regs for interrupt handler.
 115         (riscv_use_save_libcall): Return false  for interrupt handler.
 116         (riscv_first_stack_step): Add forward declaration.
 117         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
 118         for interrupt handler with large frame.  Use it for saved reg list.
 119         (riscv_expand_prologue): Move flag_stack_usage_info support to
 120         eliminate duplication.
 121         (riscv_expand_epilogue): Generate mret for interrupt handler.
 122         (riscv_epilogue_uses): New.
 123         (riscv_can_use_return_insn): Return false for interrupt handler.
 124         (riscv_function_ok_for_sibcall): Likewise.
 125         (riscv_set_current_function): Add interrupt handler support.
 126         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
 127         * config/riscv/riscv.md (UNSPECV_MRET): New.
 128         (GP_REGNUM): New.
 129         (riscv_frflags, riscv_fsflags): Use tab after opcode.
 130         (riscv_mret): New.
 131         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
 132
 133 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
 134
 135         PR tree-optimization/85712
 136         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
 137         this candidate has already been replaced in-situ by a copy.
 138
 139 2018-05-25  Jason Merrill  <jason@redhat.com>
 140
 141         PR c++/80485 - inline function non-zero address.
 142         * symtab.c (nonzero_address): Check DECL_COMDAT.
 143
 144 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
 145
 146         PR target/83628
 147         * config/alpha/alpha.md (ashlsi3): New insn pattern.
 148         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
 149         extension of SImode operation.  Use const123_operand predicate.
 150         (*saddsi_1): Remove.
 151         (*saddl_se_1): Ditto.
 152         (*ssubsi_1): Ditto.
 153         (*ssubl_se_1): Ditto.
 154         * config/alpha/predicates.md (const123_operand): New predicate.
 155         * config/alpha/constraints.md (P): Use IN_RANGE.
 156
 157 2018-05-25  Richard Biener  <rguenther@suse.de>
 158
 159         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
 160         defaulted to true.
 161         (ref_maybe_used_by_stmt_p): Likewise.
 162         (stmt_may_clobber_ref_p): Likewise.
 163         (stmt_may_clobber_ref_p_1): Likewise.
 164         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
 165         and pass it along.
 166         (ref_maybe_used_by_stmt_p): Likewise.
 167         (stmt_may_clobber_ref_p): Likewise.
 168         (stmt_may_clobber_ref_p_1): Likewise.
 169         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
 170         the alias oracle to disambiguate DRs with stmts DR analysis
 171         couldn't handle.
 172         (vect_analyze_data_refs): Do not give up on not analyzable
 173         DRs for BB vectorization.  Remove code truncating the dataref
 174         vector.
 175
 176 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
 177
 178         PR target/85832
 179         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
 180         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
 181         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
 182
 183 2018-05-25  Richard Biener  <rguenther@suse.de>
 184
 185         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
 186         function, combining stmt data ref gathering and fatal analysis
 187         parts.
 188         (vect_analyze_data_refs): Remove now redudnant code and simplify.
 189         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
 190         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
 191         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
 192         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
 193
 194 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
 195
 196         PR tree-optimization/85720
 197         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
 198         SCC if all partitions are builtins.
 199         (version_loop_by_alias_check): New parameter.  Generate cancelable
 200         runtime alias check if all partitions are builtins.
 201         (distribute_loop): Update call to above function.
 202
 203 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
 204
 205         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
 206         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
 207         (parm_default_def_partition_arg): Ditto.
 208         (set_parm_default_def_partition): Ditto.
 209         (get_parm_default_def_partitions): Ditto and make it static.
 210         (get_undefined_value_partitions): Ditto and make it static.
 211         (remove_ssa_form): Refactor call to init_var_map here.
 212         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
 213         computation for loop region.
 214         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
 215         (register_default_def): Delete.
 216         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
 217         (parm_default_def_partition_arg): Ditto.
 218         (set_parm_default_def_partition): Ditto.
 219         (get_parm_default_def_partitions): Ditto and make it static.
 220         (get_undefined_value_partitions): Ditto and make it static.
 221         (coalesce_with_default, coalesce_with_default): Update comment.
 222         (create_coalesce_list_for_region): New func factored out from
 223         create_outofssa_var_map.
 224         (populate_coalesce_list_for_outofssa): New func factored out from
 225         create_outofssa_var_map and coalesce_ssa_name.
 226         (create_outofssa_var_map): Delete.
 227         (coalesce_ssa_name): Refactor to support live range computation.
 228         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
 229         (get_parm_default_def_partitions): Delete.
 230         (get_undefined_value_partitions): Ditto.
 231         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
 232         computation for loop region.
 233         (new_tree_live_info, loe_visit_block): Ditto.
 234         (live_worklist, set_var_live_on_entry): Ditto.
 235         (calculate_live_on_exit, verify_live_on_entry): Ditto.
 236         * tree-ssa-live.h (struct _var_map): New fields.
 237         (init_var_map): Change decl.
 238         (region_contains_p): New.
 239
 240 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
 241
 242         * tree-ssa-live.h (live_merge_and_clear): Delete.
 243
 244 2018-05-25  Richard Biener  <rguenther@suse.de>
 245
 246         PR c++/85912
 247         * tree-dump.c (dequeue_and_dump): Remove access to removed
 248         operand 2 of a SWITCH_EXPR.
 249
 250 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
 251
 252         * doc/sourcebuild.texi (vect_double_cond_arith): Include
 253         multiplication and division.
 254         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
 255         (cond_udiv@var{m}, cond_umod@var{m}): Document.
 256         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
 257         (cond_udiv_optab, cond_umod_optab): New optabs.
 258         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
 259         (IFN_COND_RDIV): New internal functions.
 260         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
 261         TRUNC_MOD_EXPR and RDIV_EXPR.
 262         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
 263         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
 264         New unspecs.
 265         (SVE_INT_BINARY): Include mult.
 266         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
 267         (optab, sve_int_op): Handle mult.
 268         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
 269         UNSPEC_COND_DIV.
 270         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
 271         for SVE_INT_BINARY_SD.
 272
 273 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
 274
 275         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
 276         (optab, sve_int_op): Handle div and udiv.
 277         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
 278         for SVE_INT_BINARY_SD.
 279         (*<optab><mode>3): New insn for the same.
 280
 281 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
 282
 283         * tree-vect-patterns.c: Include predict.h.
 284         (vect_recog_divmod_pattern): Restrict check for division support
 285         to when optimizing for size.
 286
 287 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
 288
 289         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
 290         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
 291         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
 292         (gimple_match_op::set_op): Likewise.
 293         (gimple_resimplify4): Declare.
 294         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
 295         (expr::gen_transform): Likewise.
 296         (decision_tree::gen): Generate a simplification routine for 4 operands.
 297         * gimple-match-head.c (gimple_simplify): Add an overload for
 298         4 operands.  In the top-level function, handle up to 4 call
 299         arguments and call gimple_resimplify4.
 300         (gimple_resimplify4): New function.
 301         (build_call_internal): Pass a fourth operand.
 302         (maybe_push_to_seq): Likewise.
 303         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
 304         Fold VEC_COND_EXPRs of an operation and a default value into
 305         an IFN_COND_* function if possible.
 306         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
 307         New unspecs.
 308         (SVE_COND_FP_BINARY): Include them.
 309         (optab, sve_fp_op): Handle them.
 310         (SVE_INT_BINARY_REV): New code iterator.
 311         (SVE_COND_FP_BINARY_REV): New int iterator.
 312         (commutative): New int attribute.
 313         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
 314         Declare.
 315         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
 316         function.
 317         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
 318         (*cond_<optab><mode>): New patterns for reversed operands.
 319
 320 2018-05-25  Richard Biener  <rguenther@suse.de>
 321
 322         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
 323         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
 324         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
 325         (STMT_VINFO_GROUPED_ACCESS): Adjust.
 326         * tree-vect-data-refs.c (everywhere): Adjust users.
 327         * tree-vect-loop.c (everywhere): Likewise.
 328         * tree-vect-slp.c (everywhere): Likewise.
 329         * tree-vect-stmts.c (everywhere): Likewise.
 330         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
 331
 332 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
 333
 334         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
 335         Rename to...
 336         (gcc_cv_as_section_exclude): ... this.
 337         Try Solaris as #exclude syntax.
 338         * configure: Regenerate.
 339         * config.in: Regenerate.
 340         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
 341         SECTION_EXCLUDE.
 342         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
 343         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
 344
 345         * varasm.c (default_elf_asm_named_section): Don't check if
 346         HAVE_GAS_SECTION_EXCLUDE is defined.
 347
 348 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
 349
 350         * doc/md.texi: Update the documentation of the cond_* optabs
 351         to mention the new final operand.  Fix GET_MODE_NUNITS call.
 352         Describe the scalar case too.
 353         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
 354         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
 355         instead of 2.
 356         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
 357         (get_conditional_internal_fn): Update comment.
 358         * tree-vect-loop.c (vectorizable_reduction): Pass the original
 359         accumulator value as a final argument to conditional functions.
 360         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
 361         a define_expand and add an "else" operand.  Assert for now that
 362         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
 363         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
 364         (*cond_<optab><mode>): New patterns.
 365         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
 366         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
 367         (UNSPEC_COND_EOR): Delete.
 368         (optab): Remove associated mappings.
 369         (SVE_INT_BINARY): New code iterator.
 370         (sve_int_op): Remove int attribute and add "minus" to the code
 371         attribute.
 372         (SVE_COND_INT_OP): Delete.
 373         (SVE_COND_FP_OP): Rename to...
 374         (SVE_COND_FP_BINARY): ...this.
 375
 376 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
 377
 378         * optabs.c (can_reuse_operands_p): New function.
 379         (maybe_legitimize_operands): Try to reuse the results for
 380         earlier operands.
 381
 382 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
 383
 384         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
 385         Add {q} suffix to insn mnemonic.
 386
 387 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
 388
 389         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
 390         (msp430_warn_func_return): New.
 391
 392 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
 393
 394         * fold-const.c (tree_nonzero_bits): New function.
 395         * fold-const.h (tree_nonzero_bits): Likewise.
 396         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
 397         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
 398
 399 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
 400
 401         PR target/85900
 402         PR target/85345
 403         * varasm.c (assemble_alias): Check ifunc_resolver only on
 404         FUNCTION_DECL.
 405
 406 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
 407
 408         PR target/85903
 409         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
 410         when memory input operand is handled.
 411
 412 2018-05-24  Luis Machado  <luis.machado@linaro.org>
 413
 414         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
 415         global.
 416         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
 417
 418 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
 419
 420         * match.pd: Delay FMA folds until after vectorization.
 421
 422 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
 423
 424         PR target/83009
 425         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
 426         address check not strict.
 427
 428 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
 429
 430         * gimple-match.h (gimple_match_op): New class.
 431         (mprts_hook): Replace parameters with a gimple_match_op *.
 432         (maybe_build_generic_op): Likewise.
 433         (gimple_simplified_result_is_gimple_val): Replace parameters with
 434         a const gimple_match_op *.
 435         (gimple_simplify): Replace code_helper * and tree * parameters with
 436         a gimple_match_op * parameter.
 437         (gimple_resimplify1): Replace code_helper *, tree and tree *
 438         parameters with a gimple_match_op * parameter.
 439         (gimple_resimplify2): Likewise.
 440         (gimple_resimplify3): Likewise.
 441         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
 442         parameters with a gimple_match_op * parameter.
 443         * gimple-match-head.c (gimple_simplify): Change prototypes of
 444         auto-generated functions to take a gimple_match_op * instead of
 445         separate code_helper * and tree * parameters.  Make the same
 446         change in the top-level overload and update calls to the
 447         gimple_resimplify routines.  Update calls to the auto-generated
 448         functions and to maybe_push_res_to_seq in the publicly-facing
 449         operation-specific gimple_simplify overloads.
 450         (gimple_match_op::MAX_NUM_OPS): Define.
 451         (gimple_resimplify1): Replace rcode and ops with a single res_op
 452         parameter.  Update call to gimple_simplify.
 453         (gimple_resimplify2): Likewise.
 454         (gimple_resimplify3): Likewise.
 455         (mprts_hook): Replace parameters with a gimple_match_op *.
 456         (maybe_build_generic_op): Likewise.
 457         (build_call_internal): Replace type, nargs and ops with
 458         a gimple_match_op *.
 459         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
 460         with a single gimple_match_op *.  Update calls to mprts_hook,
 461         build_call_internal and gimple_simplified_result_is_gimple_val.
 462         Factor out code that is common to the tree_code and combined_fn cases.
 463         * genmatch.c (expr::gen_transform): Replace tem_code and
 464         tem_ops with a gimple_match_op called tem_op.  Update calls
 465         to the gimple_resimplify functions and maybe_push_res_to_seq.
 466         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
 467         res_ops.  Update call to the gimple_resimplify functions.
 468         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
 469         (decision_tree::gen): Make the functions take a gimple_match_op *
 470         called res_op instead of separate res_code and res_ops parameters.
 471         Update call accordingly.
 472         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
 473         and ops with a single res_op parameter.  Update calls to
 474         maybe_build_generic_op and maybe_push_res_to_seq.
 475         (fold_stmt_1): Update calls to gimple_simplify and
 476         replace_stmt_with_simplification.
 477         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
 478         and gimple_simplified_result_is_gimple_val.
 479         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
 480         gimple_simplify.
 481         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
 482         with a gimple_match_op *.
 483         (vn_nary_build_or_lookup): Likewise.  Update call to
 484         vn_nary_build_or_lookup_1.
 485         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
 486         gimple_match_op *.  Update calls to the gimple_resimplify routines
 487         and to gimple_simplified_result_is_gimple_val.
 488         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
 489         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
 490         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
 491         (visit_nary_op): Likewise.
 492         (visit_reference_op_load): Likewise.
 493
 494 2018-05-23  Luis Machado  <luis.machado@linaro.org>
 495
 496         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
 497         modifier for printing the step amount.
 498
 499 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
 500
 501         PR target/78849
 502         * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
 503         types.
 504
 505 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
 506
 507         * doc/sourcebuild.texi (Endianness): New subsubsection.
 508
 509 2018-05-23  Luis Machado  <luis.machado@linaro.org>
 510
 511         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
 512         <prefetch_dynamic_strides>: New const bool field.
 513         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
 514         prefetch_dynamic_strides.
 515         (exynosm1_prefetch_tune): Likewise.
 516         (thunderxt88_prefetch_tune): Likewise.
 517         (thunderx_prefetch_tune): Likewise.
 518         (thunderx2t99_prefetch_tune): Likewise.
 519         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
 520         false.
 521         (aarch64_override_options_internal): Update to set
 522         PARAM_PREFETCH_DYNAMIC_STRIDES.
 523         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
 524         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
 525         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
 526         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
 527         prefetch-dynamic-strides setting.
 528
 529 2018-05-23  Luis Machado  <luis.machado@linaro.org>
 530
 531         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
 532         <minimum_stride>: New const int field.
 533         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
 534         minimum_stride field defaulting to -1.
 535         (exynosm1_prefetch_tune): Likewise.
 536         (thunderxt88_prefetch_tune): Likewise.
 537         (thunderx_prefetch_tune): Likewise.
 538         (thunderx2t99_prefetch_tune): Likewise.
 539         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
 540         <default_opt_level>: Set to 3.
 541         (aarch64_override_options_internal): Update to set
 542         PARAM_PREFETCH_MINIMUM_STRIDE.
 543         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
 544         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
 545         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
 546         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
 547         stride is constant and is below the minimum stride threshold.
 548
 549 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 550
 551         * config/arm/arm-cpus.in (mode26): Delete.
 552         (armv4): Delete mode26 reference.
 553         * config/arm/arm.c (arm_configure_build_target): Delete use of
 554         isa_bit_mode26.
 555
 556 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
 557
 558         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
 559         New insn pattern.
 560         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
 561         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
 562         for non-SSE modes.
 563         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
 564         (floatunsdidf2): Ditto.
 565
 566 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
 567
 568         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
 569         (fixuns_trunc<mode>si2_avx512f): Ditto.
 570         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
 571         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
 572         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
 573
 574 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
 575
 576         PR rtl-optimization/79985
 577         * df-scan.c (df_insn_refs_collect): Remove special case for
 578         global registers and asm statements.
 579
 580 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
 581
 582         * extend.texi (Global Register Variables): Rewrite the bullet list.
 583         Note that the register is available for allocation. Note that access
 584         via inline asm must use constraints. Add note about async-signal
 585         handlers. Remove paragraph about automagic register selection.
 586
 587 2018-05-23  Richard Biener  <rguenther@suse.de>
 588
 589         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
 590         of fixed offset from memset VN.
 591
 592 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
 593
 594         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
 595         first_interp field.
 596         (alloc_cand_and_find_basis): Initialize first_interp field.
 597         (slsr_process_mul): Modify first_interp field.
 598         (slsr_process_add): Likewise.
 599         (slsr_process_cast): Modify first_interp field for each new
 600         interpretation.
 601         (slsr_process_copy): Likewise.
 602         (dump_candidate): Dump first_interp field.
 603         (replace_mult_candidate): Process all interpretations, not just
 604         subsequent ones.
 605         (replace_rhs_if_not_dup): Likewise.
 606         (replace_one_candidate): Likewise.
 607
 608 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
 609
 610         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
 611         Add new boolean.
 612         (aarch64_needs_frame_chain): New function.
 613         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
 614
 615 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
 616
 617         PR target/84882
 618         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
 619         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
 620         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
 621         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
 622         as true for strict-align.
 623         (aarch64_can_inline_p): Perform checks even when callee has no
 624         attributes to check for strict alignment.
 625         * doc/extend.texi (AArch64 Function Attributes): Document
 626         no-strict-align.
 627         * doc/invoke.texi: (AArch64 Options): Likewise.
 628
 629 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
 630
 631         PR tree-optimization/85853
 632         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
 633         the handling of the root of the node to...
 634         (vect_slp_analyze_node_operations_1): ...this new function,
 635         and run the whole thing with the child nodes' def types
 636         set according to their SLP node's def type.
 637
 638 2018-05-23  Richard Biener  <rguenther@suse.de>
 639
 640         PR middle-end/85874
 641         * tree-data-ref.c (create_runtime_alias_checks): Defer
 642         and ignore overflow warnings.
 643
 644 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
 645
 646         PR tree-optimization/85822
 647         * tree-vrp.c (is_masked_range_test): Fix handling of negative
 648         constants.
 649
 650 2018-05-23  Richard Biener  <rguenther@suse.de>
 651
 652         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
 653         memset constants via native_interpret_expr.
 654
 655 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
 656
 657         PR target/85345
 658         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
 659         attribute.
 660         (cgraph_node::create_alias): Likewise.
 661         (cgraph_node::get_availability): Check ifunc_resolver instead
 662         of looking up ifunc attribute.
 663         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
 664         * varasm.c (do_assemble_alias): Likewise.
 665         (assemble_alias): Likewise.
 666         (default_binds_local_p_3): Likewise.
 667         * cgraph.h (cgraph_node): Add ifunc_resolver.
 668         (cgraph_node::only_called_directly_or_aliased_p): Return false
 669         for IFUNC resolver.
 670         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
 671         attribute.
 672         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
 673         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
 674         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
 675         instead of looking up ifunc attribute.
 676
 677 2018-05-22  Luis Machado  <luis.machado@linaro.org>
 678
 679         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
 680
 681 2018-05-22  Martin Sebor  <msebor@redhat.com>
 682
 683         PR middle-end/85359
 684         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
 685         only when expasion succeeds.
 686         (expand_builtin_strcmp): Same.
 687         (expand_builtin_strncmp): Same.
 688
 689 2018-05-22  Martin Sebor  <msebor@redhat.com>
 690
 691         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
 692
 693 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
 694             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 695
 696         * config/aarch64/aarch64-ldpstp.md: Replace uses of
 697         aarch64_mem_pair_operand with memory_operand and delete operand swapping
 698         code.
 699         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
 700         Add check for legitimate_address.
 701         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
 702         (aarch64_swap_ldrstr_operands): New.
 703         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
 704         Define prototype.
 705
 706 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
 707             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 708
 709         * config/aarch64/aarch64.md: New patterns to generate stp
 710         and ldp.
 711         (store_pair_sw, store_pair_dw): New patterns to generate stp for
 712         single words and double words.
 713         (load_pair_sw, load_pair_dw): Likewise.
 714         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
 715         Delete.
 716         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
 717         Delete.
 718         * config/aarch64/aarch64-ldpstp.md: Modify peephole
 719         for different mode ldpstp and add peephole for merged zero stores.
 720         Likewise for loads.
 721         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
 722         Add size check.
 723         (aarch64_gen_store_pair): Rename calls to match new patterns.
 724         (aarch64_gen_load_pair): Rename calls to match new patterns.
 725         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
 726         (load_pair<DREG:mode><DREG2:mode>): ... This.
 727         (store_pair<mode>): Rename to...
 728         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
 729         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
 730         New mode iterators.
 731         (V_INT_EQUIV): Handle SImode.
 732         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
 733         New predicate.
 734
 735 2018-05-22  Martin Sebor  <msebor@redhat.com>
 736
 737         PR c/85623
 738         * calls.c (maybe_warn_nonstring_arg): Use string length to set
 739         or ajust the presumed bound on an operation to avoid unnecessary
 740         warnings.
 741
 742 2018-05-22  Martin Sebor  <msebor@redhat.com>
 743
 744         PR tree-optimization/85826
 745         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
 746         assuming that a DECL necesarily has a constant size.
 747
 748 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
 749
 750         PR middle-end/85862
 751         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
 752
 753 2018-05-22  Richard Biener  <rguenther@suse.de>
 754
 755         PR tree-optimization/85834
 756         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
 757         non-constant and non-zero memset arguments.
 758
 759 2018-05-22  Martin Liska  <mliska@suse.cz>
 760
 761         PR ipa/85607
 762         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
 763
 764 2018-05-22  Richard Biener  <rguenther@suse.de>
 765
 766         PR tree-optimization/85863
 767         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
 768         comparisons when vectype is specified.
 769         (vectorizable_condition): Do not specify vectype for
 770         vect_is_simple_cond when SLP vectorizing.
 771
 772 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
 773
 774         PR target/85657
 775         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
 776         define __ibm128 as long double.
 777         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
 778         as a distinct type when IEEE 128-bit support is enabled.
 779         (init_float128_ieee): Fix up conversions between IFmode and IEEE
 780         128-bit types to use the correct functions.
 781         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
 782         convert between 128-bit floating point types that have different
 783         modes but the same representation, instead of using gen_lowpart to
 784         makean alias.
 785         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
 786         KFmode.
 787         (IFKF_reg): New attributes to give the register constraints for
 788         IFmode and KFmode.
 789         (extend<mode>tf2_internal): New insns to mark an explicit
 790         conversion between 128-bit floating point types that have a
 791         different mode but share the same representation.
 792
 793 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
 794
 795         PR tree-optimization/85814
 796         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
 797         a null return from get_strinfo when unsharing the next
 798         strinfo in the chain.
 799
 800 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
 801
 802         PR gcc/84923
 803         * varasm.c (weak_finish): Clean up weak_decls.
 804
 805 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
 806
 807         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
 808         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
 809         UNSPEC_UADALP values.
 810         * config/aarch64/iterators.md (ABAL): New int iterator.
 811         (ABDL2): Likewise.
 812         (ADALP): Likewise.
 813         (sur): Add mappings for the above.
 814         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
 815         New define_insn.
 816         (aarch64_<sur>abal<mode>_4): Likewise.
 817         (aarch64_<sur>adalp<mode>_3): Likewise.
 818         (<sur>sadv16qi): New define_expand.
 819
 820 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
 821
 822         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
 823         (*movdf_internal): Ditto.
 824         (*rcpsf2_sse): Ditto.
 825         (*rsqrtsf2_sse): Ditto.
 826         (*sqrt<mode>2_sse): Ditto.
 827
 828 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
 829
 830         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
 831         eor3q<mode>4.
 832         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
 833         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
 834         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
 835         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
 836         vbcaxq_s64): New.
 837         * config/aarch64/arm_neon.h: Likewise.
 838         * config/aarch64/iterators.md (VQ_I): New.
 839
 840 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
 841
 842         * config.gcc: Add arc/t-multilib-linux to tmake_file for
 843         arc*-*-linux*.
 844         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
 845         MULTILIB_DIRNAMES
 846
 847 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
 848
 849         * config/nds32/constraints.md (S): New constraint.
 850         * config/nds32/nds32.md (call_internal): Use constraint S.
 851         (call_value_internal): Likewise.
 852         (sibcall_internal): Likewise.
 853         (sibcall_value_internal): Likewise.
 854
 855 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
 856             Chung-Ju Wu  <jasonwucj@gmail.com>
 857
 858         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
 859         into consideration.
 860
 861 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
 862             Chung-Ju Wu  <jasonwucj@gmail.com>
 863
 864         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
 865         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
 866         (nds32_rtx_costs_impl): Simplify.
 867         (nds32_address_cost_impl): Simplify.
 868         (nds32_init_rtx_costs): New function.
 869         (nds32_rtx_costs_speed_prefer): Likewise.
 870         (nds32_rtx_costs_size_prefer): Likewise.
 871         (nds32_address_cost_speed_prefer): Likewise.
 872         (nds32_address_cost_speed_fwprop): Likewise.
 873         (nds32_address_cost_size_prefer): Likewise.
 874         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
 875         * config/nds32/nds32.c (nds32_option_override): Use
 876         nds32_init_rtx_costs function.
 877
 878 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
 879
 880         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
 881         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
 882         (TARGET_PIPELINE_N8): Likewise.
 883         (TARGET_PIPELINE_N10): Likewise.
 884         (TARGET_PIPELINE_N13): Likewise.
 885         (TARGET_PIPELINE_GRAYWOLF): Likewise.
 886
 887 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
 888
 889         * config/nds32/nds32-fpu.md: Update copyright year.
 890
 891 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
 892
 893         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
 894
 895 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
 896
 897         * config/nds32/nds32.c
 898         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
 899         * config/nds32/nds32.opt (minline-asm-r15): New option.
 900
 901 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
 902
 903         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
 904         MASK_HW_ABS.
 905         * config/nds32/nds32.md (abssi2): New pattern.
 906
 907 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
 908
 909         * config/i386/i386.md (rex64namesuffix): New mode attribute.
 910         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
 911         Merge insn pattern from sse_cvtsi2ss<round_name> and
 912         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
 913         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
 914         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
 915         using SWI48 mode iterator.
 916         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
 917         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
 918         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
 919         pattern from sse_cvttss2si<round_saeonly_name>
 920         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
 921         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
 922         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
 923         using SWI48 mode iterator.
 924         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
 925         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
 926         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
 927         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
 928         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
 929         using SWI48 mode iterator.
 930         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
 931         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
 932         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
 933         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
 934         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
 935         SWI48 mode iterator.
 936         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
 937         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
 938         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
 939         pattern from sse_cvttsd2si<round_saeonly_name>
 940         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
 941
 942 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
 943
 944         * config/nds32/nds32-md-auxiliary.c
 945         (nds32_valid_smw_lwm_base_p): Refine.
 946         (nds32_output_smw_single_word): Refine.
 947         (nds32_output_smw_double_word): New.
 948         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
 949
 950 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
 951
 952         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
 953         (nds32_output_stack_pop): Refine.
 954         (nds32_expand_unaligned_load): Refine.
 955         (nds32_expand_unaligned_store): Refine.
 956
 957 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
 958             Chung-Ju Wu  <jasonwucj@gmail.com>
 959
 960         * config/nds32/constants.md: Add TP_REGNUM constant.
 961         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
 962         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
 963         UNSPEC_ADD32.
 964         * config/nds32/nds32-doubleword.md: Consider flag_pic.
 965         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
 966         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
 967         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
 968         and PIC code generation.
 969         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
 970         code generation.
 971         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
 972         optimization.
 973         * config/nds32/nds32.md: Support TLS and PIC.
 974         * config/nds32/nds32.c: Support TLS and PIC.
 975         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
 976         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
 977         predicate.
 978
 979 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
 980
 981         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
 982         mode with E_ prefix.
 983
 984 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
 985             Chung-Ju Wu  <jasonwucj@gmail.com>
 986
 987         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
 988         * config/nds32/nds32-md-auxiliary.c
 989         (symbolic_reference_mentioned_p): New.
 990         (nds32_legitimize_ict_address): New.
 991         (nds32_expand_ict_move): New.
 992         (nds32_indirect_call_referenced_p): New.
 993         (nds32_symbol_binds_local_p): Delete.
 994         (nds32_long_call_p): Modify.
 995         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
 996         * config/nds32/nds32-protos.h
 997         (symbolic_reference_mentioned_p): Declare.
 998         (nds32_legitimize_ict_address): Declare.
 999         (nds32_expand_ict_move): Declare.
1000         (nds32_indirect_call_referenced_p): Declare.
1001         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1002         (nds32_relax_group): Use nds32_ict_const_p as condition.
1003         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1004         (nds32_asm_file_start): Output ict_model directive in asm code.
1005         (nds32_legitimate_address_p): Consider indirect call.
1006         (nds32_print_operand): Consider indirect call.
1007         (nds32_print_operand_address): Consider indirect call.
1008         (nds32_insert_attributes): Handle "indirect_call" attribute.
1009         (TARGET_LEGITIMATE_ADDRESS_P): Define.
1010         (TARGET_LEGITIMATE_CONSTANT_P): Define.
1011         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1012         (TARGET_DELEGITIMIZE_ADDRESS): Define.
1013         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1014         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1015         (TARGET_ICT_MODEL_SMALL): Define.
1016         (TARGET_ICT_MODEL_LARGE): Define.
1017         * config/nds32/nds32.md (movsi): Consider ict model.
1018         (call, call_value): Consider ict model.
1019         (sibcall, sibcall_value): Consider ict model.
1020         * config/nds32/nds32.opt (mict-model): New option.
1021         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1022         model.
1023
1024 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
1025             Monk Chiang  <sh.chiang04@gmail.com>
1026             Jim Wilson <jimw@sifive.com>
1027
1028         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1029         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
1030         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1031         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1032         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
1033         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1034         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1035         compute save_libcall_adjustment properly.
1036         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1037         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1038         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1039         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1040         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1041         (ABI_SPEC): Handle mabi=ilp32e.
1042         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1043         (RVE): Add RVE mask.
1044         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1045         <-march>: Add rv32e as an example.
1046
1047 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
1048
1049         PR c++/82899
1050         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1051         (intra_create_variable_infos): Handle C++ constructors.
1052
1053 2018-05-18  Martin Liska  <mliska@suse.cz>
1054
1055         * passes.def: Remove a redundant pass.
1056
1057 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
1058
1059         PR bootstrap/85838
1060         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1061
1062 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1063
1064         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1065         (ARMv4): Update.
1066         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1067         (ARMv6m): Update.
1068         (armv2, armv2a, armv3, armv3m): Delete architectures.
1069         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1070         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1071         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1072         Delete cpus.
1073         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1074         (*mulsidi3adddi): Likewise.
1075         (mulsidi3): Likewise.
1076         (*mulsidi3_nov6): Likewise.
1077         (umulsidi3): Likewise.
1078         (umulsidi3_nov6): Likewise.
1079         (umaddsidi4): Likewise.
1080         (*umulsidi3adddi): Likewise.
1081         (smulsi3_highpart): Likewise.
1082         (*smulsi3_highpart_nov6): Likewise.
1083         (umulsi3_highpart): Likewise.
1084         (*umulsi3_highpart_nov6): Likewise.
1085         * config/arm/arm.h (arm_arch3m): Delete.
1086         * config/arm/arm.c (arm_arch3m): Delete.
1087         (arm_option_override_internal): Update armv3-related comment.
1088         (arm_configure_build_target): Delete use of isa_bit_mode32.
1089         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1090         (arm_rtx_costs_internal): Delete check of arm_arch3m.
1091         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1092         (mulsa3): Likewise.
1093         (mulusa3): Likewise.
1094         * config/arm/arm-protos.h (arm_arch3m): Delete.
1095         * config/arm/arm-tables.opt: Regenerate.
1096         * config/arm/arm-tune.md: Likewise.
1097         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1098         deleted architectures.
1099
1100 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1101
1102         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1103         (armv5t, armv5te): New features.
1104         (ARMv5, ARMv5e): Delete fgroups.
1105         (ARMv5t, ARMv5te): Adjust for above changes.
1106         (ARMv6m): Likewise.
1107         (armv5, armv5e): Delete arches.
1108         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1109         arm_arch5.
1110         (*call_reg_arm): Likewise.
1111         (*call_value_reg_armv5): Likewise.
1112         (*call_value_reg_arm): Likewise.
1113         (*call_symbol): Likewise.
1114         (*call_value_symbol): Likewise.
1115         (*sibcall_insn): Likewise.
1116         (*sibcall_value_insn): Likewise.
1117         (clzsi2): Likewise.
1118         (prefetch): Likewise.
1119         (define_split and define_peephole2 dependent on arm_arch5):
1120         Likewise.
1121         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1122         arm_arch5e.
1123         (TARGET_ARM_QBIT): Likewise.
1124         (TARGET_DSP_MULTIPLY): Likewise.
1125         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1126         (arm_arch5, arm_arch5e): Delete.
1127         (arm_arch5t, arm_arch5te): Declare.
1128         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1129         (arm_arch5t): Declare.
1130         (arm_option_reconfigure_globals): Update for the above.
1131         (arm_options_perform_arch_sanity_checks): Update comment, replace
1132         use of arm_arch5 with arm_arch5t.
1133         (use_return_insn): Likewise.
1134         (arm_emit_call_insn): Likewise.
1135         (output_return_instruction): Likewise.
1136         (arm_final_prescan_insn): Likewise.
1137         (arm_coproc_builtin_available): Likewise.
1138         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1139         arm_arch5e with arm_arch5t and arm_arch5te.
1140         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1141         (arm_arch5t, arm_arch5te): Declare.
1142         * config/arm/arm-tables.opt: Regenerate.
1143         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1144         * config/arm/t-multilib: Likewise.
1145         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1146         instead of arm_arch5.
1147         (*call_reg_thumb1): Likewise.
1148         (*call_value_reg_thumb1_v5): Likewise.
1149         (*call_value_reg_thumb1): Likewise.
1150         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1151         unreachable path.
1152         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1153
1154 2018-05-18  Martin Liska  <mliska@suse.cz>
1155
1156         PR gcov-profile/84846
1157         * doc/gcov.texi: Document -t option of gcov tool.
1158
1159 2018-05-18  Martin Liska  <mliska@suse.cz>
1160
1161         PR gcov-profile/84846
1162         * gcov.c (print_usage): Add new -t option.
1163         (process_args): Handle the option.
1164         (generate_results): Use stdout as output when requested by
1165         the option.
1166
1167 2018-05-18  Martin Liska  <mliska@suse.cz>
1168
1169         PR gcov-profile/84846
1170         * coverage.c (coverage_init): Write PWD to .gcno file.
1171         * doc/gcov.texi: Document how working directory is printed.
1172         * gcov-dump.c (dump_gcov_file): Print PWD.
1173         * gcov.c (output_intermediate_file): Likewise.
1174         (read_graph_file): Read PWD string.
1175         (output_lines): Print PWD.
1176
1177 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
1178
1179         PR middle-end/85817
1180         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1181         for retval and return false if all args to phi are zero.
1182
1183 2018-05-18  Richard Biener  <rguenther@suse.de>
1184
1185         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1186         method.
1187         (evrp_dom_walker::before_dom_children): Call it.
1188
1189 2018-05-18  Richard Biener  <rguenther@suse.de>
1190
1191         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1192         results when processing array refs with variable index.
1193
1194 2018-05-18  Toon Moene  <toon@moene.org>
1195
1196         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1197         directly after that of -floop-interchange. Indicate that both
1198         options are enabled by default when specifying -O3.
1199
1200 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1201
1202         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1203         iterator.  Delete separate integer-mode vec_set<mode> expander.
1204         (aarch64_simd_vec_setv2di): Delete.
1205         (vec_setv2di): Delete.
1206         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1207         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
1208         the "w, r" alternative.
1209
1210 2018-05-18  Martin Liska  <mliska@suse.cz>
1211
1212         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1213         * tree-pass.h (make_pass_lower_switch_O0): New function.
1214         * tree-switch-conversion.c (node_has_low_bound): Remove.
1215         (node_has_high_bound): Likewise.
1216         (node_is_bounded): Likewise.
1217         (class pass_lower_switch): Make it a template type and create
1218         two instances.
1219         (pass_lower_switch::execute): Add template argument.
1220         (make_pass_lower_switch): New function.
1221         (make_pass_lower_switch_O0): New function.
1222         (do_jump_if_equal): Remove.
1223         (emit_case_nodes): Simplify to just handle all 3 cases and leave
1224         all the hard work to tree optimization passes.
1225
1226 2018-05-18  Martin Liska  <mliska@suse.cz>
1227
1228         * dbgcnt.c (limit_low): Renamed from limit.
1229         (limit_high): New variable.
1230         (dbg_cnt_is_enabled): Check for upper limit.
1231         (dbg_cnt): Adjust dumping.
1232         (dbg_cnt_set_limit_by_index): Add new argument for high
1233         value.
1234         (dbg_cnt_set_limit_by_name): Likewise.
1235         (dbg_cnt_process_single_pair): Parse new format.
1236         (dbg_cnt_process_opt): Use strtok.
1237         (dbg_cnt_list_all_counters): Remove 'value' and add
1238         'limit_high'.
1239         * doc/invoke.texi: Document changes.
1240
1241 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
1242
1243         * doc/sourcebuild.texi (scalar_all_fma): Document.
1244         * tree.def (FMA_EXPR): Delete.
1245         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1246         * internal-fn.c (ternary_direct): New macro.
1247         (expand_ternary_optab_fn): Likewise.
1248         (direct_ternary_optab_supported_p): Likewise.
1249         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1250         * builtins.c (fold_builtin_fma): Delete.
1251         (fold_builtin_3): Don't call it.
1252         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1253         * expr.c (expand_expr_real_2): Likewise.
1254         * fold-const.c (operand_equal_p): Likewise.
1255         (fold_ternary_loc): Likewise.
1256         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1257         * gimple.c (DEFTREECODE): Likewise.
1258         * gimplify.c (gimplify_expr): Likewise.
1259         * optabs-tree.c (optab_for_tree_code): Likewise.
1260         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1261         * tree-eh.c (operation_could_trap_p): Likewise.
1262         (stmt_could_throw_1_p): Likewise.
1263         * tree-inline.c (estimate_operator_cost): Likewise.
1264         * tree-pretty-print.c (dump_generic_node): Likewise.
1265         (op_code_prio): Likewise.
1266         * tree-ssa-loop-im.c (stmt_cost): Likewise.
1267         * tree-ssa-operands.c (get_expr_operands): Likewise.
1268         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1269         * fold-const-call.h (fold_fma): Delete.
1270         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1271         CFN_FNMA and CFN_FNMS.
1272         (fold_fma): Delete.
1273         * genmatch.c (combined_fn): New enum.
1274         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1275         (commutative_op): New function.
1276         (commutate): Use it.  Handle more than 2 operands.
1277         (dt_operand::gen_gimple_expr): Use commutative_op.
1278         (parser::parse_expr): Allow :c to be used with non-binary
1279         operators if the commutative operand is known.
1280         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1281         CFN_FMS, CFN_FNMA and CFN_FNMS.
1282         (backprop::process_assign_use): Remove FMA_EXPR handling.
1283         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1284         (gen_hsa_fma): New function.
1285         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1286         IFN_FNMA and IFN_FNMS.
1287         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1288         * gimple-fold.h (follow_all_ssa_edges): Declare.
1289         * gimple-fold.c (follow_all_ssa_edges): New function.
1290         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1291         gimple_build interface and use follow_all_ssa_edges to fold the result.
1292         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1293         instead of checking for optabs directly.
1294         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1295         rather than FMA_EXPRs.
1296         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1297         call to IFN_FMA instead of an FMA_EXPR.
1298
1299 2018-05-17  Jim Wilson  <jimw@sifive.com>
1300
1301         * expr.c (do_tablejump): When converting index to Pmode, if we have a
1302         sign extended promoted subreg, and the range does not have the sign bit
1303         set, then do a sign extend.
1304
1305         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1306         test, check for sign extended subreg and/or constant operands, and
1307         do a sign extend in that case.
1308
1309 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
1310
1311         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1312         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1313         Add untyped.
1314         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1315         Change logics_shift_reg to logics_shift_imm.
1316         (thunderx2t99_fp_loadpair_basic): Delete.
1317         (thunderx2t99_fp_storepair_basic): Delete.
1318         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1319         (thunderx2t99_asimd_polynomial): Delete.
1320         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1321         and neon_fp_mul_d_scalar_q.
1322         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1323         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1324         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1325         (thunderx2t99_asimd_lut): Add missing tbl types.
1326         (thunderx2t99_asimd_ext): Delete.
1327         (thunderx2t99_asimd_load1_1_mult): Delete.
1328         (thunderx2t99_asimd_load1_2_mult): Delete.
1329         (thunderx2t99_asimd_load1_ldp): New.
1330         (thunderx2t99_asimd_load1): New.
1331         (thunderx2t99_asimd_load2): Add missing *load2* types.
1332         (thunderx2t99_asimd_load3): New.
1333         (thunderx2t99_asimd_load4): New.
1334         (thunderx2t99_asimd_store1_1_mult): Delete.
1335         (thunderx2t99_asimd_store1_2_mult): Delete.
1336         (thunderx2t99_asimd_store2_mult): Delete.
1337         (thunderx2t99_asimd_store2_onelane): Delete.
1338         (thunderx2t99_asimd_store_stp): New.
1339         (thunderx2t99_asimd_store1): New.
1340         (thunderx2t99_asimd_store2): New.
1341         (thunderx2t99_asimd_store3): New.
1342         (thunderx2t99_asimd_store4): New.
1343
1344 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
1345
1346         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1347         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
1348
1349 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
1350             Segher Boessenkool  <segher@kernel.crashing.org>
1351
1352         PR target/85698
1353         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1354         operand.
1355
1356 2018-05-17  Richard Biener  <rguenther@suse.de>
1357
1358         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1359         for pruning loop and prune defs feeding only already visited PHIs.
1360
1361 2018-05-17  Richard Biener  <rguenther@suse.de>
1362
1363         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1364
1365 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
1366             Richard Biener  <rguenther@suse.de>
1367
1368         PR tree-optimization/85793
1369         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1370         for VMAT_ELEMENTWISE.
1371
1372 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
1373
1374         * internal-fn.h (lookup_internal_fn): Declare
1375         * internal-fn.c (lookup_internal_fn): New function.
1376         * gimple.c (gimple_build_call_from_tree): Handle calls to
1377         internal functions.
1378         * gimple-pretty-print.c (dump_gimple_call): Print "." before
1379         internal function names.
1380         * tree-pretty-print.c (dump_generic_node): Likewise.
1381         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1382
1383 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
1384
1385         * gimple-fold.h (gimple_build): Make the function forms take
1386         combined_fn rather than built_in_function.
1387         (gimple_simplify): Likewise.
1388         * gimple-match-head.c (gimple_simplify): Likewise.
1389         * gimple-fold.c (gimple_build): Likewise.
1390         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1391         rather than gimple_build_call_internal.
1392         (get_initial_defs_for_reduction): Likewise.
1393         (vect_create_epilog_for_reduction): Likewise.
1394         (vectorizable_live_operation): Likewise.
1395
1396 2018-05-17  Martin Liska  <mliska@suse.cz>
1397
1398         * gimple-ssa-sprintf.c (format_directive): Do not use
1399         space in between 'G_' and '('.
1400
1401 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
1402
1403         PR target/85323
1404         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1405         even if the mask is not all ones.
1406
1407         PR target/85323
1408         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1409         vector.
1410         (ix86_gimple_fold_builtin): Likewise.
1411
1412         PR target/85323
1413         * config/i386/i386.c: Include tree-vector-builder.h.
1414         (ix86_vector_shift_count): New function.
1415         (ix86_fold_builtin): Fold shift builtins by scalar count.
1416         (ix86_gimple_fold_builtin): Likewise.
1417
1418         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1419         _mm512_setzero): New intrinsics.
1420
1421 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
1422             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1423
1424         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1425         code generation for cases where splatting a value is not useful.
1426         * simplify-rtx.c (simplify_ternary_operation): Simplify
1427         vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1428         mode to a vec_concat.
1429
1430 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
1431
1432         * config.gcc: Support "goldmont-plus".
1433         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1434         "goldmont-plus".
1435         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1436         PROCESSOR_GOLDMONT_PLUS.
1437         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1438         (processor_target_table): Add "goldmont-plus".
1439         (PTA_GOLDMONT_PLUS): Define.
1440         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1441         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1442         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1443         (fold_builtin_cpu): Add "goldmont-plus".
1444         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1445         (ix86_option_override_internal): Add "goldmont-plus".
1446         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1447         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1448         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1449         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1450
1451 2018-05-17  Richard Biener  <rguenther@suse.de>
1452
1453         PR tree-optimization/85757
1454         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1455         remove defs that only feed that PHI from further processing.
1456
1457 2018-05-16  Jim Wilson  <jimw@sifive.com>
1458
1459         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1460         asterisk to name.
1461         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1462
1463 2018-05-16  Mark Wielaard  <mark@klomp.org>
1464
1465         * dwarf2out.c (count_index_strings): New function.
1466         (output_indirect_strings): Call count_index_strings and generate
1467         header for dwarf_version >= 5.
1468
1469 2018-05-16  Mark Wielaard  <mark@klomp.org>
1470
1471         * dwarf2out.c (dwarf_FORM): New function.
1472         (set_indirect_string): Use dwarf_FORM.
1473         (reset_indirect_string): Likewise.
1474         (size_of_die): Likewise.
1475         (value_format): Likewise.
1476         (output_die): Likewise.
1477         (add_skeleton_AT_string): Likewise.
1478         (output_macinfo_op): Likewise.
1479         (index_string): Likewise.
1480         (output_index_string_offset): Likewise.
1481         (output_index_string): Likewise.
1482         (count_index_strings): Likewise.
1483
1484 2018-05-16  Carl Love  <cel@us.ibm.com>
1485
1486         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1487         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1488
1489 2018-05-16  Martin Jambor  <mjambor@suse.cz>
1490
1491         * ipa-prop.c (ipa_free_all_edge_args): Remove.
1492         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1493
1494 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
1495
1496         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1497         (fnma<mode>4): Likewise.
1498         (fms<mode>4): Likewise.
1499         (fnms<mode>4): Likewise.
1500         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1501         (aarch64_fnma<mode>4): Likewise.
1502         (aarch64_fms<mode>4): Likewise.
1503         (aarch64_fnms<mode>4): Likewise.
1504         (aarch64_fnmadd<mode>4): Likewise.
1505
1506 2018-05-16  Jason Merrill  <jason@redhat.com>
1507
1508         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
1509
1510 2018-05-16  Richard Biener  <rguenther@suse.de>
1511
1512         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1513         (dump_stmt_cost): Declare.
1514         (add_stmt_cost): Dump cost we add.
1515         (add_stmt_costs): New function.
1516         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1517         No longer exported.
1518         (vect_analyze_stmt): Adjust prototype.
1519         (vectorizable_condition): Likewise.
1520         (vectorizable_live_operation): Likewise.
1521         (vectorizable_reduction): Likewise.
1522         (vectorizable_induction): Likewise.
1523         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1524         cost vector to pass to vectorizable_ and record afterwards.
1525         (vect_model_reduction_cost): Take cost vector argument and adjust.
1526         (vect_model_induction_cost): Likewise.
1527         (vectorizable_reduction): Likewise.
1528         (vectorizable_induction): Likewise.
1529         (vectorizable_live_operation): Likewise.
1530         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1531         SLP_TREE_NUMBER_OF_VEC_STMTS.
1532         (vect_analyze_slp_cost_1): Remove.
1533         (vect_analyze_slp_cost): Likewise.
1534         (vect_slp_analyze_node_operations): Take visited args and
1535         a target cost vector.  Avoid processing already visited stmt sets.
1536         (vect_slp_analyze_operations): Use a local cost vector to gather
1537         costs and register those of non-discarded instances.
1538         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1539         (vect_schedule_slp_instance): Remove copying of
1540         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
1541         zero.
1542         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1543         adding cost.  Record cost entry location.
1544         (vect_prologue_cost_for_slp_op): Function to compute cost of
1545         a constant or invariant generated for SLP vect in the prologue,
1546         split out from vect_analyze_slp_cost_1.
1547         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
1548         (vect_model_promotion_demotion_cost): Likewise.
1549         (vect_model_store_cost): Likewise, make static.
1550         (vect_model_load_cost): Likewise.
1551         (vectorizable_bswap): Add cost vector arg and adjust.
1552         (vectorizable_call): Likewise.
1553         (vectorizable_simd_clone_call): Likewise.
1554         (vectorizable_conversion): Likewise.
1555         (vectorizable_assignment): Likewise.
1556         (vectorizable_shift): Likewise.
1557         (vectorizable_operation): Likewise.
1558         (vectorizable_store): Likewise.
1559         (vectorizable_load): Likewise.
1560         (vectorizable_condition): Likewise.
1561         (vectorizable_comparison): Likewise.
1562         (can_vectorize_live_stmts): Likewise.
1563         (vect_analyze_stmt): Likewise.
1564         (vect_transform_stmt): Adjust calls to vectorizable_*.
1565         * tree-vectorizer.c: Include gimple-pretty-print.h.
1566         (dump_stmt_cost): New function.
1567
1568 2018-05-16  Richard Biener  <rguenther@suse.de>
1569
1570         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1571         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1572         * tree-ssa-dse.c: Include tree-ssa-loop.h.
1573         (check_name): New callback.
1574         (dse_classify_store): Track cycles via a visited bitmap of PHI
1575         defs and simplify handling of in-loop and across loop dead stores
1576         and properly fail for loop-variant refs.  Handle byte-tracking with
1577         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1578         limiting the walk.
1579
1580 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
1581
1582         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1583         (vect_get_mask_type_for_stmt): Likewise.
1584         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1585         split out from...
1586         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
1587         to determine the statement's vector type and the vector type that
1588         should be used for calculating nunits.  Deal with cases in which
1589         the type has to be deferred.
1590         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1591         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1592         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1593         (vect_determine_vf_for_stmt): New functions, split out from...
1594         (vect_determine_vectorization_factor): ...here.
1595         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1596         (vect_get_mask_type_for_stmt): New functions, split out from
1597         vect_determine_vectorization_factor.
1598
1599 2018-05-16  Richard Biener  <rguenther@suse.de>
1600
1601         * tree-cfg.c (verify_gimple_assign_ternary): Properly
1602         verify the [VEC_]COND_EXPR embedded comparison.
1603
1604 2018-05-15  Martin Sebor  <msebor@redhat.com>
1605
1606         PR tree-optimization/85753
1607         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1608         RECORD_TYPE in addition to ARRAY_TYPE.
1609
1610 2018-05-15  Martin Sebor  <msebor@redhat.com>
1611
1612         PR middle-end/85643
1613         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1614
1615 2018-05-15  Richard Biener  <rguenther@suse.de>
1616
1617         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1618         add by_clobber_p one.  Change algorithm to collect all defs
1619         representing uses we need to walk and try reducing them to
1620         a single one before failing.
1621         (dse_dom_walker::dse_optimize_stmt): Adjust.
1622
1623 2018-05-13  Mark Wielaard  <mark@klomp.org>
1624
1625         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1626         (size_of_loc_descr): Likewise.
1627         (output_loc_operands): Likewise.
1628         (output_loc_operands_raw): Likewise.
1629         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1630         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1631         (hash_loc_operands): Likewise.
1632         (compare_loc_operands): Likewise.
1633
1634 2018-05-14  Mark Wielaard  <mark@klomp.org>
1635
1636         * dwarf2out.c (count_index_addrs): New function.
1637         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1638
1639 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
1640
1641         PR tree-optimization/83648
1642         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1643         return value as malloc candidate.
1644
1645 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
1646
1647         PR ipa/85734
1648         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1649         as true in call to suggest_attribute.
1650
1651 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
1652
1653         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1654         -mreadonly-in-sdata.
1655
1656 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1657
1658         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1659         New pattern.
1660         (aarch64_crypto_aesd_fused): Likewise.
1661
1662 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
1663
1664         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1665         (movsi_aarch64): Likewise.
1666         (load_pairsi): Likewise.
1667         (load_pairdi): Likewise.
1668         (store_pairsi): Likewise.
1669         (store_pairdi): Likewise.
1670         (load_pairsf): Likewise.
1671         (load_pairdf): Likewise.
1672         (store_pairsf): Likewise.
1673         (store_pairdf): Likewise.
1674         (zero_extend): Likewise.
1675         (trunc): Swap alternatives.
1676         (fcvt_target): Add '?' to prefer w over r.
1677
1678 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
1679
1680         PR target/85756
1681         * config/i386/i386.md: Disallow non-commutative arithmetics in
1682         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1683         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
1684         in the peephole2 before it.
1685
1686 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
1687
1688         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1689         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1690         (ix86_handle_option): Handle -mcldemote.
1691         * config.gcc: New header.
1692         * config/i386/cldemoteintrin.h: New file.
1693         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1694         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1695         -mcldemote.
1696         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1697         OPTION_MASK_ISA_CLDEMOTE.
1698         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1699         (ix86_valid_target_attribute_inner_p): Ditto.
1700         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1701         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1702         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1703         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1704         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1705         (cldemote): New.
1706         * config/i386/i386.opt: Add -mcldemote.
1707         * config/i386/x86intrin.h: New header.
1708         * doc/invoke.texi: Add -mcldemote.
1709
1710 2018-05-14  Richard Biener  <rguenther@suse.de>
1711
1712         * doc/match-and-simplify.texi: Adjust :s documentation.
1713
1714 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
1715
1716         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1717         intended memcpy size.
1718         (REORDER_45): Likewise.
1719
1720 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
1721
1722         * sort.cc: New file.
1723         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1724         * vec.c (qsort_chk): Use gcc_qsort.
1725         * Makefile.in (OBJS-libcommon): Add sort.o.
1726         (build/sort.o): New target.  Use it...
1727         (BUILD_RTL): ... here, and...
1728         (build/gencfn-macros): ... here, and...
1729         (build/genmatch): ... here.
1730
1731 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
1732             Chung-Ju Wu  <jasonwucj@gmail.com>
1733
1734         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1735         * config/nds32/nds32-graywolf.md: New file.
1736         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1737         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1738         pipeline.
1739         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1740         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1741         * config/nds32/nds32.md (pipeline_model): Add graywolf.
1742         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1743         * config/nds32/pipelines.md: Include n15 settings.
1744
1745 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
1746             Chung-Ju Wu  <jasonwucj@gmail.com>
1747
1748         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1749         * config/nds32/nds32-n13.md: New file.
1750         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1751         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1752         pipeline.
1753         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1754         * config/nds32/nds32.md (pipeline_model): Add n13.
1755         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1756         * config/nds32/pipelines.md: Include n13 settings.
1757
1758 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
1759             Chung-Ju Wu  <jasonwucj@gmail.com>
1760
1761         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1762         * config/nds32/nds32-n10.md: New file.
1763         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1764         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1765         pipeline.
1766         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1767         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1768         * config/nds32/nds32.md (pipeline_model): Add n10.
1769         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1770         * config/nds32/pipelines.md: Include n10 settings.
1771
1772 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
1773             Kito Cheng  <kito.cheng@gmail.com>
1774             Chung-Ju Wu  <jasonwucj@gmail.com>
1775
1776         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1777         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1778         Add enum values for DSP extension instructions.
1779         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1780         New constraints.
1781         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1782         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1783         New code iterators.
1784         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1785         * config/nds32/nds32-dspext.md: New file for DSP implementation.
1786         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1787         * config/nds32/nds32-intrinsic.md: Likewise.
1788         * config/nds32/nds32_intrinsic.h: Likewise.
1789         * config/nds32/nds32-md-auxiliary.c: Likewise.
1790         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1791         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1792         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1793         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1794         * config/nds32/nds32-protos.h: New declarations for DSP extension.
1795         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1796         TYPE_DMAC in switch statement.
1797         * config/nds32/nds32.c: New checking and implementation for DSP
1798         extension instructions.
1799         * config/nds32/nds32.h: Likewise.
1800         * config/nds32/nds32.md: Likewise.
1801         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1802         * config/nds32/predicates.md: Implement new predicates for DSP
1803         extension.
1804
1805 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
1806
1807         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1808         Reformat alternatives and attributes so it is easier to identify
1809         which constraints/attributes go with which instruction.
1810         (mov<mode>_hardfloat32, FMOVE64): Likewise.
1811         (mov<mode>_softfloat32, FMOVE64): Likewise.
1812         (mov<mode>_hardfloat64, FMOVE64): Likewise.
1813         (mov<mode>_softfloat64, FMOVE64): Likewise.
1814
1815 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
1816
1817         * doc/extend.texi (PowerPC Built-in Functions): Rename this
1818         subsection.
1819         (Basic PowerPC Built-in Functions): The new name of the
1820         subsection previously known as "PowerPC Built-in Functions".
1821         (Basic PowerPC Built-in Functions Available on all Configurations):
1822         New subsubsection.
1823         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1824         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1825         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1826         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1827
1828 2018-05-11  Martin Jambor  <mjambor@suse.cz>
1829
1830         PR ipa/85655
1831         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1832         single const.
1833
1834 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
1835
1836         PR target/85733
1837         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1838
1839 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
1840
1841         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1842         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1843         (ix86_handle_option): Handle -mwaitpkg.
1844         * config.gcc: New header.
1845         * config/i386/cpuid.h (bit_WAITPKG): New bit.
1846         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1847         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1848         function type.
1849         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1850         OPTION_MASK_ISA_WAITPKG.
1851         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1852         (ix86_option_override_internal): Add PTA_WAITPKG.
1853         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1854         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1855         IX86_BUILTIN_TPAUSE.
1856         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1857         __builtin_ia32_umwait and __builtin_ia32_tpause.
1858         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1859         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1860         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1861         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1862         UNSPECV_TPAUSE): New.
1863         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1864         * config/i386/i386.opt: Add -mwaitpkg.
1865         * config/i386/waitpkgintrin.h: New file.
1866         * config/i386/x86intrin.h: New header.
1867         * doc/invoke.texi: Add -mwaitpkg.
1868
1869 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
1870
1871         PR target/85606
1872         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1873         equivalent.
1874         (cortex-m0): Use armv6s-m isa.
1875         (cortex-m0plus): Likewise.
1876         (cortex-m1): Likewise.
1877         (cortex-m0.small-multiply): Likewise.
1878         (cortex-m0plus.small-multiply): Likewise.
1879         (cortex-m1.small-multiply): Likewise.
1880
1881 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
1882             Jakub Jelinek  <jakub@redhat.com>
1883
1884         PR tree-optimization/85692
1885         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1886         source permute as well.
1887
1888 2018-05-11  Martin Liska  <mliska@suse.cz>
1889
1890         PR sanitizer/85556
1891         * doc/extend.texi: Document LLVM style format for no_sanitize
1892         attribute.
1893
1894 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
1895
1896         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1897         mode_supports_vsx_dform_quad to mode_supports_dq_form.
1898         (mode_supports_vsx_dform_quad): Likewise.
1899         (mode_supports_vmx_dform): Move these functions to be next to the
1900         other mode_supports functions.
1901         (mode_supports_dq_form): Likewise.
1902         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1903         mode_supports_dq_form.
1904         (reg_offset_addressing_ok_p): Likewise.
1905         (offsettable_ok_by_alignment): Likewise.
1906         (rs6000_legitimate_offset_address_p): Likewise.
1907         (legitimate_lo_sum_address_p): Likewise.
1908         (rs6000_legitimize_address): Likewise.
1909         (rs6000_legitimize_reload_address): Likewise.
1910         (rs6000_secondary_reload_inner): Likewise.
1911         (rs6000_preferred_reload_class): Likewise.
1912         (rs6000_output_move_128bit): Likewise.
1913
1914 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
1915
1916         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1917         Generate SImode target register for null target.
1918         <case IX86_BUILTIN_XGETBV>: Ditto.
1919         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1920         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1921
1922 2018-05-10  Carl Love  <cel@us.ibm.com>
1923
1924         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1925         dcbtt and dcbtstt if operands[2] is 0.
1926
1927 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
1928
1929         PR target/85693
1930         * config/i386/sse.md (usadv64qi): New expander.
1931
1932 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
1933
1934         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1935         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1936         -maltivec=be support.
1937         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1938         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1939         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1940         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1941         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1942         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1943         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1944         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1945         altivec_vsumsws): Adjust.
1946         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1947         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1948         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1949         support.
1950         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1951         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1952         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1953         (altivec_lve<VI_char>x): Delete expand.
1954         (*altivec_lve<VI_char>x_internal): Rename to...
1955         (altivec_lve<VI_char>x): ... this.
1956         (altivec_lvxl_<mode>): Delete expand.
1957         (*altivec_lvxl_<mode>_internal): Rename to ...
1958         (altivec_lvxl_<mode>): ... this.
1959         (altivec_stvxl_<mode>): Delete expand.
1960         (*altivec_stvxl_<mode>_internal): Rename to ...
1961         (altivec_stvxl_<mode>): ... this.
1962         (altivec_stve<VI_char>x): Delete expand.
1963         (*altivec_stve<VI_char>x_internal): Rename to ...
1964         (altivec_stve<VI_char>x): ... this.
1965         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1966         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1967         reduc_plus_scal_<mode>): Adjust.
1968         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1969         comment.
1970         (rs6000_cpu_cpp_builtins): Adjust.
1971         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1972         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1973         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1974         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1975         -maltivec=be support.
1976         (rs6000_split_vec_extract_var): Adjust.
1977         (rs6000_split_v4si_init): Adjust.
1978         (swap_selector_for_mode): Delete.
1979         (altivec_expand_lvx_be, altivec_expand_stvx_be,
1980         altivec_expand_stvex_be): Delete.
1981         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1982         -maltivec=be support.
1983         (rs6000_gimple_fold_builtin): Ditto.
1984         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1985         Adjust.
1986         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1987         (TARGET_DIRECT_MOVE_64BIT): Adjust.
1988         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1989         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1990         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1991         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1992         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1993         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1994         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1995         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1996         anonymous split): Adjust.
1997         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1998         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1999
2000 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
2001
2002         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2003         when --with-gxx-include-dir is also specified.
2004         * configure: Regenerate.
2005
2006 2018-05-09  Jim Wilson  <jimw@sifive.com>
2007
2008         PR target/84797
2009         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2010         * config/riscv/t-withmultilib: New.
2011         * config/riscv/withmultilib.h: New.
2012         * doc/install.texi: Document RISC-V --with-multilib-list support.
2013
2014 2018-05-09  Richard Biener  <rguenther@suse.de>
2015
2016         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2017         vector.
2018         (vect_bb_vectorization_profitable_p): Adjust.  Compute
2019         actual scalar cost using the cost vector and the add_stmt_cost
2020         machinery.
2021
2022 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
2023
2024         PR rtl-optimization/85645
2025         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2026         in the REG_CFA_REGISTER note for LR, don't leave it empty.
2027
2028 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
2029
2030         PR rtl-optimization/85645
2031         * shrink-wrap.c (spread_components): Return a boolean saying if
2032         anything was changed.
2033         (try_shrink_wrapping_separate): Iterate spread_components until
2034         nothing changes anymore.
2035
2036 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
2037
2038         PR rtl-optimization/85645
2039         * regrename.c (build_def_use): Also kill the chains that include the
2040         destination of a REG_CFA_REGISTER note.
2041
2042 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
2043
2044         PR rtl-optimization/85645
2045         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2046         insn that has a REG_CFA_REGISTER note.
2047
2048 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
2049
2050         * cfgexpand.c (expand_clobber): New function.
2051         (expand_gimple_stmt_1): Use it.
2052         * tree-vect-stmts.c (vect_clobber_variable): New function,
2053         split out from...
2054         (vectorizable_simd_clone_call): ...here.
2055         (vectorizable_store): Emit a clobber either side of an
2056         IFN_STORE_LANES sequence.
2057         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2058
2059 2018-05-09  Tom de Vries  <tom@codesourcery.com>
2060
2061         PR target/85626
2062         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2063         (define_insn "trap_if_false"): Add exit after trap.
2064
2065 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
2066
2067         PR rtl-optimization/85638
2068         * bb-reorder.c: Include common/common-target.h.
2069         (create_forwarder_block): New function extracted from...
2070         (fix_up_crossing_landing_pad): ...here.  Rename into...
2071         (dw2_fix_up_crossing_landing_pad): ...this.
2072         (sjlj_fix_up_crossing_landing_pad): New function.
2073         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2074         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2075         from both partitions and exit the loop after one iteration.
2076
2077 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
2078
2079         Revert:
2080         * doc/extend.texi (PowerPC Built-in Functions): Rename this
2081         subsection.
2082         (Basic PowerPC Built-in Functions): The new name of the
2083         subsection previously known as "PowerPC Built-in Functions".
2084         (Basic PowerPC Built-in Functions Available on all Configurations):
2085         New subsubsection.
2086         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2087         subsubsection.
2088         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2089         subsubsection.
2090         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2091         subsubsection.
2092         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2093         subsubsection.
2094
2095 2018-05-08  Jim Wilson  <jimw@sifive.com>
2096
2097         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2098         (LD_EMUL_SUFFIX): New.
2099         (LINK_SPEC): Use it.
2100
2101 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
2102
2103         * doc/extend.texi (PowerPC Built-in Functions): Rename this
2104         subsection.
2105         (Basic PowerPC Built-in Functions): The new name of the
2106         subsection previously known as "PowerPC Built-in Functions".
2107         (Basic PowerPC Built-in Functions Available on all Configurations):
2108         New subsubsection.
2109         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2110         subsubsection.
2111         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2112         subsubsection.
2113         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2114         subsubsection.
2115         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2116         subsubsection.
2117
2118 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
2119
2120         PR target/85683
2121         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2122         after cmpelim optimization.
2123
2124 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
2125
2126         * config.gcc: Support "goldmont".
2127         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2128         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2129         PROCESSOR_GOLDMONT.
2130         * config/i386/i386.c (m_GOLDMONT): Define.
2131         (processor_target_table): Add "goldmont".
2132         (PTA_GOLDMONT): Define.
2133         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2134         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2135         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2136         (fold_builtin_cpu): Add "goldmont".
2137         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2138         (ix86_option_override_internal): Add "goldmont".
2139         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2140         (processor_type): Add PROCESSOR_GOLDMONT.
2141         * config/i386/i386.md: Add CPU "glm".
2142         * config/i386/glm.md: New file.
2143         * config/i386/x86-tune.def: Add m_GOLDMONT.
2144         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2145
2146 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
2147
2148         PR target/85572
2149         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2150         E_V4DImode.
2151         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2152         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
2153         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
2154
2155         PR target/85317
2156         * config/i386/i386.c (ix86_fold_builtin): Handle
2157         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2158
2159         PR target/85480
2160         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2161         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2162
2163 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
2164
2165         PR target/85658
2166         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2167         (check_arch): Likewise.
2168         (check_fpu): Return the result rather than printing it.
2169         (end arch): Fix operator precedence.
2170         (end cpu): Likewise.
2171         (END): Print the result from check_fpu.
2172
2173 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
2174             Alan Hayward  <alan.hayward@arm.com>
2175             David Sherwood  <david.sherwood@arm.com>
2176
2177         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2178         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2179         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2180         (*fcmuo<mode>_and): New patterns.
2181
2182 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
2183
2184         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2185         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2186         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2187         (cmp_op, sve_imm_con): New code attributes.
2188         (SVE_COND_INT_CMP, imm_con): Delete.
2189         (cmp_op): Remove above unspecs from int attribute.
2190         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2191         to...
2192         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
2193         comparison-specific unspecs.
2194         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2195         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2196         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2197         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2198         (*vec_fcm<cmp_op><mode>): Rename to...
2199         (*fcm<cmp_op><mode>): ...this and adjust likewise.
2200         (*vec_fcmuo<mode>): Rename to...
2201         (*fcmuo<mode>): ...this and adjust likewise.
2202         (*pred_fcm<cmp_op><mode>): New pattern.
2203         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2204         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2205         functions.
2206         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2207         and UNORDERED.
2208         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2209         (aarch64_emit_sve_predicated_cond): New function.
2210         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2211         (aarch64_emit_unspec_cond_or): Replace with...
2212         (aarch64_emit_sve_or_conds): ...this new function.  Use
2213         aarch64_emit_sve_ptrue_op for the individual comparisons and
2214         aarch64_emit_binop to OR them together.
2215         (aarch64_emit_inverted_unspec_cond): Replace with...
2216         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
2217         aarch64_emit_sve_ptrue_op for the comparison and
2218         aarch64_emit_unop to invert the result.
2219         (aarch64_expand_sve_vec_cmp_float): Update after the above
2220         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
2221
2222 2018-05-07  Nathan Sidwell  <nathan@acm.org>
2223
2224         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2225         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2226         (Backwards Compatibility): Likewise.
2227
2228 2018-05-07  Luis Machado  <luis.machado@linaro.org>
2229
2230         PR bootstrap/85681
2231         Revert:
2232         2018-05-07  Luis Machado  <luis.machado@linaro.org>
2233
2234         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2235         <prefetch_dynamic_strides>: New const bool field.
2236         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2237         prefetch_dynamic_strides.
2238         (exynosm1_prefetch_tune): Likewise.
2239         (thunderxt88_prefetch_tune): Likewise.
2240         (thunderx_prefetch_tune): Likewise.
2241         (thunderx2t99_prefetch_tune): Likewise.
2242         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2243         (aarch64_override_options_internal): Update to set
2244         PARAM_PREFETCH_DYNAMIC_STRIDES.
2245         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2246         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2247         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2248         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2249         prefetch-dynamic-strides setting.
2250
2251         2018-05-07  Luis Machado  <luis.machado@linaro.org>
2252
2253         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2254         <minimum_stride>: New const int field.
2255         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2256         minimum_stride field.
2257         (exynosm1_prefetch_tune): Likewise.
2258         (thunderxt88_prefetch_tune): Likewise.
2259         (thunderx_prefetch_tune): Likewise.
2260         (thunderx2t99_prefetch_tune): Likewise.
2261         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2262         (aarch64_override_options_internal): Update to set
2263         PARAM_PREFETCH_MINIMUM_STRIDE.
2264         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2265         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2266         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2267         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2268         stride is constant and is below the minimum stride threshold.
2269
2270 2018-05-07  Luis Machado  <luis.machado@linaro.org>
2271
2272         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2273         to 512.
2274
2275 2018-05-07  Luis Machado  <luis.machado@linaro.org>
2276
2277         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2278         <prefetch_dynamic_strides>: New const bool field.
2279         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2280         prefetch_dynamic_strides.
2281         (exynosm1_prefetch_tune): Likewise.
2282         (thunderxt88_prefetch_tune): Likewise.
2283         (thunderx_prefetch_tune): Likewise.
2284         (thunderx2t99_prefetch_tune): Likewise.
2285         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2286         (aarch64_override_options_internal): Update to set
2287         PARAM_PREFETCH_DYNAMIC_STRIDES.
2288         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2289         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2290         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2291         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2292         prefetch-dynamic-strides setting.
2293
2294 2018-05-07  Luis Machado  <luis.machado@linaro.org>
2295
2296         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2297         <minimum_stride>: New const int field.
2298         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2299         minimum_stride field.
2300         (exynosm1_prefetch_tune): Likewise.
2301         (thunderxt88_prefetch_tune): Likewise.
2302         (thunderx_prefetch_tune): Likewise.
2303         (thunderx2t99_prefetch_tune): Likewise.
2304         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2305         (aarch64_override_options_internal): Update to set
2306         PARAM_PREFETCH_MINIMUM_STRIDE.
2307         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2308         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2309         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2310         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2311         stride is constant and is below the minimum stride threshold.
2312
2313 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
2314
2315         PR c++/85659
2316         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2317         the type is addressable.  Don't force op into register if it has
2318         BLKmode.
2319
2320 2018-05-05  Roland McGrath  <mcgrathr@google.com>
2321
2322         PR other/77609
2323         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2324         any section for which we don't know a specific type it should have,
2325         regardless of name.  Previously this was done only for the exact
2326         names ".init_array", ".fini_array", and ".preinit_array".
2327         (default_elf_asm_named_section): Add comment about
2328         relationship with default_section_type_flags and SECTION_NOTYPE.
2329         (get_section): Don't consider it a type conflict if one side has
2330         SECTION_NOTYPE and the other doesn't, as long as neither has the
2331         SECTION_BSS et al used in the default_section_type_flags logic.
2332
2333 2018-05-05  Tom de Vries  <tom@codesourcery.com>
2334
2335         PR target/85653
2336         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2337         (workaround_barsyncs): New function.
2338         (nvptx_reorg): Use workaround_barsyncs.
2339         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2340         (define_expand "nvptx_membar_cta"): New define_expand.
2341         (define_insn "*nvptx_membar_cta"): New insn.
2342
2343 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
2344
2345         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2346         To improve optimization opportunities.
2347         * builtin-types.def: The new needed builtin types for the above.
2348
2349 2018-05-04  Richard Biener  <rguenther@suse.de>
2350
2351         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2352         * gimple-ssa-store-merging.c
2353         (imm_store_chain_info::output_merged_store): Remove redundant create,
2354         release split_store vector contents on failure.
2355         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2356         scalar stmt vector on cache hit.
2357
2358 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
2359
2360         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2361         Xilinx FP support.
2362         * config.gcc (powerpc-xilinx-eabi*): Remove.
2363         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2364         support.
2365         (fusion_addis_mem_combo_load): Ditto.
2366         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2367         FP support.
2368         (rs6000_cpu_cpp_builtins): Ditto.
2369         * config/rs6000/rs6000-linux.c
2370         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2371         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2372         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2373         support.
2374         (rs6000_setup_reg_addr_masks): Ditto.
2375         (rs6000_init_hard_regno_mode_ok): Ditto.
2376         (rs6000_option_override_internal): Ditto.
2377         (legitimate_lo_sum_address_p): Ditto.
2378         (rs6000_legitimize_address): Ditto.
2379         (rs6000_legitimize_reload_address): Ditto.
2380         (rs6000_legitimate_address_p): Ditto.
2381         (abi_v4_pass_in_fpr): Ditto.
2382         (setup_incoming_varargs): Ditto.
2383         (rs6000_gimplify_va_arg): Ditto.
2384         (rs6000_split_multireg_move): Ditto.
2385         (rs6000_savres_strategy): Ditto.
2386         (rs6000_emit_prologue_components): Ditto.
2387         (rs6000_emit_epilogue_components): Ditto.
2388         (rs6000_emit_prologue): Ditto.
2389         (rs6000_emit_epilogue): Ditto.
2390         (rs6000_elf_file_end): Ditto.
2391         (rs6000_function_value): Ditto.
2392         (rs6000_libcall_value): Ditto.
2393         * config/rs6000/rs6000.h: Ditto.
2394         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2395         (TARGET_MINMAX): ... this.  New.
2396         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2397         * config/rs6000/rs6000.md: Remove Xilinx FP support.
2398         (*movsi_internal1_single): Delete.
2399         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2400         mfpu=, mxilinx-fpu): Delete.
2401         * config/rs6000/singlefp.h: Delete.
2402         * config/rs6000/sysv4.h: Remove Xilinx FP support.
2403         * config/rs6000/t-rs6000: Ditto.
2404         * config/rs6000/t-xilinx: Delete.
2405         * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2406         * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2407         (VStype_simple): Delete.
2408         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2409         * config/rs6000/xfpu.h: Delete.
2410         * config/rs6000/xfpu.md: Delete.
2411         * config/rs6000/xilinx.h: Delete.
2412         * config/rs6000/xilinx.opt: Delete.
2413         * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2414         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2415
2416 2018-05-04  Tom de Vries  <tom@codesourcery.com>
2417
2418         PR libgomp/85639
2419         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2420         if ignore == 0.
2421
2422 2018-05-04  Richard Biener  <rguenther@suse.de>
2423
2424         PR middle-end/85627
2425         * tree-complex.c (update_complex_assignment): We are always in SSA form.
2426         (expand_complex_div_wide): Likewise.
2427         (expand_complex_operations_1): Likewise.
2428         (expand_complex_libcall): Preserve EH info of the original stmt.
2429         (tree_lower_complex): Handle removed blocks.
2430         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2431         on complex multiplication and division libcall builtins.
2432
2433 2018-05-04  Richard Biener  <rguenther@suse.de>
2434
2435         PR middle-end/85574
2436         * fold-const.c (negate_expr_p): Restrict negation of operand
2437         zero of a division to when we know that can happen without
2438         overflow.
2439         (fold_negate_expr_1): Likewise.
2440
2441 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
2442
2443         PR libstdc++/85466
2444         * real.h (real_nextafter): Declare.
2445         * real.c (real_nextafter): New function.
2446         * fold-const-call.c (fold_const_nextafter): New function.
2447         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2448         CASE_CFN_NEXTTOWARD.
2449         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2450         even when arg1_mode is different from arg0_mode.
2451
2452 2018-05-03  Nathan Sidwell  <nathan@acm.org>
2453
2454         * doc/extend.texi (Deprecated Features): Remove
2455         -ffriend-injection.
2456         (Backwards Compatibility): Likewise.
2457         * doc/invoke.texi (C++ Language Options): Likewise.
2458         (C++ Dialect Options): Likewise.
2459
2460 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
2461
2462         PR target/85530
2463         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2464         _mm512_mask_mullox_epi64): New intrinsics.
2465
2466 2018-05-03  Tom de Vries  <tom@codesourcery.com>
2467
2468         PR testsuite/85106
2469         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2470         dump files): Add offload-tree.
2471
2472 2018-05-03  Richard Biener  <rguenther@suse.de>
2473
2474         PR tree-optimization/85615
2475         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2476         to loops not nested in BBs loop father to avoid creating multi-entry
2477         loops.
2478
2479 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2480
2481         PR tree-optimization/70291
2482         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2483         arguments.  Change return type to tree.  Emit libcall as a new
2484         statement rather than replacing existing one when inplace_p is true.
2485         (expand_complex_multiplication_components): New function.
2486         (expand_complex_multiplication): Expand floating-point complex
2487         multiplication using the above.
2488         (expand_complex_division): Rename inner_type parameter to type.
2489         Update expand_complex_libcall call-site.
2490         (expand_complex_operations_1): Update expand_complex_multiplication
2491         and expand_complex_division call-sites.
2492
2493 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
2494
2495         PR target/85582
2496         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2497         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2498         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2499         the highest significant bit of the shift count mask is clear.  In
2500         check whether and[sq]i3 is needed verify that all significant bits
2501         of the shift count other than the highest are set.
2502
2503 2018-05-02  Tom de Vries  <tom@codesourcery.com>
2504
2505         PR libgomp/82428
2506         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2507         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2508         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2509         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2510         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2511         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2512         __builtin_goacc_parlevel_size.
2513
2514 2018-05-02  Richard Biener  <rguenther@suse.de>
2515
2516         PR tree-optimization/85597
2517         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2518         do not use split vect_get_vec_defs call but call vect_get_slp_defs
2519         directly.
2520
2521 2018-05-02  Tom de Vries  <tom@codesourcery.com>
2522
2523         PR testsuite/85106
2524         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2525         dump files): Add ltrans-tree.
2526
2527 2018-05-02  Tom de Vries  <tom@codesourcery.com>
2528
2529         PR testsuite/85106
2530         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2531         dump files): Add wpa-ipa.
2532
2533 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
2534
2535         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
2536         powerpc*-*-linux*paired* target.
2537         * config/rs6000/750cl.h: Delete.
2538         * config/rs6000/paired.h: Delete.
2539         * config/rs6000/paired.md: Delete.
2540         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2541         float support.
2542         * config/rs6000/rs6000-builtin.def: Remove paired float support.
2543         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2544         comment.  Remove paired float support.
2545         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2546         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2547         VECTOR_PAIRED.
2548         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2549         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2550         declarations.
2551         * config/rs6000/rs6000.c: Remove paired float support.
2552         (paired_expand_vector_init, paired_expand_vector_move,
2553         paired_emit_vector_compare, paired_emit_vector_cond_expr,
2554         (paired_expand_lv_builtin, paired_expand_stv_builtin,
2555         paired_expand_builtin, paired_expand_predicate_builtin,
2556         paired_init_builtins): Delete.
2557         * config/rs6000/rs6000.h: Remove paired float support.
2558         * config/rs6000/rs6000.md: Remove paired float support.
2559         (move_from_CR_ov_bit): Delete.
2560         * config/rs6000/rs6000.opt (mpaired): Delete.
2561         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2562         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2563
2564 2018-05-02  Richard Biener  <rguenther@suse.de>
2565
2566         PR middle-end/85567
2567         * gimplify.c (gimplify_save_expr): When in SSA form allow
2568         SAVE_EXPRs to compute to SSA vars.
2569
2570 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
2571
2572         PR target/85582
2573         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2574         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2575         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2576         clobber operands[2], instead use a new pseudo.  Formatting fixes.
2577
2578 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
2579
2580         PR tree-optimization/85586
2581         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2582         exit early for statements in the same group if the accesses are
2583         not strided.
2584
2585 2018-05-02  Tom de Vries  <tom@codesourcery.com>
2586
2587         PR lto/85451
2588         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2589         error message.
2590
2591 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
2592
2593         PR tree-optimization/85143
2594         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2595
2596 2018-05-01  Tom de Vries  <tom@codesourcery.com>
2597
2598         PR lto/85451
2599         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2600         not found" error message.
2601
2602 2018-05-01  Tom de Vries  <tom@codesourcery.com>
2603
2604         PR other/83786
2605         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2606         * vec.c (test_ordered_remove_if): New function.
2607         (vec_c_tests): Call test_ordered_remove_if.
2608         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2609         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2610         * tree-vect-patterns.c (vect_pattern_recog_1): Use
2611         VEC_ORDERED_REMOVE_IF.
2612
2613 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
2614
2615         PR tree-optimization/82665
2616         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2617         pointer subtraction where arguments come from a memchr call.
2618
2619 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
2620
2621         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2622         --push-state --as-needed and --pop-state instead of --as-needed and
2623         --no-as-needed if ld supports it.
2624         * configure: Regenerated.
2625
2626         PR web/85578
2627         * doc/install.texi2html: Replace _002d with - and _002a with * in
2628         generated html files using sed.
2629
2630 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
2631
2632         PR c++/85523
2633         * gcc-rich-location.c (blank_line_before_p): New function.
2634         (use_new_line): New function.
2635         (gcc_rich_location::add_fixit_insert_formatted): New function.
2636         * gcc-rich-location.h
2637         (gcc_rich_location::add_fixit_insert_formatted): New function.
2638
2639 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
2640
2641         * selftest.c (assert_streq): Rename "expected" and "actual" to
2642         "val1" and "val2".  Extend NULL-handling to cover both inputs
2643         symmetrically, while still requiring both to be non-NULL for a pass.
2644         * selftest.h (assert_streq): Rename "expected" and "actual" to
2645         "val1" and "val2".
2646         (ASSERT_EQ): Likewise.
2647         (ASSERT_EQ_AT): Likewise.
2648         (ASSERT_KNOWN_EQ): Likewise.
2649         (ASSERT_KNOWN_EQ_AT): Likewise.
2650         (ASSERT_NE): Likewise.
2651         (ASSERT_MAYBE_NE): Likewise.
2652         (ASSERT_MAYBE_NE_AT): Likewise.
2653         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
2654         the assertion to pass.
2655         (ASSERT_STREQ_AT): Likewise.
2656
2657 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
2658
2659         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2660         interaction with -pie.
2661
2662 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
2663
2664         * selftest.h: Fix alphabetization of per-source-file selftest
2665         declarations.
2666
2667 2018-04-30  Jason Merrill  <jason@redhat.com>
2668
2669         PR c++/61982 - dead stores to destroyed objects.
2670         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2671         of clobber.
2672
2673 2018-04-30  Jason Merrill  <jason@redhat.com>
2674
2675         * tree.c (build_clobber): New.
2676         * tree.h: Declare it.
2677         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2678
2679 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
2680
2681         * diagnostic-show-locus.c (layout::layout): Update for
2682         location_get_source_line returning a char_span.
2683         (struct char_span): Move to input.h.
2684         (struct correction): Update for fields in char_span becoming
2685         private.
2686         (struct source_line): Update for location_get_source_line
2687         returning a char_span.
2688         (layout::print_line): Likewise.
2689         * edit-context.c (edited_file::print_content): Likewise.
2690         (edited_file::print_diff_hunk): Likewise.
2691         (edited_file::print_run_of_changed_lines): Likewise.
2692         (edited_file::get_num_lines): Likewise.
2693         (edited_line::edited_line): Likewise.
2694         * final.c (asm_show_source): Likewise.
2695         * input.c (location_get_source_line): Convert return type
2696         from const char * to char_span, losing the final "line_len"
2697         param.
2698         (dump_location_info): Update for the above.
2699         (get_substring_ranges_for_loc): Likewise.  Use a char_span
2700         when handling the literal within the line.
2701         (test_reading_source_line): Update for location_get_source_line
2702         returning a char_span.
2703         * input.h (class char_span): Move here from
2704         diagnostic-show-locus.c, converting from a struct to a class.
2705         Make data members private.
2706         (char_span::operator bool): New.
2707         (char_span::length): New.
2708         (char_span::get_buffer): New.
2709         (char_span::operator[]): New.
2710         (char_span::subspan): Make const.
2711         (char_span::xstrdup): New.
2712         (location_get_source_line): Convert return type from const char *
2713         to char_span, losing the final "line_size" param.
2714
2715 2018-04-30  Jan Hubicka  <jh@suse.cz>
2716
2717         * lto-wrapper.c (ltrans_priorities): New static var.
2718         (cmp_priority): New.
2719         (run_gcc): Read priorities and if doing parallel build order
2720         the Makefile by them.
2721
2722 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
2723
2724         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2725
2726 2018-04-30  Richard Biener  <rguenther@suse.de>
2727
2728         * tree-cfg.c (verify_address): Remove base argument, add
2729         flag whether to check TREE_ADDRESSABLE and do that.
2730         (verify_expr): Remove.
2731         (verify_types_in_gimple_reference): Add pieces from verify_expr.
2732         (verify_gimple_assign_single): Likewise.
2733         (verify_gimple_switch): Likewise.
2734         (verify_expr_location_1): Dereference tp once.  Add (disabled)
2735         piece from verify_expr.
2736         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2737
2738 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
2739
2740         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2741
2742 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
2743
2744         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2745         (small_data_pattern): Likewise.
2746         (arc_rewrite_small_data): Likewise.
2747         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2748         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2749         (get_symbol_alignment): New function.
2750         (legitimate_small_data_address_p): Likewise.
2751         (legitimate_scaled_address): Update, call
2752         legitimate_small_data_address_p.
2753         (output_sdata): New static variable.
2754         (arc_print_operand): Update how we handle small data operands.
2755         (arc_print_operand_address): Likewise.
2756         (arc_legitimate_address_p): Update, use
2757         legitimate_small_data_address_p.
2758         (arc_rewrite_small_data_p): Remove.
2759         (arc_rewrite_small_data_1): Likewise.
2760         (arc_rewrite_small_data): Likewise.
2761         (small_data_pattern): Likewise.
2762         (compact_sda_memory_operand): Update to use
2763         legitimate_small_data_address_p and get_symbol_alignment.
2764         (prepare_move_operands): Don't rewite sdata pattern.
2765         (prepare_extend_operands): Remove.
2766         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2767         pattern.
2768         (zero_extendqisi2): Likewise.
2769         (zero_extendhisi2): Likewise.
2770         (extendqihi2): Likewise.
2771         (extendqisi2): Likewise.
2772         (extendhisi2): Likewise.
2773         (addsi3): Likewise.
2774         (subsi3): Likewise.
2775         (andsi3): Likewise.
2776         * config/arc/constraints.md (Usd): Change it to memory constraint.
2777
2778 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
2779
2780         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2781         as source of std instructions.
2782         * config/arc/arc.md (movsi_insn): Update pattern predicate to
2783         allow 6-bit constants as source for store instructions.
2784         (movdi_insn): Update instruction pattern to allow 6-bit constants
2785         as source for store instructions.
2786
2787 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
2788
2789         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2790
2791 2018-04-30  Nathan Sidwell  <nathan@acm.org>
2792             Sandra Loosemore <sandra@codesourcery.com>
2793
2794         * dumpfile.c (dump_open): Allow '-' for stdout.
2795         * doc/invoke.texi (Developer Options): Document dump filename
2796         determination early.  Document stdin/stdout selection.
2797
2798 2018-04-30 Andrew Sadek  <andrew.sadek.se@gmail.com>
2799
2800         Microblaze Target: PIC data text relative
2801
2802         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2803         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2804         Add declaration.
2805         * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2806         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2807         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2808         New addressing mode for data-text relative position indepenedent code.
2809         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2810         'ADDRESS_SYMBOLIC_TXT_REL'.
2811         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2812         (microblaze_legitimate_pic_operand): Exclude function calls from
2813         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2814         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2815         addresses cases.
2816         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2817         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2818         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2819         for 'address + offset'.
2820         (microblaze_expand_prologue): Add new function prologue call for
2821         'r20' assignation.
2822         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2823         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2824         table in case of TARGET_PIC_DATA_TEXT_REL.
2825         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2826         * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2827         Add new macros 'UNSPEC_TEXT',
2828         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2829         + exclude function calls from 'UNSPEC_PLT' in case of data text
2830         relative mode.
2831         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2832         new target hook for generating address diff vector tables in case of
2833         flag_pic.
2834         * doc/tm.texi : Regenerate.
2835         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2836         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2837         of addr diff vector generation.
2838         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2839         target hook definition.
2840         * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2841         Add default function for generate_pic_addr_diff_vec -> flag_pic.
2842         * doc/invoke.texi (Add new pic option): Add new microblaze pic
2843         option for data text relative.
2844
2845 2018-04-30  Richard Biener  <rguenther@suse.de>
2846
2847         * tree-chrec.h (evolution_function_is_constant_p): Remove
2848         redundant check.
2849         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2850
2851 2018-04-30  Richard Biener  <rguenther@suse.de>
2852
2853         PR bootstrap/85571
2854         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2855
2856 2018-04-30  Richard Biener  <rguenther@suse.de>
2857
2858         PR tree-optimization/28364
2859         PR tree-optimization/85275
2860         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2861         copying first exit test.
2862
2863 2018-04-28  Mark Wielaard  <mark@klomp.org>
2864
2865         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2866         dwarf_version >= 5.
2867         (dwarf_AT): Handle DW_AT_addr_base.
2868         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2869
2870 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
2871
2872         PR target/84431
2873         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2874         (*ashl<dwi>3_doubleword_mask_1): Ditto.
2875         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2876         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2877
2878 2018-04-28  Richard Biener  <rguenther@suse.de>
2879
2880         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2881         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2882         to reflect use.  Only add interesting stmts.
2883
2884 2018-04-27  Martin Jambor  <mjambor@suse.cz>
2885
2886         PR ipa/85549
2887         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2888         the jump function allows for passing through aggregate values.
2889
2890 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
2891
2892         * input.h (in_system_header_at): Convert from macro to inline
2893         function.
2894         (from_macro_expansion_at): Likewise.
2895         (from_macro_definition_at): Likewise.
2896
2897 2018-04-27  Jeff Law  <law@redhat.com>
2898
2899         * config.gcc: Mark tile* targets as deprecated/obsolete.
2900
2901 2018-04-27  Richard Biener  <rguenther@suse.de>
2902
2903         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2904         fix for ILP32.
2905
2906 2018-04-27  Richard Biener  <rguenther@suse.de>
2907
2908         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2909
2910 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
2911
2912         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2913         with Yd constraint. Set "preferred_for_speed" attribute from
2914         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2915         with Yd constraint.
2916         (*movdi_internal): Ditto.
2917         (movti_interunit splitters): Remove
2918         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2919         (movdi_interunit splitters): Ditto.
2920         * config/i386/constraints.md (Ye): Remove.
2921         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2922
2923 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2924
2925         PR target/85512
2926         * config/aarch64/constraints.md (Usg): Limit to 31.
2927         (Usj): Limit to 63.
2928
2929 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
2930
2931         PR tree-optimization/85529
2932         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2933         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2934         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2935         zero extension or masking of the MSB bit.
2936         (optimize_range_tests): Add FIRST_BB argument, pass it through
2937         to optimize_range_tests_var_bound.
2938         (maybe_optimize_range_tests, reassociate_bb): Adjust
2939         optimize_range_tests callers.
2940
2941 2018-04-26  Richard Biener  <rguenther@suse.de>
2942             Jakub Jelinek  <jakub@redhat.com>
2943
2944         * cgraph.h (symbol_table): Just declare debug method here.
2945         * symtab.c (symbol_table::debug): Define.
2946
2947 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
2948
2949         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2950
2951 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
2952
2953         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2954         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2955         (*movdi_internal): Substitute Yi and Yj constraint with x
2956         and Ym and Yn constraint with y constraint.  Update "isa"
2957         attribute and set "preferred_for_speed" attribute from
2958         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2959         (*movsi_internal): Ditto.
2960         (*movdf_internal): Ditto.
2961         (*movsf_internal): Ditto.
2962         (*zero_extendsidi2): Ditto.
2963         * config/i386/sse.md (vec_set<mode>_0): Ditto.
2964         (sse2_loadld): Ditto.
2965         (*vec_extract<ssevecmodelower>_0): Ditto.
2966         (*vec_extractv4si_0_zext_sse4): Ditto.
2967         (vec_concatv2di): Ditto.
2968         (*vec_dup<mode>): Ditto.
2969         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2970         * config/i386/constraints.md (Yi): Remove.
2971         (Yj): Remove.
2972         (Ym): Remove.
2973         (Yn): Remove.
2974
2975 2018-04-26  Nathan Sidwell  <nathan@acm.org>
2976
2977         * dumpfile.c (dump_open): New.
2978         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2979         (dump_finish): Detect stdio/stderr by value not name.
2980
2981 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
2982
2983         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2984
2985 2018-04-26  Tom de Vries  <tom@codesourcery.com>
2986
2987         PR target/84952
2988         * config/nvptx/nvptx.c (verify_neutering_jumps)
2989         (verify_neutering_labels): New function
2990         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2991
2992 2018-04-26  Tom de Vries  <tom@codesourcery.com>
2993
2994         PR target/84025
2995         * config/nvptx/nvptx.c (needs_neutering_p): New function.
2996         (nvptx_single): Use needs_neutering_p to skip over insns that do not
2997         need neutering.
2998
2999 2018-04-26  Richard Biener <rguenther@suse.de>
3000             Tom de Vries  <tom@codesourcery.com>
3001
3002         PR lto/85422
3003         * lto-streamer-out.c (output_function): Fixup loops if required to match
3004         discovery done in the reader.
3005
3006 2018-04-26  Richard Biener  <rguenther@suse.de>
3007
3008         PR tree-optimization/85116
3009         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3010         have a loop exit from the single latch predecessor.  Remove
3011         case of header with just condition.
3012         (ch_base::copy_headers): Exclude infinite loops from any
3013         processing.
3014         (pass_ch::execute): Record exits.
3015
3016 2018-04-26  Richard Biener  <rguenther@suse.de>
3017
3018         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3019         prologue cost vector and pass it to vect_get_load_cost.
3020         (vect_get_peeling_costs_all_drs): Likewise.
3021         (vect_peeling_hash_get_lowest_cost): Likewise.
3022         (vect_enhance_data_refs_alignment): Likewise.
3023
3024 2018-04-26  Richard Biener  <rguenther@suse.de>
3025
3026         PR middle-end/85450
3027         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3028         checking of integer<->pointer conversions.
3029         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3030         sign-/zero-extending pointer types.
3031         (expand_omp_for_static_chunk): Likewise.
3032
3033 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
3034             Jean Lee  <xiaoyur347@gmail.com>
3035
3036         * config/mips/mips.c (mips_asan_shadow_offset): New function.
3037         (TARGET_ASAN_SHADOW_OFFSET): Define.
3038         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3039         true for -fsanitize=address.
3040
3041 2018-04-25  Mark Wielaard  <mark@klomp.org>
3042
3043         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3044         shorter ones.
3045
3046 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
3047
3048         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3049         than "alu", remove explicit "memory" and "imm_disp" attributes.
3050         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3051
3052         PR middle-end/85414
3053         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3054         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3055         gen_lowpart_no_emit.
3056
3057 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
3058
3059         PR target/85473
3060         * config/i386/i386.c (ix86_expand_builtin): Change memory
3061         operand to XI, extend p0 to Pmode.
3062         * config/i386/i386.md: Change unspec volatile and operand
3063         1 mode to XI, change operand 0 mode to P.
3064
3065 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
3066
3067         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3068         GET_MODE_MASK before any checking.
3069         (nds32_can_use_bset_p): Likewise.
3070         (nds32_can_use_btgl_p): Likewise.
3071
3072 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
3073
3074         * config/nds32/nds32-doubleword.md: New define_split pattern for
3075         illegal register number.
3076
3077 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
3078
3079         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3080
3081 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
3082
3083         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3084
3085 2018-04-25  Richard Biener  <rguenther@suse.de>
3086
3087         * lto-streamer.h (LTO_major_version): Bump to 8.
3088
3089 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
3090
3091         * BASE-VER: Set to 9.0.0.
3092
3093 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
3094
3095         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3096         in __abskf2 and __powikf2.
3097
3098 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3099
3100         PR target/85512
3101         * config/aarch64/constraints.md (Usg, Usj): New constraints.
3102         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3103         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3104         Use the above on operand 2.  Reindent.
3105         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3106
3107 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
3108
3109         PR target/85485
3110         * common/config/i386/i386-common.c (ix86_handle_option): Don't
3111         handle OPT_mcet.
3112         * config/i386/i386.opt (mcet): Removed.
3113         * doc/install.texi: Remove -mcet documentation.
3114         * doc/invoke.texi: Likewise.
3115
3116 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
3117
3118         PR target/85485
3119         * doc/install.texi: Remove -mcet from bootstrap-cet.
3120
3121 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
3122
3123         PR target/85511
3124         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3125         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3126         if TARGET_64BIT.
3127
3128         PR target/85503
3129         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3130         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3131         containing a CONST_VECTOR.
3132
3133 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
3134
3135         * doc/install.texi: Update newlib dependency for nvptx.
3136
3137 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
3138
3139         PR target/85508
3140         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3141         instead of INTVAL when shifting x left.
3142
3143 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
3144
3145         PR tree-optimization/85478
3146         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3147         vect_grouped_store_supported for single element vectors.
3148
3149 2018-04-24  Richard Biener  <rguenther@suse.de>
3150
3151         PR target/85491
3152         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3153         load cost increase to the case of non-constant step.
3154
3155 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
3156
3157         PR target/84828
3158         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3159         destination if any_malformed_asm.
3160
3161 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
3162
3163         PR middle-end/85496
3164         * expr.c (store_field): In the bitfield case, if the value comes from
3165         a function call and is returned in registers by means of a PARALLEL,
3166         do not change the mode of the temporary unless BLKmode and VOIDmode.
3167
3168 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
3169
3170         PR rtl-optimization/85423
3171         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3172         dependencies to debug insns when the previous insn is non-debug.
3173
3174 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
3175
3176         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3177         enums into a single definition.
3178         (fls): Fix predicates and printing.
3179         (seti): Likewise.
3180
3181 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
3182
3183         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3184         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3185         and short u6 immediate.
3186         (check_if_valid_sleep_operand): Remove.
3187         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3188
3189 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
3190
3191         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3192         flag_always_save_lp condition.
3193         * config/nds32/nds32.opt (malways-save-lp): New option.
3194
3195 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
3196
3197         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3198         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3199         * config/nds32/nds32.h
3200         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3201         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3202
3203 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
3204
3205         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3206         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3207
3208 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
3209             Chung-Ju Wu  <jasonwucj@gmail.com>
3210
3211         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3212         Declare.
3213         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3214         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3215
3216 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
3217
3218         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3219
3220 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
3221
3222         * config/nds32/nds32-protos.h (nds32_data_alignment,
3223         nds32_local_alignment): Declare.
3224         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3225         nds32_local_alignment): New functions.
3226         (TARGET_CONSTANT_ALIGNMENT): Define.
3227         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3228
3229 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
3230
3231         * config/nds32/nds32.c
3232         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3233         (TARGET_MODES_TIEABLE_P): Likewise.
3234
3235 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
3236
3237         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3238         level Ofast and Og.
3239
3240 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
3241             Chung-Ju Wu  <jasonwucj@gmail.com>
3242
3243         * config/nds32/constants.md (unspec_volatile_element): Add enum values
3244         for unaligned access.
3245         * config/nds32/nds32-intrinsic.c: Implementation of expanding
3246         unaligned access.
3247         * config/nds32/nds32-intrinsic.md: Likewise.
3248         * config/nds32/nds32_intrinsic.h: Likewise.
3249         * config/nds32/nds32.h (nds32_builtins): Likewise.
3250         * config/nds32/nds32.opt (munaligned-access): New option.
3251         * config/nds32/nds32.c (nds32_asm_file_start): Display
3252         flag_unaligned_access status.
3253
3254 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
3255
3256         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3257         -mno-relax is present.
3258         * config/riscv/linux.h (LINK_SPEC): Ditto.
3259
3260 2018-04-20  Martin Sebor  <msebor@redhat.com>
3261
3262         PR c/85365
3263         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3264         for null pointers.
3265         (gimple_fold_builtin_stxcpy_chk): Same.
3266         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3267
3268 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
3269
3270         PR target/85456
3271         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3272         __powikf2 when long double is IEEE 128-bit.
3273
3274 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
3275
3276         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3277         step to make sure stack always aligned.
3278
3279 2018-04-20  Carl Love  <cel@us.ibm.com>
3280
3281         PR target/83402
3282         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3283         size check for arg0.
3284
3285 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
3286             Tom de Vries  <tom@codesourcery.com>
3287
3288         PR target/85445
3289         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3290         Emit insns for calls too.
3291         (nvptx_find_par): Always look for worker-level predecessor insn.
3292         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
3293         calls.
3294         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3295         (nvptx_process_pars): Propagate frames for calls.
3296
3297 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
3298
3299         PR target/85469
3300         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3301         Removed.
3302         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3303         (ix86_handle_option): Don't handle OPT_mibt.
3304         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3305         __SHSTK__.
3306         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3307         has_ibt and ibt.
3308         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3309         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3310         (ix86_target_macros): Define __CET__ with flag_cf_protection
3311         for -fcf-protection.
3312         * config/i386/i386.c (isa2_opts): Remove -mibt.
3313         * config/i386/i386.h (TARGET_IBT): Removed.
3314         (TARGET_IBT_P): Likewise.
3315         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3316         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3317         * config/i386/i386.opt (mcet): Update help message.
3318         (mshstk): Likewise.
3319         (mibt): Removed.
3320         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
3321         -mcet as an alias for -mshstk.
3322
3323 2018-04-20  Richard Biener <rguenther@suse.de>
3324
3325         PR middle-end/85475
3326         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3327         complexity by forcing a single use of the multiply operand.
3328
3329 2018-04-20  Martin Jambor  <mjambor@suse.cz>
3330
3331         ipa/85449
3332         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3333         recursion dependency to only apply to non-clones.
3334
3335 2018-04-20  Martin Jambor  <mjambor@suse.cz>
3336
3337         ipa/85447
3338         * ipa-cp.c (create_specialized_node): Check that clones of
3339         self-recursive edges exist during IPA-CP.
3340
3341 2018-04-19  Toon Moene  <toon@moene.org>
3342
3343         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3344         by -O3.
3345
3346 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
3347
3348         PR tree-optimization/85467
3349         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3350         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
3351         VECTOR_CST element to type.
3352
3353 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
3354
3355         PR target/85397
3356         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3357         * config/i386/i386.md (builtin_setjmp_setup): Removed.
3358         (builtin_longjmp): Likewise.
3359         (save_stack_nonlocal): New pattern.
3360         (restore_stack_nonlocal): Likewise.
3361
3362 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
3363
3364         PR target/85404
3365         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3366         Replace ASM_OUTPUT_LABEL with fprintf.
3367
3368 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
3369
3370         PR target/85417
3371         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3372         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3373         * config/i386/i386-c.c (ix86_target_macros_internal): Also
3374         define __IBT__ and __SHSTK__ for -fcf-protection.
3375         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3376         TARGET_IBT.
3377         (ix86_trampoline_init): Likewise.
3378         (x86_output_mi_thunk): Likewise.
3379         (ix86_notrack_prefixed_insn_p): Likewise.
3380         (ix86_option_override_internal): Don't disallow -fcf-protection.
3381         * config/i386/i386.md (rdssp<mode>): Also enable for
3382         -fcf-protection.
3383         (incssp<mode>): Likewise.
3384         (nop_endbr): Likewise.
3385         * config/i386/i386.opt (mcet): Change help message to built-in
3386         functions only.
3387         (mibt): Likewise.
3388         (mshstk): Likewise.
3389         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3390         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
3391         enable CET built-in functions.
3392
3393 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
3394
3395         * common/config/i386/i386-common.c
3396         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3397         OPTION_MASK_ISA_MOVDIRI_UNSET,
3398         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3399         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3400         * config.gcc (movdirintrin.h): New header.
3401         * config/i386/cpuid.h (bit_MOVDIRI,
3402         bit_MOVDIR64B): New bits.
3403         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3404         and -mmvodir64b.
3405         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3406         (VOID, PVOID, PCVOID)): New function types.
3407         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3408         __builtin_ia32_directstoreu_u64,
3409          __builtin_ia32_movdir64b): New builtins.
3410         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3411         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3412         and -mmovdiri.
3413         (ix86_valid_target_attribute_inner_p): Ditto.
3414         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3415         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3416         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3417         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3418         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3419         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3420         (movdiri<mode>, movdir64b_<mode>): New.
3421         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3422         * config/i386/immintrin.h: Include movdirintrin.h.
3423         * config/i386/movdirintrin.h: New file.
3424         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3425
3426 2018-04-19  Richard Biener  <rguenther@suse.de>
3427
3428         PR middle-end/85455
3429         * cfg.c (clear_bb_flags): When loop state says we have
3430         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3431
3432 2018-04-19  Richard Biener  <rguenther@suse.de>
3433
3434         PR tree-optimization/84737
3435         * tree-vect-data-refs.c (vect_copy_ref_info): New function
3436         copying restrict info.
3437         (vect_setup_realignment): Use it.
3438         * tree-vectorizer.h (vect_copy_ref_info): Declare.
3439         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3440         the first DR to all generated stores.
3441         (vectorizable_load): Likewise for loads.
3442
3443 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
3444
3445         PR tree-optimization/85446
3446         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3447         the integral and pointer types to have the same precision.
3448
3449         * doc/install.texi: Document --disable-cet being the default and
3450         --enable-cet=auto.
3451
3452 2018-04-18  Martin Liska  <mliska@suse.cz>
3453
3454         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3455         style.
3456
3457 2018-04-18  Martin Liska  <mliska@suse.cz>
3458
3459         Revert
3460         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
3461
3462         PR ipa/83983
3463         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3464         arguments if they are comparable.
3465
3466 2018-04-18  Martin Liska  <mliska@suse.cz>
3467
3468         Revert
3469         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
3470
3471         PR lto/84805
3472         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3473         incomplete types.
3474
3475 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
3476
3477         PR target/85388
3478         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3479         ENDBR after calling __morestack.
3480
3481 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
3482
3483         PR jit/85384
3484         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3485         by using gcc_base_ver to generate a gcc_driver_version, and use
3486         it when generating GCC_DRIVER_NAME.
3487         * configure: Regenerate.
3488
3489 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
3490
3491         PR target/81084
3492         * config.gcc: Obsolete powerpc*-*-*spe*.
3493
3494 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
3495
3496         PR debug/84637
3497         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3498         (stabstr_D): Change type of unum from unsigned int to
3499         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
3500         type.
3501
3502 2018-04-17  Jim Wilson  <jimw@sifive.com>
3503
3504         PR 84856
3505         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3506         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3507         Set arg_pointer_offset after using pretend_args_size.
3508
3509 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
3510
3511         PR rtl-optimization/85431
3512         * dse.c (record_store): Ignore zero width stores.
3513
3514         PR sanitizer/85230
3515         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
3516         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3517         __builtin_stack_restore rather than after it.
3518         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3519         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3520         argument instead of virtual_dynamic_stack_rtx.
3521
3522 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
3523
3524         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3525         New prototype.
3526         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3527         Add note to error message to explain internal mapping of overloaded
3528         built-in function name to non-overloaded built-in function name.
3529         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3530         function.
3531
3532 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
3533
3534         PR target/85424
3535         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3536         where the inputs overlap with the output.
3537
3538 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
3539
3540         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3541         (=v, v) alternative and explicit "memory" attribute.
3542         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
3543         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3544         attributes.
3545         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3546         "sselog1" type instead of "sselog".
3547         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3548         "sselog".  Remove explicit "memory" attribute.
3549         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3550         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3551         attributes.
3552         (vec_extract_hi_v32hi): Merge all alternatives into one, use
3553         "sselog1" type instead of "sselog".  Remove explicit "memory"
3554         attribute.
3555         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3556         use "sselog1" type instead of "sselog".  Remove explicit "memory"
3557         attribute.
3558         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3559         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3560         attributes.
3561         (vec_extract_hi_v64qi): Merge all alternatives into one, use
3562         "sselog1" type instead of "sselog".  Remove explicit "memory"
3563         attribute.
3564         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3565         use "sselog1" type instead of "sselog".  Remove explicit "memory"
3566         attribute.
3567
3568         PR target/85430
3569         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3570
3571         PR middle-end/85414
3572         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3573         on a SUBREG.
3574
3575 2018-04-17  Martin Jambor  <mjambor@suse.cz>
3576
3577         PR ipa/85421
3578         * ipa-cp.c (create_specialized_node): Call
3579         expand_all_artificial_thunks if necessary.
3580
3581 2018-04-17  Martin Liska  <mliska@suse.cz>
3582
3583         PR lto/85405
3584         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3585         in message, remote space in between '_G' and '('.
3586
3587 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
3588
3589         PR target/85281
3590         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3591         avx512f_vmcmp<mode>3<round_saeonly_name>,
3592         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3593         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3594         avx512f_rndscale<mode><round_saeonly_name>,
3595         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3596         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3597         Use %<iptr>2 instead of %2 for -masm=intel.
3598         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3599         avx512f_vcvttss2usi<round_saeonly_name>,
3600         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3601         -masm=intel.
3602         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3603         avx512f_vcvttsd2usi<round_saeonly_name>,
3604         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3605         Use %q1 instead of %1 for -masm=intel.
3606         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3607         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3608         of %3 for -masm=intel.
3609         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3610         -masm=intel.
3611         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3612         -masm=intel.
3613         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3614         -masm=intel.
3615         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3616         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3617         %g1.
3618         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3619         -masm=intel.
3620         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3621         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3622         %g1 and one with %0 and %1.
3623         (avx512er_vmrcp28<mode><round_saeonly_name>,
3624         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3625         %1 for -masm=intel.
3626         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3627         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3628         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3629         of %0 and %{%4%} for -masm=intel.
3630         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3631         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3632         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3633         order of %0 and %{%5%}%{z%} for -masm=intel.
3634
3635 2018-04-17  Jan Hubicka  <jh@suse.cz>
3636
3637         PR lto/85405
3638         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3639
3640 2018-04-17  Martin Liska  <mliska@suse.cz>
3641
3642         PR ipa/85329
3643         * multiple_target.c (create_dispatcher_calls): Set apostrophes
3644         for target_clone error message.  Make default implementation
3645         clone to be a local declaration.
3646         (separate_attrs): Add new argument and check for an empty
3647         string.
3648         (expand_target_clones): Handle it.
3649         (ipa_target_clone): Make redirection just for target_clones
3650         functions.
3651
3652 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
3653             Tom de Vries  <tom@codesourcery.com>
3654
3655         PR middle-end/84955
3656         * omp-expand.c (expand_oacc_for): Add dummy false branch for
3657         tiled basic blocks without omp continue statements.
3658
3659 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
3660
3661         PR target/83660
3662         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3663         vec_extract expression as having side effects to make sure it gets
3664         a cleanup point.
3665
3666 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
3667
3668         PR target/85403
3669         * config/i386/i386.c (get_builtin_code_for_version): Check
3670         error_mark_node.
3671
3672 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
3673
3674         PR target/84331
3675         * gcc/config.gcc: Support "skylake".
3676         * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3677         PROCESSOR_SKYLAKE.
3678         * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3679         (processor_target_table): Add "skylake".
3680         (ix86_option_override_internal): Add "skylake".
3681         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3682         PROCESSOR_CANNONLAKE.
3683         (get_builtin_code_for_version): Fix priority for
3684         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3685         PROCESSOR_SKYLAKE-AVX512.
3686         * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3687         (processor_type): Add PROCESSOR_SKYLAKE.
3688
3689 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
3690             Jason Merrill  <jason@redhat.com>
3691
3692         PR c++/85112
3693         * convert.c (convert_to_integer_1): Use direct recursion for
3694         enumeral types and types with a precision less than the number
3695         of bits in their mode.
3696
3697 2018-04-16  Julia Koval  <julia.koval@intel.com>
3698
3699         PR target/84413
3700         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3701         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3702
3703 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
3704
3705         PR target/85293
3706         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3707         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3708         and -mno-direct-move.
3709
3710 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
3711
3712         PR target/83402
3713         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3714         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3715         Ensure negative shifts result in {0}.
3716
3717 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
3718
3719         PR rtl-optimization/79916
3720         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3721         regs (if any) to define how to gnerate SD moves when LRA is in
3722         progress.
3723
3724 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
3725
3726         PR rtl-optimization/85393
3727         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3728         * except.c (expand_dw2_landing_pad_for_region): Make static.
3729         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3730         a label and unconditional jump to old_bb, rather than
3731         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3732         basic block.
3733
3734         PR rtl-optimization/85376
3735         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3736         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3737         instead of a specific value.
3738
3739 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
3740             Bin Cheng  <bin.cheng@arm.com>
3741
3742         PR tree-optimization/82965
3743         PR tree-optimization/83991
3744         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3745         by_profile_only parameter.
3746         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3747         information if the loop was predicted to iterate too many times.
3748         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3749
3750 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
3751
3752         PR lto/71991
3753         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3754         always inline.
3755
3756 2018-04-13  Martin Liska  <mliska@suse.cz>
3757             Jakub Jelinek  <jakub@redhat.com>
3758
3759         PR middle-end/81657
3760         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3761         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3762         * builtins.c (expand_builtin_memory_copy_args): Use
3763         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3764         handle dest_addr == pc_rtx.
3765
3766 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
3767
3768         PR target/85291
3769         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3770         asked to not generate direct moves.
3771         (fix_trunc<mode>si2_stfiwx): Similar.
3772         (fix_trunc<mode>si2_internal): Similar.
3773
3774 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
3775
3776         PR debug/83157
3777         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3778         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3779         lookup if dest in some wider mode is known to be const0_rtx and
3780         if so, record permanent equivalence for it to be ZERO_EXTEND of
3781         the narrower mode destination.
3782
3783 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
3784
3785         * lto-streamer-out.c (output_function): Revert 259346.
3786         * omp-expand.c (expand_oacc_for): Likewise.
3787
3788 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
3789
3790         PR rtl-optimization/85354
3791         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3792         * sel-sched.c (sel_global_init): ... here.
3793
3794 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
3795
3796         PR target/85238
3797         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3798         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3799         mode for PE-COFF targets.
3800         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3801         (i386_pe_asm_lto_end): Likewise.
3802         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3803         (TARGET_ASM_LTO_END): Likewise.
3804         * config/i386/winnt.c (saved_debug_info_level): New static variable.
3805         (i386_pe_asm_lto_start): New function.
3806         (i386_pe_asm_lto_end): Likewise.
3807
3808 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
3809             Richard Biener  <rguenther@suse.de>
3810
3811         PR middle-end/84955
3812         * lto-streamer-out.c (output_function): Fix CFG loop state before
3813         streaming out.
3814         * omp-expand.c (expand_oacc_for): Handle calls to internal
3815         functions like regular functions.
3816
3817 2018-04-12  Richard Biener  <rguenther@suse.de>
3818
3819         PR lto/85371
3820         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3821         for the early LTO debug to properly generate references to it
3822         during DIE emission.  Do not re-use that for the skeleton for
3823         split-dwarf.
3824         (dwarf2out_early_finish): Likewise.
3825
3826 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
3827
3828         PR target/85328
3829         * config/i386/sse.md
3830         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3831         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3832         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3833         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3834         and output is a reg, avoid creating invalid lowpart subreg, but
3835         instead split into a 512-bit move.  Don't split if not AVX512VL,
3836         input is xmm16+ reg and output is a mem.
3837         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3838         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3839         xmm16+ reg and output is a mem.
3840
3841 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
3842
3843         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3844         also for flag_dwarf2_cfi_asm.
3845
3846 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
3847
3848         PR rtl-optimization/85342
3849         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3850         a bool scalar var inside of the loop instead.  Don't try to update
3851         recog_data.operand after failed apply_change_group.
3852
3853 2018-04-12  Tom de Vries  <tom@codesourcery.com>
3854
3855         PR target/85296
3856         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3857         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
3858         array with flexible array member as array without given dimension.
3859         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3860         argument for undefined param to true.
3861
3862 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
3863
3864         PR target/85321
3865         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3866         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3867         from PowerPC section.
3868         * config/rs6000/sysv4.opt (mcall-): Improve help text.
3869         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3870         help text that is too long.
3871         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3872         help text that is too long.
3873         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3874         help text that is too long.
3875
3876 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
3877
3878         * config/alpha/alpha.md (stack_probe_internal): Rename
3879         from "probe_stack".  Update all callers.
3880
3881 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
3882
3883         PR rtl-optimization/84566
3884         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3885         sched_macro_fuse_insns.
3886
3887 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
3888
3889         PR target/84301
3890         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3891         (compute_block_dependences): ... from here.
3892
3893 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
3894
3895         PR tree-optimization/85331
3896         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3897         from int to HOST_WIDE_INT.
3898
3899 2018-04-11  Martin Jambor  <mjambor@suse.cz>
3900
3901         PR ipa/84149
3902         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3903         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3904         not the same as the source val.
3905         (cgraph_edge_brings_value_p): New parameter.
3906         (gather_edges_for_value): Pass destination value to
3907         cgraph_edge_brings_value_p.
3908         (perhaps_add_new_callers): Likewise.
3909         (get_info_about_necessary_edges): Likewise and exclude values brought
3910         only by self-recursive edges.
3911         (create_specialized_node): Redirect only clones of self-calling edges.
3912         (+self_recursive_pass_through_p): New function.
3913         (find_more_scalar_values_for_callers_subset): Use it.
3914         (find_aggregate_values_for_callers_subset): Likewise.
3915         (known_aggs_to_agg_replacement_list): Removed.
3916         (decide_whether_version_node): Re-calculate known constants for all
3917         remaining context clones.
3918
3919 2018-04-11  Richard Biener  <rguenther@suse.de>
3920
3921         PR lto/85339
3922         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3923         from early DWARF output.
3924         (dwarf2out_early_finish): Output line info unconditionally into
3925         early DWARF and add reference to it.
3926
3927 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
3928
3929         PR target/85281
3930         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3931         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3932         other than V2DFmode using iptr mode attribute.
3933         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3934
3935 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
3936
3937         PR rtl-optimization/84659
3938         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3939
3940 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
3941
3942         PR debug/85302
3943         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3944         SIZEP is NULL.
3945         (output_loc_list): Pass address of a dummy size variable even in the
3946         locview handling loop.
3947         (index_location_lists): Add comment on why skip_loc_list_entry can't
3948         call size_of_locs.
3949
3950 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
3951
3952         PR target/85261
3953         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3954         into register.
3955
3956 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
3957
3958         PR target/85321
3959         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3960         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3961         and -mstring-compare-inline-limit.
3962
3963 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
3964
3965         PR target/85287
3966         * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3967         for stack clash protection in a register whenever we need it to be in
3968         a register.
3969
3970 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
3971
3972         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3973         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3974
3975 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
3976
3977         PR target/85321
3978         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3979         the help text.
3980         (mlong-double-): Ditto.
3981         * config/rs6000/sysv4.opt (msdata=): Ditto.
3982         (mtls-size=): Ditto.
3983
3984 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
3985
3986         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3987         erroneous entries for
3988         "vector int vec_ldl (int, long int *)", and
3989         "vector unsigned int vec_ldl (int, unsigned long int *)".
3990         Add comments and entries for
3991         "vector bool char vec_ldl (int, bool char *)",
3992         "vector bool short vec_ldl (int, bool short *)",
3993         "vector bool int vec_ldl (int, bool int *)",
3994         "vector bool long long vec_ldl (int, bool long long *)",
3995         "vector pixel vec_ldl (int, pixel *)",
3996         "vector long long vec_ldl (int, long long *)",
3997         "vector unsigned long long vec_ldl (int, unsigned long long *)".
3998         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3999         type tree bool_long_long_type_node and correct definition of
4000         bool_V2DI_type_node to make reference to this new type tree.
4001         (rs6000_mangle_type): Replace erroneous reference to
4002         bool_long_type_node with bool_long_long_type_node.
4003         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4004         comments to emphasize sign distinctions for char and int types and
4005         replace RS6000_BTI_bool_long constant with
4006         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
4007         use of RS6000_BTI_pixel.
4008         (bool_long_type_node): Remove this macro definition.
4009         (bool_long_long_type_node): New macro definition
4010
4011 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
4012
4013         PR rtl-optimization/85300
4014         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4015         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4016         simplify_unary_operation fails.
4017
4018 2018-04-10  Martin Liska  <mliska@suse.cz>
4019
4020         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4021         cgraph_edge and ipa_ref.
4022
4023 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
4024
4025         PR target/85177
4026         PR target/85255
4027         * config/i386/sse.md
4028         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4029         computation of the VEC_MERGE selector from mask.
4030         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4031         Fix decoding of the VEC_MERGE selector into mask.
4032
4033 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
4034
4035         PR tree-optimization/85286
4036         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4037
4038 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
4039
4040         * final.c (final_1): Set insn_last_address as well as
4041         insn_current_address.
4042
4043 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4044
4045         PR target/85173
4046         * explow.c (emit_stack_probe): Call validize_mem on memory location
4047         before passing it to gen_probe_stack.  Create address operand and
4048         legitimize it for the probe_stack_address case.
4049
4050 2018-04-09  Jan Hubicka  <jh@suse.cz>
4051
4052         PR lto/85078
4053         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4054         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4055         * tree.c (free_lang_data_in_type): Fix handling of binfos;
4056         walk basetypes.
4057         (free_lang_data): Rebuild type inheritance graph.
4058
4059 2018-04-09  Martin Sebor  <msebor@redhat.com>
4060
4061         * invoke.texi (-finline-small-functions): Mention other optimization
4062         options.
4063         (-findirect-inlining, -fpartial-inlining): Same.
4064         (-finline-functions-called-once): Same.
4065         (-freorder-blocks-and-partition): Same.
4066
4067 2018-04-09  Jan Hubicka  <jh@suse.cz>
4068
4069         PR rtl/84058
4070         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4071         jumps; choose last target that matches the criteria (i.e.
4072         no partition changes for non-crossing jumps).
4073         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4074         support for redirecting crossing jumps to non-crossing.
4075
4076 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
4077
4078         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4079         also for naked functions.
4080
4081 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
4082
4083         * config/arc/arc.md (add_shift): New pattern.
4084         (add_shift2): Likewise.
4085         (sub_shift): Likewise.
4086         (sub_shift_cmp0_noout): Likewise.
4087         (compare_si_ashiftsi): Likewise.
4088         (xbfu_cmp0_noout): New combine pattern.
4089         (xbfu_cmp0"): Likewise.
4090         (movsi_set_cc_insn): Place the predicable variant first.
4091         (commutative_binary_cmp0_noout): Remove clobber.
4092         (commutative_binary_cmp0): New pattern.
4093         (noncommutative_binary_cmp0): Likewise.
4094         (noncommutative_binary_cmp0_noout): Likewise.
4095         (noncommutative_binary_comparison_result_used): Removed.
4096         (rsub_cmp0): New pattern.
4097         (rsub_cmp0_noout): Likewise.
4098         (extzvsi): Changed, keep only meaningful variants.
4099         (SQH, SEZ): New iterators.
4100         (SQH_postfix): New mode attribute.
4101         (SEZ_prefix): New code attribute.
4102         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4103         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4104         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4105         of numerical value.
4106         (noncommutative_operator): Check the availability of barrel
4107         shifter option.
4108
4109 2018-04-09  Richard Biener  <rguenther@suse.de>
4110
4111         PR tree-optimization/85284
4112         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4113         Only use the niter constraining form of simple_iv when the exit
4114         is always executed.
4115
4116 2018-04-09  Tom de Vries  <tom@codesourcery.com>
4117
4118         PR target/84041
4119         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4120         (define_expand "*memory_barrier"): New define_expand.
4121         (define_insn "memory_barrier"): New insn.
4122
4123 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
4124
4125         PR rtl-optimization/80463
4126         PR rtl-optimization/83972
4127         PR rtl-optimization/83480
4128
4129         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4130         correct producer for the insn.
4131         (tidy_control_flow): Fixup seqnos in case of debug insns.
4132
4133 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
4134
4135         PR rtl-optimization/83913
4136
4137         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4138         different sched-times when merging exprs.
4139
4140 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
4141
4142         PR rtl-optimization/83962
4143
4144         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4145         tidy_fallthru_edge and tidy_control_flow.
4146
4147 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
4148
4149         PR rtl-optimization/83530
4150
4151         * sel-sched.c (force_next_insn): New global variable.
4152         (remove_insn_for_debug): When force_next_insn is true, also leave only
4153         next insn in the ready list.
4154         (sel_sched_region): When the region wasn't scheduled, make another pass
4155         over it with force_next_insn set to 1.
4156
4157 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
4158
4159         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4160         into tm_file.
4161         * config/nds32/constants.md (unspec_volatile_element): Add enum values
4162         for interrupt control.
4163         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4164         functions for interrupt control.
4165         * config/nds32/nds32-intrinsic.md: Likewise.
4166         * config/nds32/nds32_intrinsic.h: Likewise.
4167         * config/nds32/nds32.h (nds32_builtins): Likewise.
4168
4169 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
4170
4171         * config/nds32/nds32.c (nds32_init_machine_status,
4172         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4173         strict_aligned_p field.
4174         (nds32_expand_to_rtl_hook): New function.
4175         (TARGET_EXPAND_TO_RTL_HOOK): Define.
4176         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4177
4178 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
4179             Chung-Ju Wu  <jasonwucj@gmail.com>
4180
4181         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4182         * config/nds32/nds32-n7.md: New file.
4183         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4184         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4185         pipeline.
4186         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4187         * config/nds32/nds32.md (pipeline_model): Add n7.
4188         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4189         * config/nds32/pipelines.md: Include n7 settings.
4190
4191 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
4192             Chung-Ju Wu  <jasonwucj@gmail.com>
4193
4194         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4195         * config/nds32/nds32-e8.md: New file.
4196         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4197         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4198         pipeline.
4199         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4200         * config/nds32/nds32.md (pipeline_model): Add e8.
4201         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4202         * config/nds32/pipelines.md: Include e8 settings.
4203
4204 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
4205             Chung-Ju Wu  <jasonwucj@gmail.com>
4206
4207         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4208         * config/nds32/nds32-n8.md: New file.
4209         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4210         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4211         pipeline.
4212         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4213         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4214         * config/nds32/nds32.md (pipeline_model): Add n8.
4215         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4216         * config/nds32/pipelines.md: Include n8 settings.
4217
4218 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
4219             Chung-Ju Wu  <jasonwucj@gmail.com>
4220
4221         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4222         * config/nds32/nds32-n9-2r1w.md: New file.
4223         * config/nds32/nds32-n9-3r2w.md: New file.
4224         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4225         nds32_register_ports): New or modify for cpu n9.
4226         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4227         pipeline.
4228         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4229         * config/nds32/nds32-utils.c: New file.
4230         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4231         TARGET_MUL_SLOW): Define.
4232         * config/nds32/nds32.md (pipeline_model): New attribute.
4233         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4234         New options that support cpu n9.
4235         * config/nds32/pipelines.md: Include n9 settings.
4236         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4237
4238 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
4239
4240         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4241         information if necessary.
4242         (output_cond_branch_compare_zero): Likewise.
4243         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4244         (nds32_target_alignment): Refine for alignment.
4245         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4246         (FUNCTION_BOUNDARY): Modify.
4247         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4248         align case.
4249         * config/nds32/nds32.opt (malways-align, malign-functions): New.
4250
4251 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
4252
4253         * config/nds32/constants.md (unspec_volatile_element): Add values for
4254         TLB operation and data prefetch.
4255         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4256         functions for TLB operation and data prefetch.
4257         * config/nds32/nds32-intrinsic.md: Likewise.
4258         * config/nds32/nds32_intrinsic.h: Likewise.
4259         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4260         (nds32_print_operand): Likewise.
4261         * config/nds32/nds32.h (nds32_builtins): Likewise.
4262
4263 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
4264         Andrew Pinski <pinsika@gcc.gnu.org>
4265
4266         PR middle-end/82976
4267         * match.pd: Use constant_boolean_node of correct type instead of
4268         boolean_true_node or boolean_false_node for simplifying
4269         pointer comparisons to zero.
4270
4271 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
4272
4273         PR tree-optimization/80021
4274         * tree.c (verify_type_variant): Make error call in verify_variant_match
4275         translatable and remove final full stop.
4276
4277 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
4278
4279         * config/nds32/constants.md (unspec_volatile_element): Add
4280         UNSPEC_VOLATILE_EH_RETURN.
4281         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4282         nds32_output_stack_pop): Support dwarf exception handling process.
4283         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4284         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4285         exception handling process.
4286         (nds32_compute_stack_frame): Likewise.
4287         (nds32_return_addr_rtx): Likewise.
4288         (nds32_initial_elimination_offset): Likewise.
4289         (nds32_expand_prologue): Likewise.
4290         (nds32_expand_epilogue): Likewise.
4291         (nds32_dynamic_chain_address): New function.
4292         * config/nds32/nds32.h (machine_function): Add fields for dwarf
4293         exception handling.
4294         (DYNAMIC_CHAIN_ADDRESS): Define.
4295         (EH_RETURN_DATA_REGNO): Define.
4296         (EH_RETURN_STACKADJ_RTX): Define.
4297         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4298         patterns for dwarf exception handling.
4299
4300 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
4301
4302         * config/nds32/nds32.h: Clean up obsolete macros.
4303
4304 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
4305
4306         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4307         Add enum values for particular instructions.
4308         * config/nds32/nds32-intrinsic.c: Implementation of expanding
4309         particular intrinsic functions.
4310         * config/nds32/nds32-intrinsic.md: Likewise.
4311         * config/nds32/nds32_intrinsic.h: Likewise.
4312         * config/nds32/nds32.h (nds32_builtins): Likewise.
4313         * config/nds32/nds32.md (type): Add pbsad and pbsada.
4314         (btst, ave): New patterns for particular instructions.
4315
4316 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
4317
4318         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4319         Add enum values for atomic load/store and memory sync.
4320         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4321         and memory sync.
4322         * config/nds32/nds32-intrinsic.md: Likewise.
4323         * config/nds32/nds32_intrinsic.h: Likewise.
4324         * config/nds32/nds32.h (nds32_builtins): Likewise.
4325
4326 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
4327
4328         PR tree-optimization/85257
4329         * fold-const.c (native_encode_vector): If not all elts could fit
4330         and off is -1, return 0 rather than offset.
4331         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4332         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4333         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
4334         adjust buffer in native_interpret_expr call.
4335
4336 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
4337
4338         * config/nds32/constants.md (unspec_volatile_element): Add cache
4339         control enum values.
4340         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4341         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4342         * config/nds32/nds32.c (nds32_cctl_names): New.
4343         (nds32_print_operand): Handle cache control register names.
4344         * config/nds32/nds32.h (nds32_builtins): New enum values.
4345         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4346         macros.
4347         * config/nds32/nds32.md (type): Add mmu.
4348         * config/nds32/pipelines.md (simple_insn): Add mmu.
4349
4350 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
4351
4352         * config/nds32/nds32.md (type): Remove call.
4353         * config/nds32/pipelines.md (simple_insn): Likewise.
4354
4355 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
4356
4357         * config/nds32/constants.md (unspec_volatile_element): Add
4358         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4359         UNSPEC_VOLATILE_FMFCFG.
4360         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4361         description for fmfcfg and fmfcsr.
4362         (bdesc_1arg): Add fmtcsr.
4363         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4364         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4365         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4366         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4367         unspec_fmfcfg): New patterns.
4368         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4369         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4370         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4371         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4372         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4373         __nds32__fmfcfg): Define.
4374
4375 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
4376
4377         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4378         intrinsic register names.
4379         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4380         intrinsic register enum values and macros.
4381
4382 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
4383
4384         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4385         for load/store addressing form.
4386         (nds32_print_operand_address): Likewise.
4387
4388 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
4389
4390         PR target/85196
4391         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4392         based on LABEL_REF.  Remove useless assertion.
4393         (pic_address_needs_scratch): Fix formatting.
4394         (sparc_legitimize_pic_address): Minor tweaks.
4395         (sparc_delegitimize_address): Adjust assertion accordingly.
4396         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4397         into symbolic_operand.
4398         (movsi_high_pic_label_ref): Likewise.
4399         (movsi_lo_sum_pic_label_ref): Likewise.
4400         (movdi_pic_label_ref): Likewise.
4401         (movdi_high_pic_label_ref): Likewise.
4402         (movdi_lo_sum_pic_label_ref): Likewise.
4403
4404 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
4405
4406         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4407         custom LIB_SPEC setup.
4408
4409 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
4410             Kito Cheng  <kito.cheng@gmail.com>
4411
4412         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4413         * config/riscv/freebsd.h: New.
4414
4415 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
4416
4417         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4418         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4419         file.
4420
4421 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
4422             Kito Cheng  <kito.cheng@gmail.com>
4423
4424         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4425         nds32_output_call, nds32_symbol_binds_local_p): New functions.
4426         * config/nds32/nds32-protos.h (nds32_output_call,
4427         nds32_output_return): Declare.
4428         * config/nds32/nds32.md: Refine all the call and return patterns.
4429
4430 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
4431
4432         PR debug/85252
4433         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4434         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4435
4436         PR rtl-optimization/84872
4437         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4438         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4439         EDGE_CROSSING edge.
4440
4441 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
4442
4443         * expr.c (copy_blkmode_to_reg): Revert 254862.
4444         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4445
4446 2018-04-06  Richard Biener  <rguenther@suse.de>
4447
4448         PR middle-end/85244
4449         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4450         after seeing a component reference with an adjacent field.  Treat
4451         refs to arrays at struct end of external decls similar to
4452         refs to unconstrained commons.
4453
4454 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
4455
4456         PR sanitizer/85213
4457         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4458         look through SAVE_EXPRs with non-side-effects argument.  Adjust
4459         recursive calls.
4460         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4461         save_p here.
4462
4463 2018-04-06  Richard Biener  <rguenther@suse.de>
4464
4465         PR middle-end/85180
4466         * alias.c (find_base_term): New wrapper around find_base_term
4467         unwinding CSELIB_VAL_PTR changes.
4468         (find_base_term): Do not restore CSELIB_VAL_PTR during the
4469         recursion.
4470
4471 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
4472
4473         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4474         instructions.
4475         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4476         constant definitions.
4477         ("nop"): lr 0,0 -> nopr r0
4478         ("nop_lr0", "nop_lr1"): New insn definitions.
4479
4480 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
4481
4482         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4483         NDS32_V3PUSH_AVAILABLE_P macro.
4484
4485 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
4486             Chung-Ju Wu  <jasonwucj@gmail.com>
4487
4488         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4489         (nds32*-*-*): Add float and fpu_config into supported_defaults.
4490         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4491         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4492         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4493         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4494         * config/nds32/constraints.md: New constraints and checking for hard
4495         float configuration.
4496         * config/nds32/iterators.md: New mode iterator and attribute for hard
4497         float configuration.
4498         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4499         patterns.
4500         * config/nds32/nds32-fpu.md: New file.
4501         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4502         deal with hard float code generation.
4503         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4504         ARCH_V3S.
4505         (abi_type, float_reg_number): New enum type.
4506         * config/nds32/nds32-predicates.c: New predicates for hard float.
4507         * config/nds32/nds32-protos.h: Declare functions for hard float.
4508         * config/nds32/nds32.c: Implementation for hard float configuration.
4509         * config/nds32/nds32.h: Definitions for hard float configuration.
4510         * config/nds32/nds32.md: Include hard float machine description and
4511         modify patterns for hard float configuration.
4512         * config/nds32/nds32.opt: New options for hard float configuration.
4513         * config/nds32/predicates.md: New predicates for hard float
4514         configuration.
4515
4516 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
4517
4518         * common/config/nds32/nds32-common.c
4519         (nds32_option_optimization_table): Enable -mreleax-hint by default.
4520
4521 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
4522
4523         PR middle-end/85195
4524         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4525         CONSTRUCTOR_ELT (ctor, ...)->value.
4526
4527 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
4528
4529         PR target/85193
4530         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4531
4532 2018-04-05  Tom de Vries  <tom@codesourcery.com>
4533
4534         PR target/85204
4535         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4536         cond jump.
4537
4538 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
4539             Kito Cheng  <kito.cheng@gmail.com>
4540
4541         * config/nds32/constraints.md (U33): Fine-tune checking condition.
4542         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4543         * config/nds32/nds32.h (nds32_16bit_address_type): Add
4544         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4545
4546 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
4547             Kito Cheng  <kito.cheng@gmail.com>
4548
4549         * config/nds32/constraints.md (Ufe): New memory constraint.
4550         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4551         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4552         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4553         operands.
4554         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4555         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4556
4557 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
4558
4559         * config/nds32/nds32.md: Use optimize_size in the condition for
4560         alu-shift instructions.
4561
4562 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
4563
4564         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4565
4566 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
4567
4568         * config/nds32/nds32.md (negsi2): Refine pattern.
4569
4570 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
4571             Chung-Ju Wu  <jasonwucj@gmail.com>
4572
4573         * config/nds32/iterators.md (shift_rotate): New code iterator.
4574         (shift): New code attribute.
4575         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4576         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4577         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4578         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4579         bit-wise operations.
4580         (andsi3, *andsi3): Ditto.
4581         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4582         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4583         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4584         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4585         nds32_ior_operand, nds32_xor_operand): New predicates.
4586
4587 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
4588
4589         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4590         (addsi3, subsi3): ... this.
4591
4592 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
4593
4594         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4595
4596 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
4597
4598         * config/nds32/nds32.md: Adjust indention.
4599
4600 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
4601
4602         * config/nds32/nds32.md (feature): New attribute.
4603
4604 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
4605
4606         * config/nds32/nds32.md (subtype): New attribute.
4607
4608 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
4609
4610         PR target/85203
4611         * config/arm/arm-builtins.c (arm_expand_builtin): Change
4612         expansion to perform a bitwise AND of the argument followed by a
4613         boolean negation of the result.
4614
4615 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
4616
4617         PR rtl-optimization/84878
4618         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4619         the basic block.  Assert the use reference is not artificial and that
4620         it has an associated insn.
4621
4622 2018-04-04  Michael Matz  <matz@suse.de>
4623
4624         * builtins.c (compute_objsize): Pass correct operand
4625         to array_at_struct_end_p.
4626
4627 2018-04-04  Richard Biener  <rguenther@suse.de>
4628
4629         PR lto/85176
4630         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4631         from contexts for DINFO_LEVEL_TERSE and below.
4632
4633 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
4634
4635         * config/nds32/nds32-doubleword.md (move_<mode>): Require
4636         resiter_operand condition.
4637         * config/nds32/nds32.md (*move<mode>): Ditto.
4638
4639 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
4640             Monk Chiang  <sh.chiang04@gmail.com>
4641
4642         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4643
4644 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
4645
4646         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4647
4648 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
4649             Kito Cheng  <kito.cheng@gmail.com>
4650
4651         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4652         nds32_cond_code_str, output_cond_branch,
4653         output_cond_branch_compare_zero, nds32_expand_cbranch,
4654         nds32_expand_cstore, nds32_expand_movcc,
4655         nds32_output_cbranchsi4_equality_zero,
4656         nds32_output_cbranchsi4_equality_reg,
4657         nds32_output_cbranchsi4_equality_reg_or_const_int,
4658         nds32_output_cbranchsi4_greater_less_zero: New functions.
4659         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4660         nds32_expand_cstore, nds32_expand_movcc,
4661         nds32_output_cbranchsi4_equality_zero,
4662         nds32_output_cbranchsi4_equality_reg,
4663         nds32_output_cbranchsi4_equality_reg_or_const_int,
4664         nds32_output_cbranchsi4_greater_less_zero): Declare.
4665         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4666         nds32_rimm11s_operand): New predicates.
4667         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4668         * config/nds32/nds32.md: Rewrite all the branch and conditional move
4669         patterns.
4670
4671 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
4672
4673         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4674         * config/nds32/nds32.md: Ditto.
4675         * config/nds32/pipelines.md: Ditto.
4676
4677 2018-04-04  Richard Biener  <rguenther@suse.de>
4678
4679         PR tree-optimization/85168
4680         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4681         propagating abnormals.
4682
4683 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
4684
4685         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4686
4687 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
4688             Kito Cheng  <kito.cheng@gmail.com>
4689
4690         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4691         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4692         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4693         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4694         * config/nds32/nds32.md (sibcall_internal): New.
4695         (sibcall_register): Remove.
4696         (sibcall_immediate): Remove.
4697         (sibcall_value_internal): New.
4698         (sibcall_value_register): Remove.
4699         (sibcall_value_immediate): Remove.
4700         * config/nds32/predicates.md (nds32_general_register_operand): New.
4701         (nds32_call_address_operand): New.
4702
4703 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
4704
4705         PR rtl-optimization/85167
4706         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4707         bb_defs if *split_p, instead preinitialize it to NULL.
4708
4709         PR tree-optimization/85156
4710         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4711         evaluating the argument multiple times.
4712
4713 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
4714
4715         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4716         than vector.
4717         (_mm_cvtpd_ps): Likewise.
4718         (_mm_cvttpd_epi32): Likewise.
4719         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4720         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4721         vector, pixel, and bool following altivec.h include.
4722
4723 2018-04-03  Martin Sebor  <msebor@redhat.com>
4724
4725         * doc/extend.texi (Common Function Attributes): Clarify.
4726         (const attribute): Likewise.
4727         (pure attribute): Likewise.
4728
4729 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
4730
4731         PR target/85169
4732         * config/i386/i386.c (ix86_expand_vector_set): Use
4733         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
4734
4735 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
4736
4737         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4738         instructions when changing rounding bits to preserve precision bits
4739         in the x87 control word.
4740
4741 2018-04-03  Martin Liska  <mliska@suse.cz>
4742
4743         PR tree-optimization/82491
4744         * rtl.h (strip_offset_and_add): Replace += suboffset with
4745         poly_uint64 () + suboffset.
4746
4747 2018-03-29  Martin Liska  <mliska@suse.cz>
4748             Martin Jambor  <mjambor@suse.cz>
4749
4750         PR ipa/84947
4751         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4752         param_type is not an integral or pointer type.
4753
4754 2018-04-03  Richard Biener  <rguenther@suse.de>
4755
4756         * sese.h (recompute_all_dominators): Remove.
4757
4758 2018-04-02  Martin Sebor  <msebor@redhat.com>
4759
4760         * doc/invoke.texi (-Wrestrict): Fix typos.
4761
4762 2018-04-02  Jim Wilson  <jimw@sifive.com>
4763
4764         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4765         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4766         (<optab>di3, <optab>si3_extend): Likewise.
4767         (<optab>si3_mask, <optab>si3_mask_1): New.
4768         (<optab>di3_mask, <optab>di3_mask_1): New.
4769         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4770         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4771         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4772
4773 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
4774
4775         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4776         example.
4777
4778 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
4779
4780         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4781         (nds32_canonicalize_comparison): New function.
4782
4783 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
4784             Kito Cheng  <kito.cheng@gmail.com>
4785             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
4786
4787         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4788         * config/nds32/constants.md (unspec_volatile_element): Add
4789         UNSPEC_VOLATILE_RELAX_GROUP.
4790         * config/nds32/nds32-relax-opt.c: New file.
4791         * config/nds32/nds32-predicates.c
4792         (nds32_symbol_load_store_p): New function.
4793         * config/nds32/nds32-protos.h
4794         (nds32_symbol_load_store_p): Declare function.
4795         (make_pass_nds32_relax_opt): Declare new rtl pass function.
4796         * config/nds32/nds32.c
4797         (nds32_register_pass): New function to register pass.
4798         (nds32_register_passes): New function to register passes.
4799         * config/nds32/nds32.md (relax_group): New pattern.
4800         * config/nds32/nds32.opt (mrelax-hint): New option.
4801         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4802
4803 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
4804
4805         * config/nds32/t-nds32: Modify files dependency.
4806
4807 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
4808
4809         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4810         (PROFILE_HOOK): Define its implementation.
4811
4812 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
4813
4814         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4815         type and 32-bit size.
4816
4817 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
4818
4819         PR middle-end/85090
4820         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4821         (V_128_256): New mode iterator.
4822         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4823         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4824         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4825         of V.
4826         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4827         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4828
4829 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
4830
4831         PR target/83315
4832         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4833         NaN inputs correctly.
4834
4835 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
4836
4837         PR target/80546
4838         * config/rs6000/vsx.md (??r): New mode attribute.
4839         (*vsx_mov<mode>_64bit): Use it.
4840         (*vsx_mov<mode>_32bit): Likewise.
4841
4842 2018-03-30  Martin Sebor  <msebor@redhat.com>
4843
4844         PR tree-optimization/84818
4845         * builtins.c (check_access): Use warning_n.
4846
4847 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
4848
4849         PR target/83822
4850         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4851         condition.
4852         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4853         condition.
4854
4855 2018-03-30  Julia Koval  <julia.koval@intel.com>
4856
4857         PR target/84413
4858         * x86-tune.def (movx, partial_reg_dependency): Enable for
4859         m_SKYLAKE_AVX512.
4860
4861 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
4862
4863         PR inline-asm/84985
4864         * lra-constraints.c (process_alt_operands): Move setting
4865         this_alternative_matches below.
4866
4867 2018-03-29  Martin Liska  <mliska@suse.cz>
4868
4869         PR lto/84995.
4870         * doc/invoke.texi: Document how LTO works with debug info.
4871         Describe auto-load support of binutils.  Mention 'x86-64'
4872         as valid option value of -march option.
4873
4874 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
4875
4876         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4877
4878         PR c/85094
4879         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4880         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4881         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4882         checking.
4883
4884 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
4885
4886         PR target/84912
4887         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4888         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4889         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4890         for RS6000_BTM_POWERPC64.
4891         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4892         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4893         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4894         definition.
4895         (DIVDE): Use it.
4896         (DIVDEU): Likewise.
4897
4898 2018-03-28 Carl Love  <cel@us.ibm.com>
4899
4900         Revert
4901         2017-09-27  Carl Love  <cel@us.ibm.com>
4902
4903         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4904         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4905         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4906         fctiw instruction.
4907
4908 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4909
4910         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4911         instead of __vector bool.
4912         (_mm_max_pu8): Likewise.
4913         (_mm_min_pi16): Likewise.
4914
4915 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
4916
4917         PR target/84912
4918         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4919         (DIVWEUO): Likewise.
4920         (DIVDEO): Likewise.
4921         (DIVDEUO): Likewise.
4922         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4923         DIVWEUO and DIVDEUO.
4924         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4925         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4926         (div_extend): Likewise.
4927         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4928         builtin function.
4929         (__builtin_divweuo): Likewise.
4930         (__builtin_divdeo): Likewise.
4931         (__builtin_divdeuo): Likewise.
4932
4933 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
4934
4935         PR target/85095
4936         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4937         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4938
4939         PR tree-optimization/82004
4940         * gimple-match-head.c (optimize_pow_to_exp): New function.
4941         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4942         Don't fold to exp if optimize_pow_to_exp is false.
4943
4944 2018-03-28  Martin Liska  <mliska@suse.cz>
4945
4946         PR other/84819
4947         * calls.c (initialize_argument_information): Fix trailing space.
4948         * common.opt: Fix typo and provide better explanation for
4949         -fsanitize-coverage option.
4950         * config/i386/i386.opt: Fix typo.
4951
4952 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
4953             Martin Liska  <mliska@suse.cz>
4954
4955         PR sanitizer/85081
4956         * gimplify.c (asan_poison_variable): Don't do the check for
4957         gimplify_omp_ctxp here.
4958         (gimplify_decl_expr): Do it here.
4959         (gimplify_target_expr): Likewise.
4960
4961 2018-03-28  Martin Liska  <mliska@suse.cz>
4962
4963         PR target/84988
4964         * config/i386/i386.c (ix86_function_arg_advance): Do not call
4965         chkp_type_bounds_count if MPX is not enabled.
4966
4967 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
4968
4969         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4970
4971 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
4972
4973         PR target/84914
4974         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4975         function to create the function decl for complex long double
4976         multiply and divide for -mabi=ieeelongdouble.
4977         (init_float128_ieee): Call it.
4978
4979 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
4980
4981         PR target/85044
4982         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4983         -fcf-protection=branch -mibt.
4984         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4985
4986 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
4987
4988         PR target/81863
4989         * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4990
4991 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
4992
4993         PR target/85056
4994         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4995         extern array declarations.
4996
4997 2018-03-27  Richard Biener  <rguenther@suse.de>
4998
4999         PR middle-end/84067
5000         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5001         explicit single_use checks.
5002
5003 2018-03-27  Richard Biener  <rguenther@suse.de>
5004
5005         PR tree-optimization/85082
5006         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5007         Valueize the VUSE.
5008
5009 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5010
5011         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5012         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5013         Turn on fasynchronous-unwind-tables and funwind-tables.
5014
5015 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
5016
5017         PR target/85073
5018         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5019         (*bmi_blsr_<mode>_ccz): Ditto.
5020
5021 2018-03-26  Tom de Vries  <tom@codesourcery.com>
5022
5023         PR tree-optimization/85063
5024         * omp-general.c (offloading_function_p): New function.  Factor out
5025         of ...
5026         * omp-offload.c (pass_omp_target_link::gate): ... here.
5027         * omp-general.h (offloading_function_p): Declare.
5028         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5029         with attribute omp declare target for offloading functions.
5030
5031 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
5032
5033         PR tree-optimization/84005
5034         * tree-data-ref.h (get_base_for_alignment): Declare.
5035         * tree-data-ref.c (get_base_for_alignment_1): New function.
5036         (get_base_for_alignment): Likewise.
5037         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5038         get_base_for_alignment to find a suitable base object, instead
5039         of always using drb->base_address.
5040
5041 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
5042
5043         PR inline-asm/85022
5044         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5045         known size by default.
5046
5047 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
5048
5049         PR inline-asm/85030
5050         * lra-constraints.c (process_alt_operands): Don't match BLKmode
5051         and non BLKmode operands.
5052
5053 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5054
5055         PR target/85026
5056         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5057         Clean up attributes.
5058
5059 2018-03-23  Richard Biener  <rguenther@suse.de>
5060
5061         PR debug/85020
5062         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5063         we are going to emit early debug for LTO.
5064
5065 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
5066
5067         PR inline-asm/85034
5068         * function.c (match_asm_constraints_1): Don't optimize if input
5069         doesn't satisfy general_operand predicate for output's mode.
5070
5071         PR inline-asm/85022
5072         * alias.c (write_dependence_p): Don't require for x_canonicalized
5073         non-VOIDmode if x has VOIDmode.
5074
5075         PR sanitizer/85029
5076         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5077         just don't try to optimize it rather than assert it never happens.
5078
5079 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5080
5081         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5082         macro expansions for definition of ST_INTERNAL_<mode> and
5083         LD_INTERNAL_<mode> builtins.
5084         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5085         Remove prototype.
5086         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5087         function.
5088         (altivec_expand_st_builtin): Likewise.
5089         (altivec_expand_builtin): Remove calls to deleted functions.
5090         (rs6000_address_for_altivec): Delete this function.
5091         * config/rs6000/vector.md: Remove expands for
5092         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5093
5094 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
5095
5096         PR target/84826
5097         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5098         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5099         re-computing once computed.
5100         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5101         (arm_init_machine_status): Initialize
5102         machine->static_chain_stack_bytes.
5103
5104 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5105
5106         PR target/84760
5107         * doc/extend.texi: Add four new prototypes for vec_ld.
5108         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5109         definitions for more logical presentation.
5110         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5111         entries for V1TI variants of __builtin_altivec_ld builtin.
5112         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5113         handling of V1TI variant of LVX icode pattern.
5114         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5115         (rs6000_gimple_fold_builtin): Likewise.
5116         (altivec_init_builtins): Add code to define
5117         __builtin_altivec_lvx_v1ti function.
5118
5119 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
5120
5121         PR inline-asm/84941
5122         * function.c (match_asm_constraints_1): Don't do the optimization
5123         if input isn't a REG, SUBREG, MEM or constant.
5124
5125 2018-03-22  Tom de Vries  <tom@codesourcery.com>
5126
5127         PR tree-optimization/84956
5128         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5129         bb_has_abnormal_pred.
5130
5131 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
5132
5133         PR sanitizer/85018
5134         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5135         DECL_INITIAL (decl) to decl at the end.
5136         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5137         adjust the comment.
5138
5139 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
5140
5141         * doc/extend.texi (__builtin_tgmath): Document when complex
5142         integer types are treated as _Complex _Float64.
5143
5144 2018-03-21  Tom de Vries  <tom@codesourcery.com>
5145
5146         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5147
5148 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
5149
5150         PR tree-optimization/84960
5151         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5152         if it is ENTRY block, move them into single succ of ENTRY in that case.
5153
5154 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
5155
5156         PR tree-optimization/84811
5157         * poly-int.h (poly_span_traits): Remove the T3 parameter and
5158         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5159         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5160         (known_subrange_p): Update accordingly.  Cast each value involved
5161         in the size comparison, rather than casting the result of the
5162         subtraction.
5163
5164 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
5165
5166         PR tree-optimization/84982
5167         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5168         by flipping the least significant bit rather than all bits from
5169         bitpos to bitpos + bitsize - 1.
5170
5171 2018-03-21  Nathan Sidwell  <nathan@acm.org>
5172
5173         * doc/extend.texi (Deprecated Features): Remove mention of
5174         long-deleted deprecations.
5175
5176 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
5177
5178         PR jit/84288
5179         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5180         * configure: Regenerate.
5181
5182 2018-03-21  Tom de Vries  <tom@codesourcery.com>
5183
5184         PR tree-optimization/83126
5185         * tree-parloops.c (num_phis): New function.
5186         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5187
5188 2018-03-21  Nathan Sidwell  <nathan@acm.org>
5189
5190         * doc/extend.texi (Deprecated Features): Update deprecated flags,
5191         mention anon-struct/union members and trailing attributes.
5192
5193 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
5194
5195         PR tree-optimization/84969
5196         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5197         builtin memset partitions if they set different rhs values.
5198
5199 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
5200
5201         PR rtl-optimization/84989
5202         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5203         VEC_DUPLICATE with scalar result mode.
5204
5205 2018-03-21  Martin Liska  <mliska@suse.cz>
5206
5207         PR ipa/84963
5208         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5209         not intended return statement.
5210
5211 2018-03-21  Martin Liska  <mliska@suse.cz>
5212
5213         PR target/84988
5214         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5215         (chkp_find_bound_slots_1): Limit number of iterations.
5216
5217 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
5218
5219         PR target/84838
5220         * Minor grammar fixes for x86 options.
5221
5222 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
5223
5224         PR debug/84875
5225         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5226         holding REG_CFA_RESTORE notes, instead turn them into a USE.
5227
5228 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
5229
5230         PR target/83789
5231         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5232         (altivec_lvx_<mode>_1op): Likewise.
5233         (altivec_stvx_<mode>_2op): Likewise.
5234         (altivec_stvx_<mode>_1op): Likewise.
5235         (altivec_lvx_<VM2:mode>): New define_expand.
5236         (altivec_stvx_<VM2:mode>): Likewise.
5237         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5238         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5239         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5240         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5241         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5242         (rs6000_gen_lvx): Likewise.
5243         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5244         (altivec_expand_stv_builtin): Likewise.
5245         (altivec_expand_builtin): Likewise.
5246         * config/rs6000/vector.md: Likewise.
5247
5248 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5249
5250         PR target/82518
5251         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5252         BYTES_BIG_ENDIAN.
5253
5254 2018-03-20  Richard Biener  <rguenther@suse.de>
5255
5256         PR target/84986
5257         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5258         sign-conversions as zero, fall back to standard scalar_stmt
5259         cost for the rest.
5260
5261 2018-03-20  Martin Liska  <mliska@suse.cz>
5262
5263         PR ipa/84825
5264         * predict.c (rebuild_frequencies): Handle case when we have
5265         PROFILE_ABSENT, but flag_guess_branch_prob is false.
5266
5267 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
5268
5269         PR target/84990
5270         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5271         flag_section_anchors.
5272         * varasm.c (use_blocks_for_decl_p): Remove hack for
5273         dw2_force_const_mem.
5274
5275         PR target/84845
5276         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5277         to ...
5278         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
5279         be created, use lowpart_subreg of operands[0] rather than operands[0]
5280         itself.
5281         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5282         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5283         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5284         and n constraint instead of aarch64_shift_imm_di and Usd.
5285         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5286         (*aarch64_<optab>_reg_minus<mode>3): ... this.
5287
5288 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
5289
5290         PR target/82989
5291         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5292         to favor GPR over NEON registers.
5293         (<shift>di3_neon): Likewise.
5294
5295 2018-03-20  Tom de Vries  <tom@codesourcery.com>
5296
5297         PR target/84952
5298         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5299         (nvptx_process_pars): Emit bar.sync asap and alap.
5300
5301 2018-03-20  Tom de Vries  <tom@codesourcery.com>
5302
5303         PR target/84954
5304         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5305         seen_label if seen_label is already set.
5306
5307 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
5308
5309         PR target/84945
5310         * config/i386/i386.c (fold_builtin_cpu): For features above 31
5311         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5312         Use 1U instead of 1.  Formatting fixes.
5313
5314         PR c/84953
5315         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5316         instead of TREE_TYPE (s1) for the return value.
5317
5318 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
5319
5320         PR tree-optimization/84946
5321         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5322         bitsize + bitsize in poly_uint64 rather than poly_int64.
5323
5324         PR sanitizer/78651
5325         * dwarf2asm.c: Include fold-const.c.
5326         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5327         of decl rather than decl itself.
5328
5329         PR rtl-optimization/84643
5330         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5331
5332 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
5333
5334         PR sanitizer/78651
5335         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5336         calling assemble_variable.
5337
5338 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
5339
5340         PR target/81647
5341         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5342         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5343
5344 2018-03-19  Jim Wilson  <jimw@sifive.com>
5345
5346         PR bootstrap/84856
5347         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5348         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5349         (riscv_first_stack_step): Likewise.
5350         (riscv_option_override): Use STACK_BOUNDARY instead of
5351         MIN_STACK_BOUNDARY.
5352         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5353         MIN_STACK_BOUNDARY.
5354         (BIGGEST_ALIGNMENT): Set to 128.
5355         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5356         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5357         STACK_BOUNDARY.
5358
5359 2018-03-19  Richard Biener  <rguenther@suse.de>
5360
5361         PR tree-optimization/84933
5362         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5363         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5364
5365 2018-03-19  Richard Biener  <rguenther@suse.de>
5366
5367         PR tree-optimization/84859
5368         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5369         (cond_if_else_store_replacement): Perform sinking operation on
5370         single-store BBs regardless of MAX_STORES_TO_SINK setting.
5371         Generalize what a BB with a single eligible store is.
5372
5373 2018-03-19  Richard Biener  <rguenther@suse.de>
5374
5375         PR tree-optimization/84929
5376         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5377         chrec_is_positive against non-chrec arg.
5378
5379 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
5380
5381         PR target/84711
5382         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5383
5384 2018-03-18  Martin Liska  <mliska@suse.cz>
5385
5386         PR rtl-optimization/84635
5387         * regrename.c (build_def_use): Use matches_mode only when
5388         matches >= 0.
5389
5390 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
5391
5392         PR tree-optimization/84913
5393         * tree-vect-loop.c (vectorizable_reduction): Don't try to
5394         vectorize chains of COND_EXPRs.
5395
5396 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
5397
5398         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5399
5400 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
5401
5402         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5403
5404 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
5405
5406         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5407
5408 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
5409             Kito Cheng  <kito.cheng@gmail.com>
5410
5411         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5412         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5413         (nds32_adjust_reg_alloc_order): New function.
5414         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5415
5416 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
5417
5418         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5419         nds32_print_operand, nds32_print_operand_address): Use
5420         HOST_WIDE_INT_PRINT_DEC instead.
5421
5422 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
5423
5424         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5425
5426 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
5427
5428         PR target/84902
5429         * config/i386/i386.c (initial_ix86_tune_features,
5430         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5431         unsigned long long.
5432         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5433         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5434         rather than 1u << ix86_tune.  Formatting fix.
5435         (ix86_option_override_internal): Change ix86_arch_mask from
5436         unsigned int to unsigned HOST_WIDE_INT, initialize to
5437         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5438         (ix86_function_specific_restore): Likewise.
5439
5440 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
5441
5442         PR target/84899
5443         * postreload.c (reload_combine_recognize_pattern): Perform
5444         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5445         truncate_int_for_mode the result for the destination's mode.
5446
5447         PR c/84909
5448         * hsa-gen.c (mem_type_for_type): Fix comment typo.
5449         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5450         Likewise.
5451         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5452         Likewise.
5453
5454 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
5455
5456         PR target/84876
5457         * lra-assigns.c (lra_split_hard_reg_for): Don't use
5458         regno_allocno_class_array and sorted_pseudos.
5459         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5460         insns where regno is used.
5461
5462 2018-03-16  Martin Liska  <mliska@suse.cz>
5463
5464         PR ipa/84833
5465         * multiple_target.c (create_dispatcher_calls): Redirect
5466         reference in the symbol table.
5467
5468 2018-03-16  Martin Liska  <mliska@suse.cz>
5469
5470         PR ipa/84722
5471         * multiple_target.c (create_dispatcher_calls): Redirect also
5472         an alias.
5473
5474 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
5475
5476         PR c++/79937
5477         PR c++/82410
5478         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5479         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5480         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5481
5482 2018-03-16  Julia Koval  <julia.koval@intel.com>
5483
5484         * doc/invoke.texi (Skylake Server): Add CLWB.
5485         Cannonlake): Remove CLWB.
5486
5487 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
5488
5489         PR tree-optimization/84841
5490         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5491         1 << 3.
5492         (FLOAT_ONE_CONST_TYPE): Define.
5493         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5494         (sort_by_operand_rank): Put entries with higher constant_type last
5495         rather than first to match comments.
5496
5497 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
5498
5499         * config/nios2/nios2.md (movsi_internal): Fix thinko in
5500         split predicate.
5501
5502 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
5503
5504         PR c++/79085
5505         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5506         check and use address of target always.
5507
5508 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
5509
5510         PR target/84574
5511         * config/i386/i386.c (indirect_thunk_needed): Update comments.
5512         (indirect_thunk_bnd_needed): Likewise.
5513         (indirect_thunks_used): Likewise.
5514         (indirect_thunks_bnd_used): Likewise.
5515         (indirect_return_needed): New.
5516         (indirect_return_bnd_needed): Likewise.
5517         (output_indirect_thunk_function): Add a bool argument for
5518         function return.
5519         (output_indirect_thunk_function): Don't generate alias for
5520         function return thunk.
5521         (ix86_code_end): Call output_indirect_thunk_function to generate
5522         function return thunks.
5523         (ix86_output_function_return): Set indirect_return_bnd_needed
5524         and indirect_return_needed instead of indirect_thunk_bnd_needed
5525         and indirect_thunk_needed.
5526
5527 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
5528
5529         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5530         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5531         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5532
5533 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
5534             Paul Hua <paul.hua.gm@gmail.com>
5535
5536         PR c/84852
5537         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5538
5539 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
5540
5541         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5542         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5543         resp. SFmode cases.
5544
5545 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
5546
5547         PR target/84711
5548         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5549         instead of GET_MODE_SIZE when comparing Units.
5550
5551 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
5552
5553         PR target/68256
5554         * varasm.c (hash_section): Return an unchangeble hash value
5555         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5556         Return !aarch64_can_use_per_function_literal_pools_p ().
5557
5558 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
5559
5560         PR target/84860
5561         * optabs.c (emit_conditional_move): Pass address of cmode's copy
5562         rather than address of cmode as last argument to prepare_cmp_insn.
5563
5564 2018-03-15  Julia Koval  <julia.koval@intel.com>
5565
5566         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5567         F_AVX512VNNI, F_AVX512BITALG): New.
5568
5569 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
5570
5571         PR target/83451
5572         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
5573         insn for floating-point loads and stores.
5574
5575 2018-03-14  Carl Love  <cel@us.ibm.com>
5576
5577         * config/rs6000/rs6000-c.c: Add macro definitions for
5578         ALTIVEC_BUILTIN_VEC_PERMXOR.
5579         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5580         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5581         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5582         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5583         UNSPEC_VPERMXOR.
5584         * config/doc/extend.texi: Add prototypes for vec_permxor.
5585
5586 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
5587
5588         PR c/84852
5589         * diagnostic-show-locus.c (class layout_point): Convert m_line
5590         from int to linenum_type.
5591         (line_span::comparator): Use linenum "compare" function when
5592         comparing line numbers.
5593         (test_line_span): New function.
5594         (layout_range::contains_point): Convert param "row" from int to
5595         linenum_type.
5596         (layout_range::intersects_line_p): Likewise.
5597         (layout::will_show_line_p): Likewise.
5598         (layout::print_source_line): Likewise.
5599         (layout::should_print_annotation_line_p): Likewise.
5600         (layout::print_annotation_line): Likewise.
5601         (layout::print_leading_fixits): Likewise.
5602         (layout::annotation_line_showed_range_p): Likewise.
5603         (struct line_corrections): Likewise for field m_row.
5604         (line_corrections::line_corrections): Likewise for param "row".
5605         (layout::print_trailing_fixits): Likewise.
5606         (layout::get_state_at_point): Likewise.
5607         (layout::get_x_bound_for_row): Likewise.
5608         (layout::print_line): Likewise.
5609         (diagnostic_show_locus): Likewise for locals "last_line" and
5610         "row".
5611         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5612         * input.c (selftest::test_linenum_comparisons): New function.
5613         (selftest::input_c_tests): Call it.
5614         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5615         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5616         * selftest.h (ASSERT_GT): New macro.
5617         (ASSERT_GT_AT): New macro.
5618         (ASSERT_LT): New macro.
5619         (ASSERT_LT_AT): New macro.
5620
5621 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
5622
5623         PR rtl-optimization/84780
5624         * combine.c (distribute_links): Don't make a link based on pc_rtx.
5625
5626 2018-03-14  Martin Liska  <mliska@suse.cz>
5627
5628         * tree.c (record_node_allocation_statistics): Use
5629         get_stats_node_kind.
5630         (get_stats_node_kind): New function extracted from
5631         record_node_allocation_statistics.
5632         (free_node): Use get_stats_node_kind.
5633
5634 2018-03-14  Richard Biener  <rguenther@suse.de>
5635
5636         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5637         that the value-set of ANTIC_IN doesn't grow.
5638
5639         Revert
5640         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5641         member.
5642         (BB_VISITED_WITH_VISITED_SUCCS): New define.
5643         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5644
5645 2018-03-14  Julia Koval  <julia.koval@intel.com>
5646
5647         * config.gcc (icelake-client, icelake-server): New.
5648         (icelake): Remove.
5649         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5650         (initial_ix86_arch_features): Ditto.
5651         (PTA_SKYLAKE): Add SGX.
5652         (PTA_ICELAKE): Remove.
5653         (PTA_ICELAKE_CLIENT): New.
5654         (PTA_ICELAKE_SERVER): New.
5655         (ix86_option_override_internal): Split up icelake on icelake client and
5656         icelake server.
5657         (get_builtin_code_for_version): Ditto.
5658         (fold_builtin_cpu): Ditto.
5659         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5660         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5661         * config/i386/i386.h (processor_type): Ditto.
5662         * doc/invoke.texi: Ditto.
5663
5664 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
5665
5666         PR sanitizer/83392
5667         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5668         INTEGER_CST offset, add it together with bitpos / 8 and
5669         sign extend based on POINTER_SIZE.
5670
5671         PR target/84844
5672         Revert
5673         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
5674
5675         PR target/78090
5676         * config/i386/constraints.md (Yc): New register constraint.
5677         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5678         Use Yc constraint for alternative 2 of operand 0.  Remove
5679         preferred_for_speed attribute.
5680
5681 2018-03-14  Richard Biener  <rguenther@suse.de>
5682
5683         PR tree-optimization/84830
5684         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5685         with the old one to avoid oscillations.
5686
5687 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
5688
5689         PR target/83712
5690         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5691         pseudos.
5692         (assign_by_spills): Return a flag of reload assignment failure.
5693         Do not process the reload assignment failures.  Do not spill other
5694         reload pseudos if they has the same reg class.  Update n if
5695         necessary.
5696         (lra_assign): Add a return arg.  Set up from the result of
5697         assign_by_spills call.
5698         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5699         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
5700         usage_insns if it is not NULL.
5701         (spill_hard_reg_in_range): New function.
5702         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5703         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5704         function prototypes.
5705         (lra_assign): Change prototype.
5706         * lra.c (lra): Add code to deal with fails by splitting hard reg
5707         live ranges.
5708
5709 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
5710
5711         * config/riscv/riscv.opt (mrelax): New option.
5712         * config/riscv/riscv.c (riscv_file_start): Emit ".option
5713         "norelax" when riscv_mrelax is disabled.
5714         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5715
5716 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
5717
5718         PR target/84743
5719         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5720         reassociation for int modes.
5721
5722 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
5723
5724         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5725         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5726         for big-endian.
5727         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5728         * config/aarch64/aarch64-sve.md
5729         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5730         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5731         (*extend<mode><Vwide>2): Rename to...
5732         (aarch64_sve_extend<mode><Vwide>2): ...this.
5733         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5734         renaming the old pattern to...
5735         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
5736         unsigned packs.
5737         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5738         define_expand, renaming the old pattern to...
5739         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5740         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5741         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5742         account when deciding which SVE instruction the optab should use.
5743         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5744
5745 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
5746
5747         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5748         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5749         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5750         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5751         (tlsdesc_small_<mode>): Turn a define_expand and use
5752         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
5753         (tlsdesc_small_advsimd_<mode>): ...this.
5754         (tlsdesc_small_sve_<mode>): New pattern.
5755
5756 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
5757
5758         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5759         (UNSPEC_UMUL_HIGHPART): New constants.
5760         (MUL_HIGHPART): New int iteraor.
5761         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5762         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5763         define_expand.
5764         (*<su>mul<mode>3_highpart): New define_insn.
5765
5766 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
5767
5768         PR lto/84805
5769         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5770         incomplete types.
5771
5772 2018-03-13  Martin Liska  <mliska@suse.cz>
5773
5774         PR ipa/84658.
5775         * (sem_item_optimizer::sem_item_optimizer): Initialize new
5776         vector.
5777         (sem_item_optimizer::~sem_item_optimizer): Release it.
5778         (sem_item_optimizer::merge_classes): Register variable aliases.
5779         (sem_item_optimizer::fixup_pt_set): New function.
5780         (sem_item_optimizer::fixup_points_to_sets): Likewise.
5781         * ipa-icf.h: Declare new variables and functions.
5782
5783 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
5784
5785         PR middle-end/84834
5786         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5787         integer_pow2p@2 and test integer_pow2p in condition.
5788         (A < 0 ? C : 0): Similarly for @1.
5789
5790         PR middle-end/84831
5791         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5792         characters starting at p contain '\0' character, don't look beyond
5793         that.
5794
5795         PR target/84827
5796         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5797         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5798
5799         PR target/84828
5800         * reg-stack.c (change_stack): Change update_end var from int to
5801         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5802         also call set_block_for_insn on the newly added insns and rescan.
5803
5804         PR target/84786
5805         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5806         on the last operand.
5807
5808         PR c++/84704
5809         * tree.c (stabilize_reference_1): Return save_expr (e) for
5810         STATEMENT_LIST even if it doesn't have side-effects.
5811
5812 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
5813
5814         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5815
5816 2018-03-12  Renlin Li  <renlin.li@arm.com>
5817
5818         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5819         aarch64_output_scalar_simd_mov_immediate.
5820
5821 2018-03-12  Martin Sebor  <msebor@redhat.com>
5822
5823         PR tree-optimization/83456
5824         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5825         for perfectly overlapping calls to memcpy.
5826         (gimple_fold_builtin_memory_chk): Same.
5827         (gimple_fold_builtin_strcpy): Handle no-warning.
5828         (gimple_fold_builtin_stxcpy_chk): Same.
5829         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5830
5831 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
5832
5833         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5834         parameter.  Use it for SFmode.
5835         (rs6000_function_arg_advance_1): Adjust.
5836         (rs6000_function_arg): Adjust.
5837         (rs6000_gimplify_va_arg): Pass false for that new parameter.
5838
5839 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
5840
5841         PR rtl-optimization/84169
5842         PR rtl-optimization/84780
5843         * combine.c (can_combine_p): Check for a 2-insn combination whether
5844         the destination register is used between the two insns, too.
5845
5846 2018-03-12  Richard Biener  <rguenther@suse.de>
5847
5848         PR tree-optimization/84803
5849         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5850         for refs DR analysis didn't process.
5851
5852 2018-03-12  Richard Biener  <rguenther@suse.de>
5853
5854         PR tree-optimization/84777
5855         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5856         force-vectorize loops ignore whether we are optimizing for size.
5857
5858 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
5859
5860         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5861         (TARGET_MD_ASM_ADJUST): Define.
5862
5863 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
5864             Kito Cheng  <kito.cheng@gmail.com>
5865             Chung-Ju Wu  <jasonwucj@gmail.com>
5866
5867         * config/nds32/nds32.c (nds32_compute_stack_frame,
5868         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5869         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5870         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5871         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5872         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5873         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5874         * config/nds32/nds32.md (prologue, epilogue): Use macro
5875         NDS32_V3PUSH_AVAILABLE_P to do checking.
5876
5877 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
5878
5879         PR debug/58150
5880         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5881         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5882         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
5883         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
5884         addition of most attributes on !orig_type_die or the attribute not
5885         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5886
5887 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
5888             Chung-Ju Wu  <jasonwucj@gmail.com>
5889
5890         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5891         __NDS32_VH__ macro.
5892         * config/nds32/nds32.opt (mvh): New option.
5893
5894 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
5895             Chung-Ju Wu  <jasonwucj@gmail.com>
5896
5897         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5898         function.
5899         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5900         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5901         definition.
5902
5903 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
5904             Chung-Ju Wu  <jasonwucj@gmail.com>
5905
5906         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5907         function.
5908         * config/nds32/nds32-multiple.md (strlensi): New pattern.
5909         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5910
5911 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
5912             Kito Cheng  <kito.cheng@gmail.com>
5913             Chung-Ju Wu  <jasonwucj@gmail.com>
5914
5915         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5916         UNSPEC_FFMISM and UNSPEC_FLMISM.
5917         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5918         for ffb, ffmism and flmism.
5919         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5920         (unspec_ffmism): Ditto.
5921         (unspec_flmism): Ditto.
5922         (nds32_expand_builtin_impl): Check if string extension is available.
5923         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5924         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5925
5926 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
5927
5928         Reverting patch:
5929         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
5930
5931         PR target/83712
5932         * lra-assigns.c (assign_by_spills): Return a flag of reload
5933         assignment failure.  Do not process the reload assignment
5934         failures.  Do not spill other reload pseudos if they has the same
5935         reg class.
5936         (lra_assign): Add a return arg.  Set up from the result of
5937         assign_by_spills call.
5938         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5939         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
5940         usage_insns if it is not NULL.
5941         (spill_hard_reg_in_range): New function.
5942         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5943         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5944         function prototypes.
5945         (lra_assign): Change prototype.
5946         * lra.c (lra): Add code to deal with fails by splitting hard reg
5947         live ranges.
5948
5949 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
5950
5951         PR target/84807
5952         * config/i386/i386.opt: Replace Enforcment with Enforcement.
5953
5954 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
5955
5956         PR debug/84620
5957         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5958         (dw_val_node): Add val_symbolic_view.
5959         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5960         (symview_upper_bound): New.
5961         (new_line_info_table): Initialize symviews_since_reset.
5962         (dwarf2out_source_line): Count symviews_since_reset and set
5963         symview_upper_bound.
5964         (dw_val_equal_p): Handle symview.
5965         (add_AT_symview): New.
5966         (print_dw_val): Handle symview.
5967         (attr_checksum, attr_checksum_ordered): Likewise.
5968         (same_dw_val_p, size_of_die): Likewise.
5969         (value_format, output_die): Likewise.
5970         (add_high_low_attributes): Use add_AT_symview for entry_view.
5971         (dwarf2out_finish): Reset symview_upper_bound, clear
5972         zero_view_p.
5973
5974 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
5975
5976         PR target/83969
5977         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5978         Add strict argument and use it.
5979         (rs6000_split_multireg_move): Update for new strict argument.
5980         (mem_operand_gpr): Disallow all non-offsettable addresses.
5981         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5982
5983 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
5984
5985         PR target/84772
5986         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5987         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5988         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5989
5990         PR c++/84767
5991         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5992         decl, use remap_type if we want to use the type.
5993
5994 2018-03-09  Martin Sebor  <msebor@redhat.com>
5995
5996         PR tree-optimization/84526
5997         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5998         Remove dead code.
5999         (builtin_access::generic_overlap): Be prepared to handle non-array
6000         base objects.
6001
6002 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
6003
6004         PR rtl-optimization/84682
6005         * lra-constraints.c (process_address_1): Check is_address flag
6006         for address constraints.
6007         (process_alt_operands): Likewise.
6008         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6009         preprocess_constraints.
6010         * recog.h (preprocess_constraints): Add oploc parameter.
6011         Adjust callers.
6012         * recog.c (preprocess_constraints): Test address_operand for
6013         CT_ADDRESS constraints.
6014
6015 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
6016
6017         PR target/83712
6018         * lra-assigns.c (assign_by_spills): Return a flag of reload
6019         assignment failure.  Do not process the reload assignment
6020         failures.  Do not spill other reload pseudos if they has the same
6021         reg class.
6022         (lra_assign): Add a return arg.  Set up from the result of
6023         assign_by_spills call.
6024         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6025         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
6026         usage_insns if it is not NULL.
6027         (spill_hard_reg_in_range): New function.
6028         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6029         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6030         function prototypes.
6031         (lra_assign): Change prototype.
6032         * lra.c (lra): Add code to deal with fails by splitting hard reg
6033         live ranges.
6034
6035 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6036
6037         PR target/83193
6038         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6039         Accept complain bool parameter.  Only emit errors if it is true.
6040         (arm_parse_cpu_option_name): Likewise.
6041         (arm_target_thumb_only): Adjust callers of the above.
6042         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6043         prototype to take a default true bool parameter.
6044         (arm_parse_arch_option_name): Likewise.
6045
6046 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
6047             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
6048
6049         PR jit/64089
6050         PR jit/84288
6051         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6052         * configure: Regenerate.
6053         * configure.ac ("linker --version-script option"): New.
6054         ("linker soname option"): New.
6055
6056 2018-03-09  Richard Biener  <rguenther@suse.de>
6057
6058         PR tree-optimization/84775
6059         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6060         immediate uses of predicate stmts and mark them modified.
6061
6062         Revert
6063         PR tree-optimization/84178
6064         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6065         to caller.
6066         (version_loop_for_if_conversion): Delay update_ssa call.
6067         (tree_if_conversion): Delay update_ssa until after predicate
6068         insertion.
6069
6070 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
6071
6072         PR target/84763
6073         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6074         when the function accesses prior frames.
6075
6076 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
6077
6078         PR debug/84456
6079         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6080         gen_llsym, otherwise call maybe_gen_llsym.
6081
6082         PR inline-asm/84742
6083         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6084         has ',' character inside of it.
6085
6086 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6087
6088         PR target/84748
6089         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6090         as clobbering CC_REGNUM.
6091
6092 2018-03-08  Richard Biener  <rguenther@suse.de>
6093
6094         PR middle-end/84552
6095         * tree-scalar-evolution.c: Include tree-into-ssa.h.
6096         (follow_copies_to_constant): Do not follow SSA names registered
6097         for update.
6098
6099 2018-03-08  Richard Biener  <rguenther@suse.de>
6100
6101         PR tree-optimization/84178
6102         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6103         to caller.
6104         (version_loop_for_if_conversion): Delay update_ssa call.
6105         (tree_if_conversion): Delay update_ssa until after predicate
6106         insertion.
6107
6108 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
6109
6110         PR tree-optimization/84178
6111         * tree-if-conv.c (release_bb_predicate): Remove the
6112         the assertion that the stmts have NULL use_ops.
6113         Discard the statements, asserting that they haven't
6114         yet been added to a BB.
6115
6116 2018-03-08  Richard Biener  <rguenther@suse.de>
6117
6118         PR tree-optimization/84746
6119         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6120         (phi_translate): Pass in destination ANTIC_OUT set.
6121         (phi_translate_1): Likewise.  For a simplified result lookup
6122         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6123         (phi_translate_set): Adjust.
6124         (do_pre_regular_insertion): Likewise.
6125         (do_pre_partial_partial_insertion): Likewise.
6126
6127 2018-03-08  Martin Liska  <mliska@suse.cz>
6128
6129         PR gcov-profile/84735
6130         * doc/gcov.texi: Document usage of profile files.
6131         * gcov-io.h: Document changes in the format.
6132
6133 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
6134
6135         PR debug/84404
6136         PR debug/84408
6137         * dwarf2out.c (struct dw_line_info_table): Update comments for
6138         view == -1.
6139         (FORCE_RESET_NEXT_VIEW): New.
6140         (FORCE_RESETTING_VIEW_P): New.
6141         (RESETTING_VIEW_P): Check for -1 too.
6142         (ZERO_VIEW_P): Likewise.
6143         (new_line_info_table): Force-reset next view.
6144         (dwarf2out_begin_function): Likewise.
6145         (dwarf2out_source_line): Simplify zero_view_p initialization.
6146         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6147         view directly.  Omit view when omitting .loc at line 0.
6148
6149 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
6150
6151         PR tree-optimization/84740
6152         * tree-switch-conversion.c (process_switch): Call build_constructors
6153         only if info.phi_count is non-zero.
6154
6155         PR tree-optimization/84739
6156         * tree-tailcall.c (find_tail_calls): Check call arguments against
6157         DECL_ARGUMENTS (current_function_decl) rather than
6158         DECL_ARGUMENTS (func) when checking for tail recursion.
6159
6160 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
6161
6162         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6163         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6164         Volker Reichelt's entry and add entries for people that perform
6165         GCC fuzzy testing and report numerous bugs.
6166
6167 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
6168
6169         PR target/82411
6170         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6171         readonly data in sdata, if that is disabled.
6172         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6173         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6174         -mreadonly-in-sdata option.
6175
6176 2018-03-07  Martin Sebor  <msebor@redhat.com>
6177
6178         PR tree-optimization/84468
6179         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6180         basic block when looking for nul assignment.
6181
6182 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
6183
6184         PR target/84277
6185         * except.h (output_function_exception_table): Adjust prototype.
6186         * except.c (output_function_exception_table): Remove FNNAME parameter
6187         and add SECTION parameter.  Ouput one part of the table at a time.
6188         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6189         the first part of the exception table and emit unwind directives.
6190         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6191         (i386_pe_seh_cold_init): Likewise.
6192         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6193         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6194         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6195         (ix86_output_call_insn): Emit a nop in one more case for SEH.
6196         * config/i386/winnt.c: Include except.h.
6197         (struct seh_frame_state): Add reg_offset, after_prologue and
6198         in_cold_section fields.
6199         (i386_pe_seh_end_prologue): Set seh->after_prologue.
6200         (i386_pe_seh_cold_init): New function.
6201         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6202         to seh->in_cold_section.
6203         (seh_emit_push): Record the offset of the push.
6204         (seh_emit_save): Record the offet of the save.
6205         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6206         Test seh->after_prologue to disregard the epilogue.
6207         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6208         (i386_pe_end_cold_function): New function.
6209
6210 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
6211
6212         PR fortran/84565
6213         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6214         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6215
6216         PR c++/84704
6217         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6218         on tmp_var.
6219         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6220         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6221
6222         PR middle-end/84723
6223         * multiple_target.c: Include tree-inline.h and intl.h.
6224         (expand_target_clones): Diagnose and fail if node->definition and
6225         !tree_versionable_function_p (node->decl).
6226
6227 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
6228
6229         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6230         sprint_ul.
6231         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6232         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6233         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6234
6235 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
6236
6237         PR target/84710
6238         * combine.c (try_combine): Use reg_or_subregno instead of handling
6239         just paradoxical SUBREGs and REGs.
6240
6241 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
6242
6243          * config/arc/arc.c (arc_finalize_pic): Remove function.
6244          (arc_must_save_register): We use single base PIC register, remove
6245          checks to save/restore the PIC register.
6246          (arc_expand_prologue): Likewise.
6247          * config/arc/arc-protos.h (arc_set_default_type_attributes):
6248          Remove.
6249          (arc_verify_short): Likewise.
6250          (arc_attr_type): Likewise.
6251          * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6252          (walk_stores): Likewise.
6253          (arc_address_cost): Make it static.
6254          (arc_verify_short): Likewise.
6255          (branch_dest): Likewise.
6256          (arc_attr_type): Likewise.
6257          * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6258          (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6259          (arc_final_prescan_insn): Remove inserting the nops due to
6260          hardware hazards.  It is done in reorg step.
6261          (insn_length_variant_t): Remove.
6262          (insn_length_parameters_t): Likewise.
6263          (arc_insn_length_parameters): Likewise.
6264          (arc_get_insn_variants): Likewise.
6265          * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6266
6267 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
6268
6269         PR inline-asm/84683
6270         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6271         assertion failure.
6272
6273         PR tree-optimization/84687
6274         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6275         on new_node->decl.
6276         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6277
6278 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
6279
6280         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6281         Rename to ppc_speculation_barrier.
6282         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6283         __builtin_ppc_speculation_barrier.
6284
6285 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
6286
6287         PR target/84700
6288         * combine.c (combine_simplify_rtx): Don't try to simplify if
6289         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6290         are equal to x.
6291
6292 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
6293
6294         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6295         to 32 bytes when compiling for POWER9.
6296
6297 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
6298
6299         PR target/84564
6300         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6301         regparm >= 3 with no arg reg available also for calls with
6302         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
6303
6304         PR target/84524
6305         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6306         orig,vex.
6307         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
6308
6309 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
6310
6311         PR target/84264
6312         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6313
6314 2018-03-05  Richard Biener  <rguenther@suse.de>
6315
6316         PR tree-optimization/84486
6317         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6318         When inserting a __builtin_assume_aligned call set the LHS
6319         SSA name alignment info accordingly.
6320
6321 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
6322
6323         PR tree-optimization/84114
6324         * config/aarch64/aarch64.c (aarch64_reassociation_width)
6325         Avoid reassociation of FLOAT_MODE addition.
6326
6327 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
6328
6329         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6330         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6331         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6332         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6333         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6334         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6335         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6336         and -mwbnoinvd.
6337         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6338         __builtin_ia32_wbinvd): New builtins.
6339         (SPECIAL_ARGS2): New.
6340         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6341         (SPECIAL_ARGS2): New.
6342         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6343         (ix86_valid_target_attribute_inner_p): Ditto.
6344         (ix86_init_mmx_sse_builtins): Add special_args2.
6345         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6346         TARGET_WBNOINVD_P): New.
6347         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6348         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6349         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6350         * config/i386/immintrin.h (_wbinvd): New intrinsic.
6351         * config/i386/pconfigintrin.h: New file.
6352         * config/i386/wbnoinvdintrin.h: Ditto.
6353         * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6354         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6355
6356 2018-03-05  Richard Biener  <rguenther@suse.de>
6357
6358         PR tree-optimization/84670
6359         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6360         member.
6361         (BB_VISITED_WITH_VISITED_SUCCS): New define.
6362         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6363         (compute_antic_aux): Only assert the number of values in ANTIC_IN
6364         doesn't grow if all successors (recursively) were visited at least
6365         once.
6366
6367 2018-03-05  Richard Biener  <rguenther@suse.de>
6368
6369         PR tree-optimization/84650
6370         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6371         if executed in the loop pipeline.
6372
6373 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
6374
6375         * doc/configfiles.texi (Configuration Files): Move info about
6376         conditionalizing $target-protos.h to...
6377         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
6378         differs from $target-protos.h.
6379
6380 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
6381             Chung-Ju Wu  <jasonwucj@gmail.com>
6382
6383         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6384         * config/nds32/nds32-multiple.md (setmemsi): Define.
6385         * config/nds32/nds32-memory-manipulation.c
6386         (nds32_gen_dup_4_byte_to_word_value): New.
6387         (emit_setmem_word_loop): New.
6388         (emit_setmem_byte_loop): New.
6389         (nds32_expand_setmem_loop): New.
6390         (nds32_expand_setmem_loop_v3m): New.
6391         (nds32_expand_setmem_unroll): New.
6392         (nds32_expand_setmem): New.
6393
6394 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
6395             Chung-Ju Wu  <jasonwucj@gmail.com>
6396
6397         * config/nds32/nds32-memory-manipulation.c
6398         (nds32_emit_load_store): New.
6399         (nds32_emit_post_inc_load_store): New.
6400         (nds32_emit_mem_move): New.
6401         (nds32_emit_mem_move_block): New.
6402         (nds32_expand_movmemsi_loop_unknown_size): New.
6403         (nds32_expand_movmemsi_loop_known_size): New.
6404         (nds32_expand_movmemsi_loop): New.
6405         (nds32_expand_movmemsi_unroll): New.
6406         (nds32_expand_movmemqi): Rename ...
6407         (nds32_expand_movmemsi): ... to this.
6408         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6409         (movmemsi): ... to this.
6410         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6411         (nds32_expand_movmemsi): ... to this.
6412
6413 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
6414             Monk Chiang  <sh.chiang04@gmail.com>
6415             Chung-Ju Wu  <jasonwucj@gmail.com>
6416
6417         * config/nds32/nds32-protos.h
6418         (nds32_expand_load_multiple): New arguments.
6419         (nds32_expand_store_multiple): Ditto.
6420         (nds32_valid_multiple_load_store): Rename ...
6421         (nds32_valid_multiple_load_store_p): ... to this.
6422         * config/nds32/nds32-memory-manipulation.c
6423         (nds32_expand_load_multiple): Refine implementation.
6424         (nds32_expand_store_multiple): Ditto.
6425         * config/nds32/nds32-multiple.md
6426         (load_multiple): Update nds32_expand_load_multiple interface.
6427         (store_multiple): Update nds32_expand_store_multiple interface.
6428         * config/nds32/nds32-predicates.c
6429         (nds32_valid_multiple_load_store): Rename ...
6430         (nds32_valid_multiple_load_store_p): ... to this and refine
6431         implementation.
6432         * config/nds32/predicates.md
6433         (nds32_load_multiple_and_update_address_operation): New predicate.
6434         (nds32_store_multiple_and_update_address_operation): New predicate.
6435
6436 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
6437             Chung-Ju Wu  <jasonwucj@gmail.com>
6438
6439         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6440         (combo): New attribute.
6441         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6442
6443 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
6444
6445         * config/nds32/nds32.opt: Change -mcmodel= default value.
6446
6447 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
6448             Monk Chiang  <sh.chiang04@gmail.com>
6449             Chung-Ju Wu  <jasonwucj@gmail.com>
6450
6451         * config/nds32/constants.md (unspec_element): New enum.
6452         * config/nds32/constraints.md (Umw): New constraint.
6453         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6454         * config/nds32/nds32-intrinsic.md: Likewise.
6455         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6456         (nds32_valid_smw_lwm_base_p): New.
6457         (nds32_output_smw_single_word): New.
6458         (nds32_output_lmw_single_word): New.
6459         (nds32_expand_unaligned_load): New.
6460         (nds32_expand_unaligned_store): New.
6461         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6462         (nds32_output_smw_single_word): Declare.
6463         (nds32_output_lmw_single_word): Declare.
6464         (nds32_expand_unaligned_load): Declare.
6465         (nds32_expand_unaligned_store): Declare.
6466         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6467         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6468         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6469         NDS32_BUILTIN_UASTORE_DW.
6470         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6471         predicate.
6472
6473 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
6474             Kito Cheng  <kito.cheng@gmail.com>
6475             Chung-Ju Wu  <jasonwucj@gmail.com>
6476
6477         * config/nds32/nds32-intrinsic.c
6478         (nds32_expand_builtin_null_ftype_reg): Delete.
6479         (nds32_expand_builtin_reg_ftype_imm): Ditto.
6480         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6481         (nds32_read_argument): New.
6482         (nds32_legitimize_target): Ditto.
6483         (nds32_legitimize_argument): Ditto.
6484         (nds32_check_constant_argument): Ditto.
6485         (nds32_expand_unop_builtin): Ditto.
6486         (nds32_expand_unopimm_builtin): Ditto.
6487         (nds32_expand_binop_builtin): Ditto.
6488         (nds32_builtin_decl_impl): Ditto.
6489         (builtin_description): Ditto.
6490         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6491         (nds32_init_builtins_impl): Ditto.
6492         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6493         (nds32_builtin_decl): New.
6494         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6495         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6496
6497 2018-03-02  Jeff Law  <law@redhat.com>
6498
6499         * reorg.c (stop_search_p): Handle DEBUG_INSN.
6500         (redundant_insn, fill_simple_delay_slots): Likewise.
6501         (fill_slots_from_thread): Likewise.
6502         * resource.c (mark_referenced_resources): Likewise.
6503         (mark_set_resources, find_dead_or_set_registers): Likewise.
6504
6505 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
6506
6507         * substring-locations.h (format_warning_va): Formatting fix for
6508         ATTRIBUTE_GCC_DIAG.
6509         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6510         argument.
6511         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6512         * substring-locations.c: Include intl.h.
6513         (format_warning_va): Turned into small wrapper around
6514         format_warning_n_va, renamed to ...
6515         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6516         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6517         use ngettext.
6518         (format_warning_at_substring_n): New function.
6519         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6520         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
6521         format_warning_at_substring with just a shorter name instead of
6522         const function pointer.
6523         (fmtwarn_n): New function.
6524         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6525         appropriate, get rid of all the fmtstr temporaries, move conditionals
6526         with G_() wrapped string literals directly into fmtwarn arguments,
6527         cast dir.len to (int), formatting fixes.
6528
6529 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
6530
6531         * doc/invoke.texi: Remove "Cilk Plus" references.
6532
6533 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
6534             Richard Biener  <rguenther@suse.de>
6535
6536         PR ipa/84628
6537         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6538         for error or warning attributes if CALL_FROM_THUNK_P is set.
6539         Formatting fixes.
6540
6541 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
6542
6543         PR target/56540
6544         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6545         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6546
6547         PR target/56540
6548         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6549         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6550
6551         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6552         instead of -1U in last predictors element's probability member.
6553
6554 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
6555
6556         PR ipa/83983
6557         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6558         arguments if they are comparable.
6559
6560 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
6561
6562         PR tree-optimization/84634
6563         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6564         masks and masked_loop_p with a single loop_masks, making sure it's
6565         null for bb vectorization.
6566
6567 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
6568
6569         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6570         (vect_analyze_data_ref_access): Use loop->safe_len rather than
6571         loop->force_vectorize to check whether there is no alias.
6572
6573 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
6574
6575         PR target/84614
6576         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6577         prototypes.
6578         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6579         comments.
6580         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6581         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6582         instead of a loop around prev_real_insn.
6583         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6584         prev_real_insn.
6585
6586         PR inline-asm/84625
6587         * config/i386/i386.c (ix86_print_operand): Use conditional
6588         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6589         zero vector.
6590
6591 2018-03-02  Richard Biener  <rguenther@suse.de>
6592
6593         PR tree-optimization/84427
6594         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6595         (bitmap_set_subtract_values): Rewrite to handle multiple
6596         exprs per value.
6597         (clean): Likewise.
6598         (prune_clobbered_mems): Likewise.
6599         (phi_translate): Take edge instead of pred/phiblock.
6600         (phi_translate_1): Likewise.
6601         (phi_translate_set): Likewise.  Insert all translated
6602         exprs for a value into the set, keeping possibly multiple
6603         expressions per value.
6604         (compute_antic_aux): Adjust for phi_translate changes.
6605         When intersecting union the expressions and prune those
6606         not in the final value set, keeping possibly multiple
6607         expressions per value.  Do not use value-insertion
6608         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6609         all expressions.  Add verification that the value-sets
6610         only shrink during iteration.
6611         (compute_partial_antic_aux): Adjust for the phi_translate changes.
6612         (do_pre_regular_insertion): Likewise.
6613         (do_pre_partial_partial_insertion): Likewise.
6614
6615 2018-03-02  Richard Biener  <rguenther@suse.de>
6616
6617         PR target/82005
6618         * config/darwin.c (saved_debug_info_level): New static global.
6619         (darwin_asm_lto_start): Disable debug info generation for LTO out.
6620         (darwin_asm_lto_end): Restore debug info generation settings.
6621
6622 2018-03-01  Martin Liska  <mliska@suse.cz>
6623
6624         PR sanitizer/82484
6625         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6626         volatile arguments.
6627
6628 2018-03-01  Richard Biener  <rguenther@suse.de>
6629
6630         PR debug/84645
6631         * dwarf2out.c (gen_variable_die): Properly handle late VLA
6632         type annotation with LTO when debug was disabled at compile-time.
6633
6634 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
6635
6636         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6637         XINT with INTVAL.
6638         (mips_final_postscan_insn): Likewise.
6639
6640 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
6641
6642         PR rtl-optimization/84528
6643         * alias.c (init_alias_target): Add commentary.
6644         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6645         a unique base value if the frame pointer is not eliminated
6646         to the stack pointer.
6647
6648 2018-03-01  Tom de Vries  <tom@codesourcery.com>
6649
6650         PR rtl-optimization/83327
6651         * lra-int.h (hard_regs_spilled_into): Declare.
6652         * lra.c (hard_regs_spilled_into): Define.
6653         (init_reg_info): Init hard_regs_spilled_into.
6654         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6655         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6656         (process_bb_lives): Handle hard_regs_spilled_into.
6657         (lra_create_live_ranges_1): Before doing liveness propagation, clear
6658         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6659
6660 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
6661
6662         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6663         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6664         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6665         * config/rs6000/aix72.h: New file.
6666
6667 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
6668
6669         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6670         instead of warning_at with conditional singular and plural messages
6671         where possible.
6672
6673         PR target/52991
6674         * stor-layout.c (update_alignment_for_field): For
6675         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6676         && !DECL_PACKED (field), do the alignment update, just use
6677         only desired_align instead of MAX (type_align, desired_align)
6678         as the alignment.
6679         (place_field): Don't do known_align < desired_align handling
6680         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6681         is non-NULL, instead do it after rli->prev_field handling and
6682         only if not within a bitfield word.  For DECL_PACKED (field)
6683         use type_align of BITS_PER_UNIT.
6684
6685 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
6686
6687         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6688         superfluous parentheses and trailing spaces.
6689
6690 2018-02-28  Richard Biener  <rguenther@suse.de>
6691
6692         PR tree-optimization/84584
6693         * graphite-scop-detection.c (scop_detection::add_scop): Discard
6694         SCoPs with fake exit edge.
6695
6696 2018-02-28  Martin Liska  <mliska@suse.cz>
6697
6698         PR testsuite/84597
6699         * timevar.c (timer::print): Fix format to properly print 100%
6700         values.
6701
6702 2018-02-28  Richard Biener  <rguenther@suse.de>
6703
6704         PR middle-end/84607
6705         * genmatch.c (capture_info::walk_match): Do not mark
6706         captured expressions without operands as expr_p given
6707         they act more like predicates and should be subject to
6708         "lost tail" side-effect preserving.
6709
6710 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
6711
6712         PR rtl-optimization/81611
6713         * auto-inc-dec.c (attempt_change): Move dead note from
6714         mem_insn if it's the next use of regno
6715         (find_address): Take address use of reg holding
6716         non-incremented value.  Add parm to limit search to the named
6717         reg only.
6718         (merge_in_block): Attempt to use a mem insn that is the next
6719         use of the original regno.
6720
6721 2018-02-27  Martin Sebor  <msebor@redhat.com>
6722
6723         PR c++/83871
6724         * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6725         * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6726
6727 2018-02-27  Martin Sebor  <msebor@redhat.com>
6728
6729         PR translation/84207
6730         * diagnostic-core.h (warning_n, error_n, inform_n): Change
6731         n argument to unsigned HOST_WIDE_INT.
6732         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6733         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
6734         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6735         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6736
6737 2018-02-27  Richard Biener  <rguenther@suse.de>
6738
6739         PR tree-optimization/84512
6740         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6741         Do not use the estimate returned from record_stmt_cost for
6742         the scalar iteration cost but sum properly using add_stmt_cost.
6743
6744 2018-02-27  Richard Biener  <rguenther@suse.de>
6745
6746         PR tree-optimization/84466
6747         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6748         Adjust last change to less strictly validate use operands.
6749
6750 2018-02-27  Martin Liska  <mliska@suse.cz>
6751
6752         PR gcov-profile/84548
6753         * gcov.c (process_file): Allow partial overlap and consider it
6754         also as group functions.
6755         (output_lines): Properly calculate range of lines for a group.
6756
6757 2018-02-27  Martin Liska  <mliska@suse.cz>
6758
6759         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6760         'ggc' suffixes.  Change first column width.
6761         (timer::print): Fix formatting of the column.
6762
6763 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
6764
6765         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6766         preserve inline entry blocks for the sake of debug inline
6767         entry point markers alone.
6768         (remove_unused_locals): Suggest in comments a better place to
6769         force the preservation of inline entry blocks that are
6770         otherwise unused, but do not preserve them.
6771
6772 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
6773
6774         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6775
6776 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
6777
6778         PR target/84039
6779         * config/i386/constraints.md (Bs): Replace
6780         ix86_indirect_branch_register with
6781         TARGET_INDIRECT_BRANCH_REGISTER.
6782         (Bw): Likewise.
6783         * config/i386/i386.md (indirect_jump): Likewise.
6784         (tablejump): Likewise.
6785         (*sibcall_memory): Likewise.
6786         (*sibcall_value_memory): Likewise.
6787         Peepholes of indirect call and jump via memory: Likewise.
6788         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6789         (*sibcall_value_GOT_32): Likewise.
6790         * config/i386/predicates.md (indirect_branch_operand): Likewise.
6791         (GOT_memory_operand): Likewise.
6792         (call_insn_operand): Likewise.
6793         (sibcall_insn_operand): Likewise.
6794         (GOT32_symbol_operand): Likewise.
6795         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6796
6797 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
6798
6799         PR rtl-optimization/83496
6800         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6801         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
6802         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6803         redundant insn, if any.
6804         (relax_delay_slots): Likewise.
6805         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6806
6807 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
6808
6809         PR tree-optimization/83965
6810         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6811         that grouped statements are part of a reduction chain.  Return
6812         true if the statement is not marked as a reduction itself but
6813         is part of a group.
6814         (vect_recog_dot_prod_pattern): Don't check whether the statement
6815         is part of a group here.
6816         (vect_recog_sad_pattern): Likewise.
6817         (vect_recog_widen_sum_pattern): Likewise.
6818
6819 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
6820
6821         PR debug/84545
6822         * final.c (rest_of_clean_state): Also look for calls inside sequences.
6823
6824 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
6825
6826         PR target/84530
6827         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6828         the bool argument.
6829         (ix86_output_indirect_function_return): New prototype.
6830         (ix86_split_simple_return_pop_internal): Likewise.
6831         * config/i386/i386.c (indirect_return_via_cx): New.
6832         (indirect_return_via_cx_bnd): Likewise.
6833         (indirect_thunk_name): Handle return va CX_REG.
6834         (output_indirect_thunk_function): Create alias for
6835         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6836         (ix86_output_indirect_jmp): Remove the bool argument.
6837         (ix86_output_indirect_function_return): New function.
6838         (ix86_split_simple_return_pop_internal): Likewise.
6839         * config/i386/i386.md (*indirect_jump): Don't pass false
6840         to ix86_output_indirect_jmp.
6841         (*tablejump_1): Likewise.
6842         (simple_return_pop_internal): Change it to define_insn_and_split.
6843         Call ix86_split_simple_return_pop_internal to split it for
6844         -mfunction-return=.
6845         (simple_return_indirect_internal): Call
6846         ix86_output_indirect_function_return instead of
6847         ix86_output_indirect_jmp.
6848
6849 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
6850
6851         PR bootstrap/84405
6852         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6853         memset and value initialization afterwards.
6854
6855 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
6856
6857         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6858
6859 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
6860
6861         PR target/84521
6862         * common/config/aarch64/aarch64-common.c
6863         (aarch_option_optimization_table[]): Switch
6864         off fomit-frame-pointer
6865
6866 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
6867             Chung-Ju Wu  <jasonwucj@gmail.com>
6868
6869         * config/nds32/nds32-multiple.md (load_multiple): Disallow
6870         volatile memory.
6871         (store_multiple): Ditto.
6872
6873 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
6874
6875         * config.gcc: Add --with-cpu support for nds32 target.
6876         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6877         * config/nds32/nds32.opt: Add -mcpu= option.
6878
6879 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
6880
6881         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6882         isel=yes): Warn for these deprecated options.
6883
6884 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
6885
6886         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6887         ISA_2_5_MASKS_EMBEDDED.
6888
6889 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
6890
6891         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6892         p->max as pointers rather than using iterative_hash_expr.
6893
6894 2018-02-23  Carl Love  <cel@us.ibm.com>
6895
6896         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6897         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6898         BU_P8V_OVERLOAD_2.
6899         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6900         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6901         P8V_BUILTIN_VEC_VUNSIGNED2.
6902
6903 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
6904
6905         PR target/81572
6906         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6907         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6908         LRA_UNKNOWN_ALT.
6909         * lra-constraints.c (curr_insn_transform): Set up
6910         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
6911         LRA_UNKNOWN_ALT.
6912         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6913         * lra-eliminations.c (spill_pseudos): Ditto.
6914         (process_insn_for_elimination): Ditto.
6915         * lra-lives.c (reg_early_clobber_p): Use the new macros.
6916         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6917         LRA_NON_CLOBBERED_ALT.
6918
6919 2018-02-22  Martin Sebor  <msebor@redhat.com>
6920
6921         PR tree-optimization/84480
6922         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6923         to maybe_diag_stxncpy_trunc.  Call it.
6924         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6925         from gimple_fold_builtin_strcpy.  Print inlining stack.
6926         (handle_builtin_stxncpy): Print inlining stack.
6927         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6928
6929 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
6930
6931         PR target/84176
6932         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6933         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6934         and -fcheck-pointer-bounds are used together.
6935         (indirect_thunk_prefix): New enum.
6936         (indirect_thunk_need_prefix): New function.
6937         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
6938         "_nt" instead of "_bnd" for NOTRACK prefix.
6939         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6940         (output_indirect_thunk_function): Likewise.
6941         (): Likewise.
6942         (ix86_code_end): Update output_indirect_thunk_function calls.
6943         (ix86_output_indirect_branch_via_reg): Replace
6944         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6945         (ix86_output_indirect_branch_via_push): Likewise.
6946         (ix86_output_function_return): Likewise.
6947         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6948         incompatible with -fcf-protection=branch and
6949         -fcheck-pointer-bounds.
6950
6951 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
6952
6953         PR target/83335
6954         * config/aarch64/aarch64.c (aarch64_print_address_internal):
6955         Change gcc_assert call to output_operand_lossage.
6956
6957 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
6958
6959         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6960
6961 2018-02-22  DJ Delorie  <dj@redhat.com>
6962             Sebastian Perta  <sebastian.perta@renesas.com>
6963             Oleg Endo  <olegendo@gcc.gnu.org>
6964
6965         * config/rx/rx.c (rx_rtx_costs): New function.
6966         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6967
6968 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6969
6970         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6971
6972 2018-02-22  Martin Liska  <mliska@suse.cz>
6973
6974         PR driver/83193
6975         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6976         Add "native" as a possible value.
6977
6978 2018-02-22  Martin Liska  <mliska@suse.cz>
6979
6980         PR driver/83193
6981         * config/i386/i386.c (ix86_option_override_internal):
6982         Add "native" as a possible value for -march and -mtune.
6983
6984 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
6985
6986         PR target/84502
6987         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6988         to all type variants.
6989
6990         PR tree-optimization/84503
6991         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6992         width as info->bitpos + info->bitsize - start.
6993         (merged_store_group::merge_overlapping): Simplify width computation.
6994         (check_no_overlap): New function.
6995         (imm_store_chain_info::try_coalesce_bswap): Compute expected
6996         start + width and last_order of the group, fail if check_no_overlap
6997         fails.
6998         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6999         to group if check_no_overlap fails.
7000
7001 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
7002
7003         * config/rs6000/altivec.md: Delete contraint arguments to
7004         define_expand, define_split, and define_peephole2, and in
7005         define_insn_and_split if always unused.
7006         * config/rs6000/darwin.md: Ditto.
7007         * config/rs6000/dfp.md: Ditto.
7008         * config/rs6000/rs6000.md: Ditto.
7009         * config/rs6000/sync.md: Ditto.
7010         * config/rs6000/vector.md: Ditto.
7011         * config/rs6000/vsx.md: Ditto.
7012
7013 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
7014
7015         * config/rs6000/altivec.md: Write output control strings as braced
7016         blocks instead of double-quoted strings.
7017         * config/rs6000/darwin.md: Ditto.
7018         * config/rs6000/rs6000.md: Ditto.
7019         * config/rs6000/vector.md: Ditto.
7020         * config/rs6000/vsx.md: Ditto.
7021
7022 2018-02-21  Jason Merrill  <jason@redhat.com>
7023
7024         PR c++/84314 - ICE with templates and fastcall attribute.
7025         * attribs.c (build_type_attribute_qual_variant): Remove assert.
7026
7027 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
7028
7029         * ipa-cp.c (determine_versionability): Fix comment typos.
7030
7031 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
7032
7033         PR c/84229
7034         * ipa-cp.c (determine_versionability): Do not version functions caling
7035         va_arg_pack.
7036
7037 2018-02-21  Martin Liska  <mliska@suse.cz>
7038
7039         PR driver/83193
7040         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7041         Add "native" as a possible value.
7042         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
7043         the macro when native cpu detection is available.
7044
7045 2018-02-21  Martin Liska  <mliska@suse.cz>
7046
7047         PR driver/83193
7048         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7049         Add "native" as a possible value.
7050         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7051         when native cpu detection is available.
7052
7053 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
7054             Martin Sebor  <msebor@redhat.com>
7055
7056         PR tree-optimization/84478
7057         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7058         false.
7059         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7060         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7061         support which is conservatively correct, for 2 only stay conservative
7062         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
7063         argument to the 2 argument get_range_strlen, adjust 6 arg
7064         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7065         false.
7066         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7067         (gimple_fold_builtin_strlen): Pass true as last argument to
7068         get_range_strlen.
7069
7070 2018-02-20  Martin Sebor  <msebor@redhat.com>
7071
7072         PR middle-end/84095
7073         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7074         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
7075         (builtin_memref::builtin_memref): Factor out parts into
7076         set_base_and_offset and call it.
7077
7078 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
7079
7080         PR middle-end/84406
7081         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7082         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7083         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
7084         search at the associated MODE_INT.
7085
7086 2018-02-20  Jeff Law  <law@redhat.com>
7087
7088         PR middle-end/82123
7089         PR tree-optimization/81592
7090         PR middle-end/79257
7091         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7092         for range data rather than using global data.
7093         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7094         range data rather than using global data.
7095         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7096         pass it to children as needed.
7097         (struct directive::fmtresult): Similarly.
7098         (struct directive::set_width): Similarly.
7099         (struct directive::set_precision): Similarly.
7100         (format_integer, format_directive, parse_directive): Similarly.
7101         (format_none): Accept unnamed vr_values parameter.
7102         (format_percent, format_floating, format_character): Similarly.
7103         (format_string, format_plain): Similarly.
7104         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7105         the EVRP range analyzer for range data rather than using global data.
7106         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7107         gimple-ssa-evrp-analyze.h
7108         (class sprintf_dom_walker): Add after_dom_children member function.
7109         Add evrp_range_analyzer member.
7110         (sprintf_dom_walker::before_dom_children): Call into the EVRP
7111         range analyzer as needed.
7112         (sprintf_dom_walker::after_dom_children): New member function.
7113         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7114         if not optimizing.
7115         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7116         (evrp_range_analyzer::pop_to_marker): Likewise.
7117
7118 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
7119
7120         PR tree-optimization/84419
7121         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7122         with the required type if its current type is compatible but
7123         different.
7124
7125 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
7126
7127         PR middle-end/82004
7128         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7129         after vectorization.
7130
7131 2018-02-20  Martin Liska  <mliska@suse.cz>
7132
7133         PR driver/83193
7134         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7135         possible values if we don't have a hint.
7136
7137 2018-02-20  Martin Liska  <mliska@suse.cz>
7138
7139         PR c/84310
7140         PR target/79747
7141         * final.c (shorten_branches): Build align_tab array with one
7142         more element.
7143         * opts.c (finish_options): Add alignment option limit check.
7144         (MAX_CODE_ALIGN): Likewise.
7145         (MAX_CODE_ALIGN_VALUE): Likewise.
7146         * doc/invoke.texi: Document maximum allowed option value for
7147         all -falign-* options.
7148
7149 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
7150
7151         PR target/84146
7152         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7153         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7154         * var-tracking.c (emit_note_insn_var_location): Remove all references
7155         to NOTE_INSN_CALL_ARG_LOCATION.
7156         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7157         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7158         Use copy_rtx_if_shared.
7159         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7160         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7161         (dwarf2out_var_location): Remove handling of
7162         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7163         on call_insn.
7164         * final.c (final_scan_insn): Remove all references to
7165         NOTE_INSN_CALL_ARG_LOCATION.
7166         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
7167         before dumping final insns.
7168         * except.c (emit_note_eh_region_end): Remove all references to
7169         NOTE_INSN_CALL_ARG_LOCATION.
7170         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7171         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7172         * config/arc/arc.c (hwloop_optimize): Likewise.
7173         * config/arm/arm.c (create_fix_barrier): Likewise.
7174         * config/s390/s390.c (s390_chunkify_start): Likewise.
7175         * config/sh/sh.c (find_barrier): Likewise.
7176         * config/i386/i386.c (rest_of_insert_endbranch,
7177         ix86_seh_fixup_eh_fallthru): Likewise.
7178         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7179         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7180         * config/frv/frv.c (frv_function_prologue): Likewise.
7181         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
7182         reg note.
7183         (note_outside_basic_block_p): Remove all references to
7184         NOTE_INSN_CALL_ARG_LOCATION.
7185         * gengtype.c (adjust_field_rtx_def): Likewise.
7186         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7187         Likewise.
7188         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7189         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7190
7191         PR c++/84444
7192         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7193         is ADDR_EXPR.
7194
7195         PR tree-optimization/84452
7196         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7197         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7198         is NULL.
7199
7200 2018-02-19  Martin Liska  <mliska@suse.cz>
7201
7202         PR sanitizer/82183
7203         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7204
7205 2018-02-19  Martin Liska  <mliska@suse.cz>
7206             Richard Sandiford  <richard.sandiford@linaro.org>
7207
7208         PR tree-optimization/82491
7209         * gimple-fold.c (get_base_constructor): Make earlier bail out
7210         to prevent ubsan.
7211
7212 2018-02-19  Carl Love  <cel@us.ibm.com>
7213
7214         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7215         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7216         BU_P8V_OVERLOAD_1.
7217         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7218         P8V_BUILTIN_VEC_NEG.
7219
7220 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
7221
7222         * config/rl78/rl78.md (movdf): New define expand.
7223
7224 2018-02-19  Martin Liska  <mliska@suse.cz>
7225
7226         PR other/80589
7227         * doc/invoke.texi: Fix typo.
7228         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7229
7230 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
7231
7232         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7233         handle rs6000_single_float and rs6000_double_float specially for
7234         e500 family CPUs.
7235
7236 2018-02-16  Jeff Law  <law@redhat.com>
7237
7238         * config/rx/rx.c (add_pop_cfi_notes): New function.;
7239         (pop_regs): Use it.
7240
7241 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
7242
7243         PR ipa/84425
7244         * ipa-inline.c (inline_small_functions): Fix a typo.
7245
7246 2018-02-16  Nathan Sidwell  <nathan@acm.org>
7247
7248         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7249
7250 2018-02-16  Carl Love  <cel@us.ibm.com>
7251
7252         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7253         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7254         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7255         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7256         expansion to P8V_BUILTIN_VEC_FLOAT2.
7257
7258 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
7259
7260         PR rtl-optimization/70023
7261         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7262         src_regno into account.
7263
7264 2018-02-16  Carl Love  <cel@us.ibm.com>
7265
7266         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7267         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7268         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7269         * config/rs6000/rs6000.c: Remove case statements for
7270         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7271         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7272         and P9V_BUILTIN_VEC_VINSERT4B.
7273         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7274         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7275         * config/rs6000/vsx.md:
7276         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7277         vec_insert4b.
7278
7279 2018-02-16  Carl Love  <cel@us.ibm.com>
7280
7281         * config/rs6000/altivec.h: Add builtin names vec_extract4b
7282         vec_insert4b.
7283         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7284         definitions.
7285         * config/rs6000/rs6000-c.c: Add the definitions for
7286         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7287         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7288         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7289         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
7290         definition for insert4b and define insn *insert3b_internal.
7291         * doc/extend.texi: Add documentation for vec_extract4b.
7292
7293 2018-02-16  Nathan Sidwell  <nathan@acm.org>
7294
7295         * doc/extend.texi (Backwards Compatibility): Mention friend
7296         injection.  Note for-scope is deprecated.
7297         * doc/invoke.texi (-ffriend-injection): Deprecate.
7298
7299 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
7300
7301         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7302         that moved to I2, also allow destinations that are a paradoxical
7303         subreg (instead of a normal reg).
7304
7305 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
7306
7307         PR target/83831
7308         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7309         to QImode.
7310
7311 2018-02-16  Richard Biener  <rguenther@suse.de>
7312
7313         PR tree-optimization/84037
7314         PR tree-optimization/84016
7315         PR target/82862
7316         * config/i386/i386.c (ix86_builtin_vectorization_cost):
7317         Adjust vec_construct for the fact we need additional higher latency
7318         128bit inserts for AVX256 and AVX512 vector builds.
7319         (ix86_add_stmt_cost): Scale vector construction cost for
7320         elementwise loads.
7321
7322 2018-02-16  Richard Biener  <rguenther@suse.de>
7323
7324         PR tree-optimization/84417
7325         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7326         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7327         (non_rewritable_lvalue_p): Likewise, use poly-ints.
7328
7329 2018-02-16  Martin Liska  <mliska@suse.cz>
7330
7331         PR sanitizer/84307
7332         * internal-fn.def (ASAN_CHECK): Set proper flags.
7333         (ASAN_MARK): Likewise.
7334
7335 2018-02-16  Julia Koval  <julia.koval@intel.com>
7336
7337         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7338         from PTA_CANNONLAKE.
7339
7340 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
7341
7342         PR target/84272
7343         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7344         Use ++iter rather than iter++ for std::list iterators.
7345         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
7346         defer deleting them until all nodes in the forest are processed.  Do
7347         free even leaf nodes.  Change to_process into auto_vec.
7348
7349         PR bootstrap/84405
7350         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7351         * vec.h (vec_default_construct): Use memset instead of placement new
7352         if BROKEN_VALUE_INITIALIZATION is defined.
7353         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7354         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7355         is defined.
7356
7357         PR rtl-optimization/83723
7358         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7359         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
7360         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
7361         recursive calls.
7362         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7363         callers.
7364         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7365
7366 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
7367
7368         PR rtl-optimization/81443
7369         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7370         from inner REGs to paradoxical SUBREGs.
7371
7372 2018-02-16  Richard Biener  <rguenther@suse.de>
7373
7374         PR tree-optimization/84399
7375         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7376         For operands we can analyze at their definition make sure we can
7377         analyze them at each use as well.
7378
7379 2018-02-16  Richard Biener  <rguenther@suse.de>
7380
7381         PR tree-optimization/84190
7382         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7383         volatile accesses if the decl isn't volatile.
7384
7385 2018-02-15  Jason Merrill  <jason@redhat.com>
7386
7387         PR c++/84314 - ICE with templates and fastcall attribute.
7388         * attribs.c (build_type_attribute_qual_variant): Don't clobber
7389         TYPE_CANONICAL on an existing type.
7390
7391 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
7392
7393         PR tree-optimization/84383
7394         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7395         dstoff nor call operand_equal_p if dstbase is NULL.
7396
7397         PR tree-optimization/84334
7398         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7399         also a CONSTANT_CLASS_P, punt.
7400
7401 2018-02-14  Jim Wilson  <jimw@sifive.com>
7402
7403         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7404         first SMALL_OPERAND check.  New local min_second_step.  Move assert
7405         to where locals are set.  Add TARGET_RVC support.
7406         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7407
7408 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
7409
7410         * doc/invoke.texi: Correct -Wformat-overflow code sample.
7411
7412 2018-02-14  Martin Sebor  <msebor@redhat.com>
7413
7414         PR tree-optimization/83698
7415         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7416         arrays constrain the offset range to their bounds.
7417         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7418         (builtin_access::overlap): Avoid setting the size of overlap if it's
7419         already been set.
7420         (maybe_diag_overlap): Also consider arrays when deciding what values
7421         of offsets to include in diagnostics.
7422
7423 2018-02-14  Martin Sebor  <msebor@redhat.com>
7424
7425         PR c/84108
7426         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7427         that correspond to the kind of a declaration.
7428
7429 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
7430
7431         PR target/83984
7432         * config/pa/pa.md: Load address of PIC label using the linkage table
7433         if the label is nonlocal.
7434
7435 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7436
7437         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7438         warning message if user requests -maltivec=be.
7439         * doc/invoke.texi: Document deprecation of -maltivec=be.
7440
7441 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
7442
7443         PR target/84220
7444         * config/rs6000/rs6000-c.c: Update definitions for
7445         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7446         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7447
7448 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
7449
7450         PR target/84239
7451         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7452         add _get_ssp intrinsics. Remove argument from
7453         __builtin_ia32_rdssp[d|q].
7454         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7455         * config/i386/i386-builtin.def: Remove argument from
7456         __builtin_ia32_rdssp[d|q].
7457         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7458         ix86_expand_special_args_builtin for _rdssp[d|q].
7459         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7460         Clear register before usage.
7461         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7462         Add documentation for new _get_ssp and _inc_ssp intrinsics.
7463
7464 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
7465
7466         PR tree-optimization/84357
7467         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7468         operand 1 of an ARRAY_REF too.
7469
7470 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
7471
7472         PR target/83831
7473         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7474         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7475         declarations.
7476         (set_of_reg): New struct.
7477         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7478         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7479         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7480         functions.
7481         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7482         Split into bitclr, bitset, bitinvert patterns if appropriate.
7483         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7484         use rx_fuse_in_memory_bitop.
7485         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7486         to named insn, correct maximum insn length.
7487
7488 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
7489
7490         PR target/79242
7491         * machmode.def: Define a complex mode for PARTIAL_INT.
7492         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7493         MODE_PARTIAL_INT.
7494         * doc/rtl.texi: Document CSPImode.
7495         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7496         handling.
7497         (msp430_hard_regno_nregs_with_padding): Likewise.
7498
7499 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
7500
7501         PR target/84279
7502         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7503
7504 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
7505
7506         PR rtl-optimization/84169
7507         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
7508         we generated a parallel as new i3 and we split that to new i2 and i3
7509         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
7510         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
7511         those to i2, not i1.  Partially rewrite this scan code.
7512
7513 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
7514
7515         PR c/82210
7516         * stor-layout.c (place_field): For variable length fields, adjust
7517         offset_align afterwards not just based on the field's alignment,
7518         but also on the size.
7519
7520         PR middle-end/84309
7521         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7522         of exps and logs in the use_exp2 case.
7523
7524 2018-02-13  Jeff Law  <law@redhat.com>
7525
7526         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7527         entry for "vector".
7528
7529         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7530         ARGS as unused.
7531
7532 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
7533
7534         PR debug/84342
7535         PR debug/84319
7536         * common.opt (gas-loc-support, gas-locview-support): New.
7537         (ginline-points, ginternal-reset-location-views): New.
7538         * doc/invoke.texi: Document them.  Use @itemx where intended.
7539         (gvariable-location-views): Adjust.
7540         * target.def (reset_location_view): New.
7541         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7542         (TARGET_RESET_LOCATION_VIEW): New.
7543         * doc/tm.texi: Rebuilt.
7544         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7545         (dwarf2out_default_as_locview_support): New.
7546         (output_asm_line_debug_info): Use option variables.
7547         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7548         (output_loc_list): Likewise.
7549         (add_high_low_attributes): Check option variables.
7550         Don't output entry view attribute in strict mode.
7551         (gen_inlined_subroutine_die): Check option variables.
7552         (dwarf2out_inline_entry): Likewise.
7553         (init_sections_and_labels): Likewise.
7554         (dwarf2out_early_finish): Likewise.
7555         (maybe_reset_location_view): New, from...
7556         (dwarf2out_var_location): ... here.  Call it.
7557         * debug.h (dwarf2out_default_as_loc_support): Declare.
7558         (dwarf2out_default_as_locview_support): Declare.
7559         * hooks.c (hook_int_rtx_insn_0): New.
7560         * hooks.h (hook_int_rtx_insn_0): Declare.
7561         * toplev.c (process_options): Take -gas-loc-support and
7562         -gas-locview-support from dwarf2out.  Enable
7563         -gvariable-location-views by default only with locview
7564         assembler support.  Enable -ginternal-reset-location-views by
7565         default only if the target defines the corresponding hook.
7566         Enable -ginline-points by default if location views are
7567         enabled; force it disabled if statement frontiers are
7568         disabled.
7569         * tree-inline.c (expand_call_inline): Check option variables.
7570         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7571
7572 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
7573
7574         PR tree-optimization/84321
7575         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7576         handling.  Also check whether the anti-range contains any values
7577         that satisfy the mask; switch to a VR_RANGE if not.
7578
7579 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
7580
7581         PR sanitizer/84340
7582         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7583
7584 2018-02-13  Martin Jambor  <mjambor@suse.cz>
7585
7586         PR c++/83990
7587         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7588         of call statements, also set location of a load to a temporary.
7589
7590 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
7591
7592         * config/rl78/rl78.c (add_vector_labels): New function.
7593         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7594         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7595         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7596         which checks that no arguments are passed.
7597         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7598         * doc/extend.texi: Documentation for the new attribute.
7599
7600 2018-02-13  Andreas Schwab  <schwab@suse.de>
7601
7602         * config/riscv/linux.h (CPP_SPEC): Define.
7603
7604 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
7605
7606         PR target/84335
7607         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7608         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7609         OPTION_MASK_ISA_AES as first argument to def_builtin_const
7610         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7611         instead of OPTION_MASK_ISA_PCLMUL as first argument to
7612         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7613         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7614         temporarily for AES and PCLMUL builtins.
7615
7616         PR tree-optimization/84339
7617         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7618         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7619         Formatting fixes.
7620
7621         PR middle-end/84309
7622         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7623         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7624         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7625         inline function.
7626         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7627         inline function.
7628         * omp-simd-clone.h: New file.
7629         * omp-simd-clone.c: Include omp-simd-clone.h.
7630         (expand_simd_clones): No longer static.
7631         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7632         cgraph.h and omp-simd-clone.h.
7633         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7634         (vect_recog_widen_shift_pattern): Formatting fix.
7635         (vect_pattern_recog_1): Don't check optab for calls.
7636
7637         PR target/84336
7638         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7639         operands[2] into a REG before using gen_lowpart on it.
7640
7641 2018-02-12  Jeff Law  <law@redhat.com>
7642
7643         PR target/83760
7644         * config/sh/sh.c (find_barrier): Consider a sibling call
7645         a barrier as well.
7646
7647         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7648         successfully back substituting a reg.
7649
7650 2018-02-12  Richard Biener  <rguenther@suse.de>
7651
7652         PR tree-optimization/84037
7653         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7654         parameter, move visited init to caller.
7655         (vect_slp_analyze_operations): Separate cost from validity
7656         check, initialize visited once for all instances.
7657         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7658         for all instances.
7659         * tree-vect-stmts.c (vect_model_simple_cost): Make early
7660         out an assert.
7661         (vect_model_promotion_demotion_cost): Likewise.
7662         (vectorizable_bswap): Guard cost modeling with !slp_node
7663         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7664         SLP stmts.
7665         (vectorizable_call): Likewise.
7666         (vectorizable_conversion): Likewise.
7667         (vectorizable_assignment): Likewise.
7668         (vectorizable_shift): Likewise.
7669         (vectorizable_operation): Likewise.
7670         (vectorizable_store): Likewise.
7671         (vectorizable_load): Likewise.
7672         (vectorizable_condition): Likewise.
7673         (vectorizable_comparison): Likewise.
7674
7675 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
7676
7677         PR sanitizer/84307
7678         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7679         (ASAN_MARK): Fix fnspec to account for return value, change pointer
7680         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7681
7682 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
7683
7684         PR middle-end/83665
7685         * params.def (inline-min-speedup): Increase from 8 to 15.
7686         (max-inline-insns-auto): Decrease from 40 to 30.
7687         * ipa-split.c (consider_split): Add some buffer for function to
7688         be considered inlining candidate.
7689         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7690         default values.
7691
7692 2018-02-12  Richard Biener  <rguenther@suse.de>
7693
7694         PR tree-optimization/84037
7695         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7696         matched stmts if we cannot swap the non-matched ones.
7697
7698 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
7699
7700         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7701         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7702         _mm_maskz_scalef_round_ss): New intrinsics.
7703         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7704         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7705         __builtin_ia32_scalefss_round): Remove.
7706         (__builtin_ia32_scalefsd_mask_round,
7707         __builtin_ia32_scalefss_mask_round): New intrinsics.
7708         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7709         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7710         ((match_operand:VF_128 2 "<round_nimm_predicate>"
7711         "<round_constraint>")): Changed to ...
7712         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7713         "<round_scalar_constraint>")): ... this.
7714         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7715         %0, %1, %2<round_op3>}"): Changed to ...
7716         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7717         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7718         %2<round_scalar_mask_op3>}"): ... this.
7719         * config/i386/subst.md (round_scalar_nimm_predicate): New.
7720
7721 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
7722
7723         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7724         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7725         (_mm_maskz_sqrt_round_ss): New intrinsics.
7726         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7727         (__builtin_ia32_sqrtsd_mask_round)
7728         (__builtin_ia32_sqrtss_mask_round): New builtins.
7729         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7730         (__builtin_ia32_sqrtss_round): Remove.
7731         (__builtin_ia32_sqrtsd_mask_round)
7732         (__builtin_ia32_sqrtss_mask_round): New builtins.
7733         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7734         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7735         ((match_operand:VF_128 1 "vector_operand"
7736         "xBm,<round_constraint>")): Changed to ...
7737         ((match_operand:VF_128 1 "vector_operand"
7738         "xBm,<round_scalar_constraint>")): ... this.
7739         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7740         %0, %2, %<iptr>1<round_op3>}): Changed to ...
7741         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7742         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7743         %<iptr>1<round_scalar_mask_op3>}): ... this.
7744         ((set_attr "prefix" "<round_prefix>")): Changed to ...
7745         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7746
7747 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
7748
7749         PR target/84266
7750         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7751         Cast vec_cmpeq result to correct type.
7752         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7753         Cast vec_cmpgt result to correct type.
7754
7755 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
7756
7757         * final.c (final_scan_insn_1): Renamed from...
7758         (final_scan_insn): ... this.  New wrapper, to recover
7759         seen from the outermost call in recursive ones.
7760         * config/sparc/sparc.c (output_return): Drop seen from call.
7761         (output_sibcall): Likewise.
7762         * config/visium/visium.c (output_branch): Likewise.
7763
7764 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
7765
7766         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7767         function label.
7768
7769 2018-02-10  Alan Modra  <amodra@gmail.com>
7770
7771         PR target/84300
7772         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7773         Specify LR as an input.
7774
7775 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
7776
7777         PR sanitizer/83987
7778         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7779         remove_member_access_dummy_vars): New functions.
7780         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7781         lower_omp_1, execute_lower_omp): Use them.
7782
7783         PR rtl-optimization/84308
7784         * shrink-wrap.c (spread_components): Release todo vector.
7785
7786 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
7787
7788         PR rtl-optimization/57193
7789         * ira-color.c (struct allocno_color_data): Add member
7790         conflict_allocno_hard_prefs.
7791         (update_conflict_allocno_hard_prefs): New.
7792         (bucket_allocno_compare_func): Add a preference based on
7793         conflict_allocno_hard_prefs.
7794         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7795         (color_allocnos): Remove a dead code.  Initiate
7796         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
7797
7798 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
7799
7800         PR target/84226
7801         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7802         constraint from =wa to wa.  Avoid a subreg on the output operand,
7803         instead use a pseudo and subreg it in a move.
7804         (p9_xxbrd_<mode>): Changed to ...
7805         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7806         (p9_xxbrd_v2df): New expander.
7807         (p9_xxbrw_<mode>): Changed to ...
7808         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7809         (p9_xxbrw_v4sf): New expander.
7810
7811 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
7812
7813         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7814
7815 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
7816
7817         PR target/83926
7818         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7819         multiply in 32-bit mode.
7820         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7821         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7822         mode.
7823
7824 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
7825
7826         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7827         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7828         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7829         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7830
7831 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
7832
7833         PR lto/84213
7834         * dwarf2out.c (is_trivial_indirect_ref): New function.
7835         (dwarf2out_late_global_decl): Do not generate a location
7836         attribute for variables that have a non-trivial DECL_VALUE_EXPR
7837         and that are not defined in the current unit.
7838
7839 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
7840
7841         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7842         instead of a libcall for UNORDERED.
7843
7844 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
7845
7846         PR target/82641
7847         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7848         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7849
7850 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
7851
7852         PR target/PR84295
7853         * config/s390/s390.c (s390_set_current_function): Invoke
7854         s390_indirect_branch_settings also if fndecl didn't change.
7855
7856 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
7857
7858         * config/rs6000/rs6000.md (blockage): Set length to zero.
7859
7860 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
7861
7862         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7863
7864 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
7865
7866         PR sanitizer/84285
7867         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7868         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7869         -static-lib*san.
7870
7871         PR debug/84252
7872         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7873         PARALLEL incoming that failed vt_get_decl_and_offset check.
7874
7875         PR middle-end/84237
7876         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7877         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7878         TREE_READONLY bit.
7879         (get_variable_section): For decls in named .bss* sections pass true as
7880         second argument to bss_initializer_p.
7881
7882 2018-02-09  Marek Polacek  <polacek@redhat.com>
7883             Jakub Jelinek  <jakub@redhat.com>
7884
7885         PR c++/83659
7886         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7887         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
7888         Sync some changes from cxx_fold_indirect_ref.
7889
7890 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
7891
7892         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7893         markers.
7894         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7895         (BLOCK_INLINE_ENTRY_LABEL): New.
7896         (dwarf2out_var_location): Disregard inline entry markers.
7897         (inline_entry_data): New struct.
7898         (inline_entry_data_hasher): New hashtable type.
7899         (inline_entry_data_hasher::hash): New.
7900         (inline_entry_data_hasher::equal): New.
7901         (inline_entry_data_table): New variable.
7902         (add_high_low_attributes): Add DW_AT_entry_pc and
7903         DW_AT_GNU_entry_view attributes if a pending entry is found
7904         in inline_entry_data_table.  Add old entry_pc attribute only
7905         if debug nonbinding markers are disabled.
7906         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7907         markers are enabled.
7908         (block_within_block_p, dwarf2out_inline_entry): New.
7909         (dwarf2out_finish): Check that no entries remained in
7910         inline_entry_data_table.
7911         * final.c (reemit_insn_block_notes): Handle inline entry notes.
7912         (final_scan_insn, notice_source_line): Likewise.
7913         (rest_of_clean_state): Skip inline entry markers.
7914         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7915         markers.
7916         * gimple.c (gimple_build_debug_inline_entry): New.
7917         * gimple.h (enum gimple_debug_subcode): Add
7918         GIMPLE_DEBUG_INLINE_ENTRY.
7919         (gimple_build_debug_inline_entry): Declare.
7920         (gimple_debug_inline_entry_p): New.
7921         (gimple_debug_nonbind_marker_p): Adjust.
7922         * insn-notes.def (INLINE_ENTRY): New.
7923         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7924         inline entry marker notes.
7925         (print_insn): Likewise.
7926         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7927         (INSN_DEBUG_MARKER_KIND): Likewise.
7928         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7929         * tree-inline.c (expand_call_inline): Build and insert
7930         debug_inline_entry stmt.
7931         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7932         inline entry blocks early, if nonbind markers are enabled.
7933         (dump_scope_block): Dump fragment info.
7934         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7935         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7936         (gimple_build_debug_inline_entry): New.
7937         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7938         Enable/disable inline entry points too.
7939         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7940         (DEBUG_INSN): Describe inline entry markers.
7941
7942         * common.opt (gvariable-location-views): New.
7943         (gvariable-location-views=incompat5): New.
7944         * config.in: Rebuilt.
7945         * configure: Rebuilt.
7946         * configure.ac: Test assembler for view support.
7947         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7948         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7949         * dwarf2out.c (var_loc_view): New typedef.
7950         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7951         (dwarf2out_locviews_in_attribute): New.
7952         (dwarf2out_locviews_in_loclist): New.
7953         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7954         (enum dw_line_info_opcode): Add LI_adv_address.
7955         (struct dw_line_info_table): Add view.
7956         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7957         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7958         (zero_view_p): New variable.
7959         (ZERO_VIEW_P): New macro.
7960         (output_asm_line_debug_info): New.
7961         (struct var_loc_node): Add view.
7962         (add_AT_view_list, AT_loc_list): New.
7963         (add_var_loc_to_decl): Add view param.  Test it against last.
7964         (new_loc_list): Add view params.  Record them.
7965         (AT_loc_list_ptr): Handle loc and view lists.
7966         (view_list_to_loc_list_val_node): New.
7967         (print_dw_val): Handle dw_val_class_view_list.
7968         (size_of_die): Likewise.
7969         (value_format): Likewise.
7970         (loc_list_has_views): New.
7971         (gen_llsym): Set vl_symbol too.
7972         (maybe_gen_llsym, skip_loc_list_entry): New.
7973         (dwarf2out_maybe_output_loclist_view_pair): New.
7974         (output_loc_list): Output view list or entries too.
7975         (output_view_list_offset): New.
7976         (output_die): Handle dw_val_class_view_list.
7977         (output_dwarf_version): New.
7978         (output_compilation_unit_header): Use it.
7979         (output_skeleton_debug_sections): Likewise.
7980         (output_rnglists, output_line_info): Likewise.
7981         (output_pubnames, output_aranges): Update version comments.
7982         (output_one_line_info_table): Output view numbers in asm comments.
7983         (dw_loc_list): Determine current endview, pass it to new_loc_list.
7984         Call maybe_gen_llsym.
7985         (loc_list_from_tree_1): Adjust.
7986         (add_AT_location_description): Create view list attribute if
7987         needed, check it's absent otherwise.
7988         (convert_cfa_to_fb_loc_list): Adjust.
7989         (maybe_emit_file): Call output_asm_line_debug_info for test.
7990         (dwarf2out_var_location): Reset views as needed.  Precompute
7991         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
7992         attribute.  Set view.
7993         (new_line_info_table): Reset next view.
7994         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7995         (dwarf2out_source_line): Likewise.  Output view resets and labels to
7996         the assembler, or select appropriate line info opcodes.
7997         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7998         (optimize_string_length): Catch it.  Adjust.
7999         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
8000         dw_val_class_view_list, and remove it if no longer needed.
8001         (hash_loc_list): Hash view numbers.
8002         (loc_list_hasher::equal): Compare them.
8003         (optimize_location_lists): Check whether a view list symbol is
8004         needed, and whether the locview attribute is present, and
8005         whether they match.  Remove the locview attribute if no longer
8006         needed.
8007         (index_location_lists): Call skip_loc_list_entry for test.
8008         (dwarf2out_finish): Call output_asm_line_debug_info for test.
8009         Use output_dwarf_version.
8010         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8011         (struct dw_val_node): Add val_view_list.
8012         * final.c (SEEN_NEXT_VIEW): New.
8013         (set_next_view_needed): New.
8014         (clear_next_view_needed): New.
8015         (maybe_output_next_view): New.
8016         (final_start_function): Rename to...
8017         (final_start_function_1): ... this.  Take pointer to FIRST,
8018         add SEEN parameter.  Emit param bindings in the initial view.
8019         (final_start_function): Reintroduce SEEN-less interface.
8020         (final): Rename to...
8021         (final_1): ... this.  Take SEEN parameter.  Output final pending
8022         next view at the end.
8023         (final): Reintroduce seen-less interface.
8024         (final_scan_insn): Output pending next view before switching
8025         sections or ending a block.  Mark the next view as needed when
8026         outputting variable locations.  Notify debug backend of section
8027         changes, and of location view changes.
8028         (rest_of_handle_final): Adjust.
8029         * toplev.c (process_options): Autodetect value for debug variable
8030         location views option.  Warn on incompat5 without -gdwarf-5.
8031         * doc/invoke.texi (gvariable-location-views): New.
8032         (gvariable-location-views=incompat5): New.
8033         (gno-variable-location-views): New.
8034
8035 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
8036
8037         PR tree-optimization/84136
8038         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8039         that the result of find_edge is non-NULL.
8040
8041 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
8042
8043         PR target/83008
8044         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8045         storing integer register in SImode.  Fix cost of 256 and 512
8046         byte aligned SSE register store.
8047
8048 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
8049
8050         * config/i386/i386.c (ix86_multiplication_cost): Fix
8051         multiplication cost for TARGET_AVX512DQ.
8052
8053 2018-02-08  Marek Polacek  <polacek@redhat.com>
8054
8055         PR tree-optimization/84238
8056         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8057         get_range_strlen.
8058
8059 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
8060
8061         PR tree-optimization/84265
8062         * tree-vect-stmts.c (vectorizable_store): Don't treat
8063         VMAT_CONTIGUOUS accesses as grouped.
8064         (vectorizable_load): Likewise.
8065
8066 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
8067
8068         PR tree-optimization/81635
8069         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8070         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8071         (test_round_for_mask): New functions.
8072         (wide_int_cc_tests): Call test_round_for_mask.
8073         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8074         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8075         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8076         range returned by get_range_info.
8077
8078 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
8079
8080         PR ipa/81360
8081         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8082         * symtab.c: Include builtins.h
8083         (symtab_node::output_to_lto_symbol_table_p): Move here
8084         from lto-streamer-out.c:output_symbol_p.
8085         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8086         (output_symbol_p): Move all logic to symtab.c
8087         (produce_symtab): Update.
8088
8089 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
8090
8091         * config/s390/s390-opts.h (enum indirect_branch): Define.
8092         * config/s390/s390-protos.h (s390_return_addr_from_memory)
8093         (s390_indirect_branch_via_thunk)
8094         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8095         (enum s390_indirect_branch_type): Define.
8096         * config/s390/s390.c (struct s390_frame_layout, struct
8097         machine_function): Remove.
8098         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8099         (indirect_branch_table_label_no, indirect_branch_table_name):
8100         Define variables.
8101         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8102         (enum s390_indirect_branch_option): Define.
8103         (s390_return_addr_from_memory): New function.
8104         (s390_handle_string_attribute): New function.
8105         (s390_attribute_table): Add new attribute handler.
8106         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8107         (s390_indirect_branch_via_thunk): New function.
8108         (s390_indirect_branch_via_inline_thunk): New function.
8109         (s390_function_ok_for_sibcall): When jumping via thunk disallow
8110         sibling call optimization for non z10 compiles.
8111         (s390_emit_call): Force indirect branch target to be a single
8112         register.  Add r1 clobber for non-z10 compiles.
8113         (s390_emit_epilogue): Emit return jump via return_use expander.
8114         (s390_reorg): Handle JUMP_INSNs as execute targets.
8115         (s390_option_override_internal): Perform validity checks for the
8116         new command line options.
8117         (s390_indirect_branch_attrvalue): New function.
8118         (s390_indirect_branch_settings): New function.
8119         (s390_set_current_function): Invoke s390_indirect_branch_settings.
8120         (s390_output_indirect_thunk_function):  New function.
8121         (s390_code_end): Implement target hook.
8122         (s390_case_values_threshold): Implement target hook.
8123         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8124         macros.
8125         * config/s390/s390.h (struct s390_frame_layout)
8126         (struct machine_function): Move here from s390.c.
8127         (TARGET_INDIRECT_BRANCH_NOBP_RET)
8128         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8129         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8130         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8131         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8132         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8133         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8134         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8135         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8136         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8137         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8138         (mnemonic attribute): Add values which aren't recognized
8139         automatically.
8140         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8141         pattern for branch conversion.  Fix mnemonic attribute.
8142         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8143         indirect branch via thunk if requested.
8144         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8145         ("*indirect_jump"): Disable for branch conversion using out of
8146         line thunks.
8147         ("indirect_jump_via_thunk<mode>_z10")
8148         ("indirect_jump_via_thunk<mode>")
8149         ("indirect_jump_via_inlinethunk<mode>_z10")
8150         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8151         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8152         ("casesi_jump_via_inlinethunk<mode>_z10")
8153         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8154         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8155         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8156         ("*indirect2_jump"): Disable for branch conversion.
8157         ("casesi_jump"): Turn into expander and expand patterns for branch
8158         conversion.
8159         ("return_use"): New expander.
8160         ("*return"): Emit return via thunk and rename it to ...
8161         ("*return<mode>"): ... this one.
8162         * config/s390/s390.opt: Add new options and and enum for the
8163         option values.
8164
8165 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
8166
8167         * lra-constraints.c (match_reload): Unconditionally use
8168         gen_lowpart_SUBREG, rather than selecting between that
8169         and equivalent gen_rtx_SUBREG code.
8170
8171 2018-02-08  Richard Biener  <rguenther@suse.de>
8172
8173         PR tree-optimization/84233
8174         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8175         changed flag instead of boguously re-using phi_inserted.
8176
8177 2018-02-08  Martin Jambor  <mjambor@suse.cz>
8178
8179         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8180         static local variables.
8181
8182 2018-02-08  Richard Biener  <rguenther@suse.de>
8183
8184         PR tree-optimization/84278
8185         * tree-vect-stmts.c (vectorizable_store): When looking for
8186         smaller vector types to perform grouped strided loads/stores
8187         make sure the mode is supported by the target.
8188         (vectorizable_load): Likewise.
8189
8190 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
8191
8192         * config/aarch64/aarch64.c (aarch64_components_for_bb):
8193         Increase LDP/STP opportunities by adding adjacent callee-saves.
8194
8195 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
8196
8197         PR rtl-optimization/84068
8198         PR rtl-optimization/83459
8199         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8200
8201 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
8202
8203         PR tree-optimization/84224
8204         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8205         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8206         non-zero arguments.
8207
8208 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
8209
8210         PR target/84113
8211         * config/rs6000/altivec.md (*restore_world): Remove LR use.
8212         * config/rs6000/predicates.md (restore_world_operation): Adjust op
8213         count, remove one USE.
8214
8215 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
8216
8217         * doc/install.texi (Configuration): Document the
8218         --with-long-double-format={ibm,ieee} PowerPC configuration
8219         options.
8220
8221         PR target/84154
8222         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8223         Convert from define_expand to be define_insn_and_split.  Rework
8224         float/double/_Float128 conversions to QI/HI/SImode to work with
8225         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
8226         conversions to QI/HImode types did a store and then a load to
8227         truncate the value.  For conversions to VSX registers, don't split
8228         the insn, instead emit the code directly.  Use the code iterator
8229         any_fix to combine signed and unsigned conversions.
8230         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8231         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8232         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8233         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8234         (fix_<mode>di2_hw): Likewise.
8235         (fixuns_<mode>di2_hw): Likewise.
8236         (fix_<mode>si2_hw): Likewise.
8237         (fixuns_<mode>si2_hw): Likewise.
8238         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8239         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8240         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8241         fix<uns>_trunc<SFDF:mode>si2_p8.
8242         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8243         used.
8244         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8245         (fix<uns>_<mode>_mem): Likewise.
8246         (fctiw<u>z_<mode>_mem): Likewise.
8247         (fix<uns>_<mode>_mem): Likewise.
8248         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8249         the register allocator from doing a direct move to the GPRs to do
8250         a store, and instead use the ISA 3.0 store byte/half-word from
8251         vector register instruction.  For IEEE 128-bit floating point,
8252         also optimize stores of 32-bit ints.
8253         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8254
8255 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
8256
8257         * genextract.c (push_pathstr_operand): New function to support
8258         [a-zA-Z].
8259         (walk_rtx): Call push_pathstr_operand.
8260         (print_path): Support [a-zA-Z].
8261
8262 2018-02-07  Richard Biener  <rguenther@suse.de>
8263
8264         PR tree-optimization/84037
8265         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8266         (cse_and_gimplify_to_preheader): Declare.
8267         (vect_get_place_in_interleaving_chain): Likewise.
8268         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8269         ivexpr_map.
8270         (_loop_vec_info::~_loop_vec_info): Delete it.
8271         (cse_and_gimplify_to_preheader): New function.
8272         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8273         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8274         (vectorizable_load): Likewise.  For grouped stores always base
8275         the IV on the first element.
8276         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8277         condition before gimplifying.
8278
8279 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
8280
8281         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8282         *DIV_EXPR and *MOD_EXPR.
8283
8284 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
8285
8286         PR target/84248
8287         * config/i386/i386.c (ix86_option_override_internal): Mask out
8288         the CF_SET bit when checking -fcf-protection.
8289
8290 2018-02-07  Tom de Vries  <tom@codesourcery.com>
8291
8292         PR libgomp/84217
8293         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8294         enough.
8295
8296 2018-02-07  Richard Biener  <rguenther@suse.de>
8297
8298         PR tree-optimization/84204
8299         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8300         this place.
8301
8302         PR tree-optimization/84205
8303         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8304         special-case isl_ast_op_zdiv_r.
8305
8306         PR tree-optimization/84223
8307         * graphite-scop-detection.c (gather_bbs::before_dom_children):
8308         Only add conditions from within the region.
8309         (gather_bbs::after_dom_children): Adjust.
8310
8311 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
8312
8313         PR target/84209
8314         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8315         * config/avr/avr.md: Only post-reload split REG-REG moves if
8316         either register is GENERAL_REG_P.
8317
8318 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
8319
8320         PR tree-optimization/84235
8321         * tree-ssa-scopedtables.c
8322         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8323         if the subtraction is performed in floating point type where NaNs are
8324         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8325         build 1.  Formatting fix.
8326
8327 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
8328
8329         PR target/84146
8330         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8331         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8332         and skip it regardless of bb boundaries.  Use CALL_P macro,
8333         don't test INSN_P (insn) together with CALL_P or JUMP_P check
8334         unnecessarily, formatting fix.
8335
8336 2018-02-06  Michael Collison  <michael.collison@arm.com>
8337
8338         * config/arm/thumb2.md:
8339         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8340         (*thumb_mov_notscc): Ditto.
8341
8342 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
8343
8344         PR target/84154
8345         * config/rs6000/rs6000.md (su code attribute): Use "u" for
8346         unsigned_fix, not "s".
8347
8348 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
8349
8350         * configure.ac (gcc_fn_eh_frame_ro): New function.
8351         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8352         correct .eh_frame permissions.
8353         * configure: Regenerate.
8354
8355 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
8356
8357         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8358         irrelevant options.
8359
8360 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
8361
8362         * config/rs6000/rs6000.c (rs6000_option_override_internal):
8363         Display warning message for -mno-speculate-indirect-jumps.
8364
8365 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
8366
8367         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8368         Undocumented.
8369         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8370
8371 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
8372
8373         PR tree-optimization/84225
8374         * tree-eh.c (find_trapping_overflow): Only call
8375         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8376
8377 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
8378
8379         PR target/84145
8380         * config/i386/i386.c: Reimplement the check of possible options
8381         -mibt/-mshstk conbination. Change error messages.
8382         * doc/invoke.texi: Fix a typo: remove extra '='.
8383
8384 2018-02-06  Marek Polacek  <polacek@redhat.com>
8385
8386         PR tree-optimization/84228
8387         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8388
8389 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
8390
8391         PR target/82641
8392         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8393         emitted arch directives.
8394         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8395         __ARM_FEATURE_COPROC before changing architectures.
8396
8397 2018-02-06  Richard Biener  <rguenther@suse.de>
8398
8399         * config/i386/i386.c (print_reg): Fix typo.
8400         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8401
8402 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
8403
8404         * configure: Regenerate.
8405
8406 2018-02-05  Martin Sebor  <msebor@redhat.com>
8407
8408         PR tree-optimization/83369
8409         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8410         inlining context.
8411
8412 2018-02-05  Martin Liska  <mliska@suse.cz>
8413
8414         * doc/invoke.texi: Cherry-pick upstream r323995.
8415
8416 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
8417
8418         * ira.c (ira_init_register_move_cost): Adjust comment.
8419
8420 2018-02-05  Martin Liska  <mliska@suse.cz>
8421
8422         PR gcov-profile/84137
8423         * doc/gcov.texi: Fix typo in documentation.
8424
8425 2018-02-05  Martin Liska  <mliska@suse.cz>
8426
8427         PR gcov-profile/83879
8428         * doc/gcov.texi: Document necessity of --dynamic-list-data when
8429         using dlopen functionality.
8430
8431 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
8432
8433         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8434         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8435         _mm_maskz_range_ss, _mm_mask_range_round_ss,
8436         _mm_maskz_range_round_ss): New intrinsics.
8437         (__builtin_ia32_rangesd128_round)
8438         (__builtin_ia32_rangess128_round): Remove.
8439         (__builtin_ia32_rangesd128_mask_round,
8440         __builtin_ia32_rangess128_mask_round): New builtins.
8441         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8442         __builtin_ia32_rangess128_round): Remove.
8443         (__builtin_ia32_rangesd128_mask_round,
8444         __builtin_ia32_rangess128_mask_round): New builtins.
8445         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8446         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8447         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8448         "<round_saeonly_constraint>")): Changed to ...
8449         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8450         "<round_saeonly_scalar_constraint>")): ... this.
8451         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8452         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8453         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8454         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8455         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8456
8457 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
8458
8459         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8460         options.
8461         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8462         Remove all values except native, 8540 and 8548.
8463
8464 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
8465
8466         * config/i386/i386.c (ix86_output_function_return): Pass
8467         INVALID_REGNUM, instead of -1, as invalid register number to
8468         indirect_thunk_name and output_indirect_thunk.
8469
8470 2018-02-02  Julia Koval  <julia.koval@intel.com>
8471
8472         * config.gcc: Add -march=icelake.
8473         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8474         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8475         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8476         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8477         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8478         (processor_target_table): Add icelake.
8479         (ix86_option_override_internal): Handle new PTAs.
8480         (get_builtin_code_for_version): Handle icelake.
8481         (M_INTEL_COREI7_ICELAKE): New.
8482         (fold_builtin_cpu): Handle icelake.
8483         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8484         * doc/invoke.texi: Add -march=icelake.
8485
8486 2018-02-02  Julia Koval  <julia.koval@intel.com>
8487
8488         * config/i386/i386.c (ix86_option_override_internal): Change flags type
8489         to wide_int_bitmask.
8490         * wide-int-bitmask.h: New.
8491
8492 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
8493
8494         PR target/84066
8495         * config/i386/i386.md: Replace Pmode with word_mode in
8496         builtin_setjmp_setup and builtin_longjmp to support x32.
8497
8498 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
8499
8500         PR target/56010
8501         PR target/83743
8502         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8503         #include "opts.h".
8504         (rs6000_supported_cpu_names): New static variable.
8505         (linux_cpu_translation_table): Likewise.
8506         (elf_platform) <cpu>: Define new static variable and use it.
8507         Translate kernel AT_PLATFORM name to canonical name if needed.
8508         Error if platform name is unknown.
8509
8510 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
8511
8512         PR target/84089
8513         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8514
8515 2018-02-01  Jeff Law  <law@redhat.com>
8516
8517         PR target/84128
8518         * config/i386/i386.c (release_scratch_register_on_entry): Add new
8519         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
8520         the scratch if RELEASE_VIA_POP is false.
8521         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8522         If we have to save a temporary register, decrement SIZE appropriately.
8523         Pass new arguments to release_scratch_register_on_entry.
8524         (ix86_adjust_stack_and_probe): Likewise.
8525         (ix86_emit_probe_stack_range): Pass new arguments to
8526         release_scratch_register_on_entry.
8527
8528 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
8529
8530         PR rtl-optimization/84157
8531         * combine.c (change_zero_ext): Use REG_P predicate in
8532         front of HARD_REGISTER_P predicate.
8533
8534 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
8535
8536         * config/avr/avr.c (avr_option_override): Move disabling of
8537         -fdelete-null-pointer-checks to...
8538         * common/config/avr/avr-common.c (avr_option_optimization_table):
8539         ...here.
8540
8541 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
8542
8543         PR tree-optimization/81635
8544         * tree-data-ref.c (split_constant_offset_1): For types that
8545         wrap on overflow, try to use range info to prove that wrapping
8546         cannot occur.
8547
8548 2018-02-01  Renlin Li  <renlin.li@arm.com>
8549
8550         PR target/83370
8551         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8552         TAILCALL_ADDR_REGS.
8553         (aarch64_register_move_cost): Likewise.
8554         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8555         TAILCALL_ADDR_REGS.
8556         (REG_CLASS_NAMES): Likewise.
8557         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8558         TAILCALL_ADDR_REGS. Remove IP registers.
8559         * config/aarch64/aarch64.md (Ucs): Update register constraint.
8560
8561 2018-02-01  Richard Biener  <rguenther@suse.de>
8562
8563         * domwalk.h (dom_walker::dom_walker): Add additional constructor
8564         for specifying RPO order and allow NULL for that.
8565         * domwalk.c (dom_walker::dom_walker): Likewise.
8566         (dom_walker::walk): Handle NULL RPO order.
8567         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8568         in RPO order.
8569         (rewrite_update_dom_walker): Likewise.
8570         (mark_def_dom_walker): Likewise.
8571
8572 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
8573
8574         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8575         (aarch64_maybe_expand_sve_subreg_move): Declare.
8576         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8577         * config/aarch64/predicates.md (aarch64_any_register_operand): New
8578         predicate.
8579         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8580         that are semantically a reverse operation.
8581         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8582         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8583         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8584         functions.
8585         (aarch64_can_change_mode_class): For big-endian, forbid changes
8586         between two SVE modes if they have different element sizes.
8587
8588 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
8589
8590         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8591         the TImode handling for big-endian targets.
8592
8593 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
8594
8595         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8596         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
8597         not just bytes.
8598         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8599         Remove BSWAP handing for big-endian targets and use the form of
8600         LD1RQ appropariate for the mode.
8601
8602 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
8603
8604         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8605         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8606         duplicated element.
8607
8608 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
8609
8610         PR tearget/83845
8611         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8612         check for operands that need to go through aarch64_sve_reload_be.
8613
8614 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
8615
8616         PR tree-optimization/81661
8617         PR tree-optimization/84117
8618         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8619         * tree-eh.c: Include gimplify.h.
8620         (find_trapping_overflow, replace_trapping_overflow,
8621         rewrite_to_non_trapping_overflow): New functions.
8622         * tree-vect-loop.c: Include tree-eh.h.
8623         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8624         * tree-data-ref.c: Include tree-eh.h.
8625         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8626
8627 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
8628
8629         PR rtl-optimization/84123
8630         * combine.c (change_zero_ext): Check if hard register satisfies
8631         can_change_dest_mode before calling gen_lowpart_SUBREG.
8632
8633 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
8634
8635         PR target/82444
8636         * ira.c (ira_init_register_move_cost): Remove assert.
8637
8638 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
8639
8640         PR rtl-optimization/84071
8641         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8642         * doc/tm.texi: Regenerate.
8643
8644 2018-01-31  Richard Biener  <rguenther@suse.de>
8645
8646         PR tree-optimization/84132
8647         * tree-data-ref.c (analyze_miv_subscript): Properly
8648         check whether evolution_function_is_affine_multivariate_p
8649         before calling gcd_of_steps_may_divide_p.
8650
8651 2018-01-31  Julia Koval  <julia.koval@intel.com>
8652
8653         PR target/83618
8654         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8655         * config/i386/i386.md (rdpid_rex64) New.
8656         (rdpid): Make 32bit only.
8657
8658 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
8659
8660         PR lto/84105
8661         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8662         an IDENTIFIER_NODE for FUNCTION_TYPE's.
8663
8664 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
8665
8666         Revert
8667         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
8668
8669         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8670
8671 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
8672
8673         PR rtl-optimization/84071
8674         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8675         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8676
8677 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
8678
8679         * config/arc/arc.c (arc_handle_aux_attribute): New function.
8680         (arc_attribute_table): Add 'aux' attribute.
8681         (arc_in_small_data_p): Consider aux like variables.
8682         (arc_is_aux_reg_p): New function.
8683         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8684         (arc_get_aux_arg): New function.
8685         (prepare_move_operands): Handle aux-register access.
8686         (arc_handle_aux_attribute): New function.
8687         * doc/extend.texi (ARC Variable attributes): Add subsection.
8688
8689 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
8690
8691         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8692         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8693         (arc_attribute_table): Add 'uncached' attribute.
8694         (arc_print_operand): Print '.di' flag for uncached memory
8695         accesses.
8696         (arc_in_small_data_p): Do not consider for small data the uncached
8697         types.
8698         (arc_is_uncached_mem_p): New function.
8699         * config/arc/predicates.md (compact_store_memory_operand): Check
8700         for uncached memory accesses.
8701         (nonvol_nonimm_operand): Likewise.
8702         * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8703
8704 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
8705
8706         PR c/84100
8707         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8708         falign-loops=): Add Optimization flag.
8709
8710 2018-01-30  Jeff Law  <law@redhat.com>
8711
8712         PR target/84064
8713         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8714         INT_REGISTERS_SAVED.  Check it prior to calling
8715         get_scratch_register_on_entry.
8716         (ix86_adjust_stack_and_probe): Similarly.
8717         (ix86_emit_probe_stack_range): Similarly.
8718         (ix86_expand_prologue): Corresponding changes.
8719
8720 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
8721
8722         PR target/40411
8723         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8724         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8725
8726 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
8727
8728         PR target/84112
8729         * lra-constraints.c (curr_insn_transform): Process AND in the
8730         address.
8731
8732 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
8733
8734         PR rtl-optimization/83986
8735         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8736         dependence against last_pending_memory_flush in addition to
8737         pending_jump_insns.
8738
8739 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
8740
8741         PR tree-optimization/81611
8742         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8743         copies.
8744
8745 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
8746
8747         PR target/83758
8748         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8749         a reg rtx.
8750
8751 2018-01-30  Richard Biener  <rguenther@suse.de>
8752             Jakub Jelinek  <jakub@redhat.com>
8753
8754         PR tree-optimization/84111
8755         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8756         inner loops added during recursion, as they don't have up-to-date
8757         SSA form.
8758
8759 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
8760
8761         PR ipa/81360
8762         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8763         (can_inline_edge_by_limits_p): ... here.
8764         (can_early_inline_edge_p, check_callers,
8765         update_caller_keys, update_callee_keys, recursive_inlining,
8766         add_new_edges_to_heap, speculation_useful_p,
8767         inline_small_functions,
8768         inline_small_functions, flatten_function,
8769         inline_to_all_callers_1): Update.
8770
8771 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
8772
8773         * profile-count.c (profile_count::combine_with_ipa_count): Handle
8774         zeros correctly.
8775
8776 2018-01-30  Richard Biener  <rguenther@suse.de>
8777
8778         PR tree-optimization/83008
8779         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8780         invariant and constant vector uses in stmts when they need
8781         more than one stmt.
8782
8783 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
8784
8785         PR bootstrap/84017
8786         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8787         * configure: Regenerate.
8788
8789 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
8790
8791         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8792         pattern.
8793         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8794         Use gen_rtx_REG rather than gen_lowpart.
8795
8796 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
8797
8798         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8799         rather than 0 when creating partial subregs.
8800
8801 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
8802
8803         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8804         of usage.
8805
8806 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
8807
8808         PR target/81550
8809         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8810         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8811         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8812         flags.  This restores the settings used before the 2017-07-24.
8813         Turning off pre increment/decrement/modify allows IVOPTS to
8814         optimize DF/SF loops where the index is an int.
8815
8816 2018-01-29  Richard Biener  <rguenther@suse.de>
8817             Kelvin Nilsen  <kelvin@gcc.gnu.org>
8818
8819         PR bootstrap/80867
8820         * tree-vect-stmts.c (vectorizable_call): Don't call
8821         targetm.vectorize_builtin_md_vectorized_function if callee is
8822         NULL.
8823
8824 2018-01-22  Carl Love  <cel@us.ibm.com>
8825
8826         * doc/extend.tex: Fix typo in second arg in
8827         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8828
8829 2018-01-29  Richard Biener  <rguenther@suse.de>
8830
8831         PR tree-optimization/84086
8832         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8833         (flush_ssaname_freelist): When SSA names were released reset
8834         the SCEV hash table.
8835
8836 2018-01-29  Richard Biener  <rguenther@suse.de>
8837
8838         PR tree-optimization/84057
8839         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8840         removed paths when removing edges.
8841
8842 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
8843
8844         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8845         -mfunction-return=@var{choice}.
8846
8847 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8848
8849         PR diagnostic/84034
8850         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8851         Handle CR like TAB.
8852         (layout::print_source_line): Likewise.
8853         (test_get_line_width_without_trailing_whitespace): Add test cases.
8854
8855 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
8856
8857         PR middle-end/84040
8858         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8859         debug insns.
8860
8861 2018-01-26  Jim Wilson  <jimw@sifive.com>
8862
8863         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8864
8865         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8866         specified.
8867
8868 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8869
8870         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8871         and CMP + SUB-immediate -> SUBS.
8872
8873 2018-01-26  Martin Sebor  <msebor@redhat.com>
8874
8875         PR tree-optimization/83896
8876         * tree-ssa-strlen.c (get_string_len): Rename...
8877         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
8878         Avoid assuming length is constant.
8879         (handle_char_store): Use HOST_WIDE_INT for string length.
8880
8881 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
8882
8883         PR target/81763
8884         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8885         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8886
8887 2018-01-26  Richard Biener  <rguenther@suse.de>
8888
8889         PR rtl-optimization/84003
8890         * dse.c (record_store): Only record redundant stores when
8891         the earlier store aliases at least all accesses the later one does.
8892
8893 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
8894
8895         PR rtl-optimization/83985
8896         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8897         REG_CFA_RESTORE insns.
8898         (delete_unmarked_insns): Don't ignore separate shrink wrapping
8899         REG_CFA_RESTORE insns here.
8900
8901         PR c/83989
8902         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8903         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8904
8905 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
8906
8907         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8908         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8909         (arc_init): Likewise.
8910         (arc_override_options): Likewise.
8911         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8912         value.
8913         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8914         support.
8915         * config/arc/arc.h (TARGET_DBNZ): Define.
8916         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8917         properly set the tune attribute.
8918         (dbnz): Use TARGET_DBNZ guard.
8919         * config/arc/arc.opt (mtune): Add core3 option.
8920
8921 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
8922
8923         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8924         recognize new pic like addresses.
8925         (arc_delegitimize_address): Clean up.
8926
8927 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
8928
8929         * config/arc/arc-arches.def: Option mrf16 valid for all
8930         architectures.
8931         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8932         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8933         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8934         * config/arc/arc-tables.opt: Regenerate.
8935         * config/arc/arc.c (arc_conditional_register_usage): Handle
8936         reduced register file case.
8937         (arc_file_start): Set must have build attributes.
8938         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8939         mrf16 option value.
8940         * config/arc/arc.opt (mrf16): Add new option.
8941         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8942         * config/arc/genmultilib.awk: Handle new mrf16 option.
8943         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8944         * config/arc/t-multilib: Regenerate.
8945         * doc/invoke.texi (ARC Options): Document mrf16 option.
8946
8947 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
8948
8949         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8950         * config/arc/arc.c (arc_handle_secure_attribute): New function.
8951         (arc_attribute_table): Add 'secure_call' attribute.
8952         (arc_print_operand): Print secure call operand.
8953         (arc_function_ok_for_sibcall): Don't optimize tail calls when
8954         secure.
8955         (arc_is_secure_call_p): New function.  * config/arc/arc.md
8956         (call_i): Add support for sjli instruction.
8957         (call_value_i): Likewise.
8958         * config/arc/constraints.md (Csc): New constraint.
8959
8960 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
8961             John Eric Martin  <John.Martin@emmicro-us.com>
8962
8963         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8964         * config/arc/arc.c (_arc_jli_section): New struct.
8965         (arc_jli_section): New type.
8966         (rc_jli_sections): New static variable.
8967         (arc_handle_jli_attribute): New function.
8968         (arc_attribute_table): Add jli_always and jli_fixed attribute.
8969         (arc_file_end): New function.
8970         (TARGET_ASM_FILE_END): Define.
8971         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8972         (arc_add_jli_section): New function.
8973         (jli_call_scan): Likewise.
8974         (arc_reorg): Call jli_call_scan.
8975         (arc_output_addsi): Remove 'S' from printing asm operand.
8976         (arc_is_jli_call_p): New function.
8977         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8978         operand.
8979         (movhi_insn): Likewise.
8980         (movsi_insn): Likewise.
8981         (movsi_set_cc_insn): Likewise.
8982         (loadqi_update): Likewise.
8983         (load_zeroextendqisi_update): Likewise.
8984         (load_signextendqisi_update): Likewise.
8985         (loadhi_update): Likewise.
8986         (load_zeroextendhisi_update): Likewise.
8987         (load_signextendhisi_update): Likewise.
8988         (loadsi_update): Likewise.
8989         (loadsf_update): Likewise.
8990         (movsicc_insn): Likewise.
8991         (bset_insn): Likewise.
8992         (bxor_insn): Likewise.
8993         (bclr_insn): Likewise.
8994         (bmsk_insn): Likewise.
8995         (bicsi3_insn): Likewise.
8996         (cmpsi_cc_c_insn): Likewise.
8997         (movsi_ne): Likewise.
8998         (movsi_cond_exec): Likewise.
8999         (clrsbsi2): Likewise.
9000         (norm_f): Likewise.
9001         (normw): Likewise.
9002         (swap): Likewise.
9003         (divaw): Likewise.
9004         (flag): Likewise.
9005         (sr): Likewise.
9006         (kflag): Likewise.
9007         (ffs): Likewise.
9008         (ffs_f): Likewise.
9009         (fls): Likewise.
9010         (call_i): Remove 'S' asm letter, add jli instruction.
9011         (call_value_i): Likewise.
9012         * config/arc/arc.op (mjli-always): New option.
9013         * config/arc/constraints.md (Cji): New constraint.
9014         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9015         operand.
9016         (subsf3_fpx): Likewise.
9017         (mulsf3_fpx): Likewise.
9018         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9019         asm operand.
9020         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9021         function attrbutes.
9022         * doc/invoke.texi (ARC): Document mjli-always option.
9023
9024 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
9025
9026         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9027         avoid addition with 0 and use incw and decw where possible.
9028
9029 2018-01-26  Richard Biener  <rguenther@suse.de>
9030
9031         PR tree-optimization/81082
9032         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9033         association if it requires casting to unsigned.
9034         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9035         from fold_plusminus_mult_expr to catch important cases late when
9036         range info is available.
9037
9038 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9039
9040         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9041         * configure.ac (hidden_linkonce): New test.
9042         * configure: Regenerate.
9043         * config.in: Regenerate.
9044
9045 2018-01-26  Julia Koval  <julia.koval@intel.com>
9046
9047         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9048         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9049         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9050         _mm_mask_bitshuffle_epi64_mask): Fix type.
9051         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9052         USI_FTYPE_V4DI_V4DI_USI): Remove.
9053         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9054         __builtin_ia32_vpshufbitqmb256_mask,
9055         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9056         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9057         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9058
9059 2018-01-26  Alan Modra  <amodra@gmail.com>
9060
9061         PR target/84033
9062         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9063         UNSPEC_VBPERMQ.  Sort other unspecs.
9064
9065 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
9066
9067         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9068
9069 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
9070
9071         PR middle-end/83055
9072         * predict.c (drop_profile): Do not push/pop cfun; update also
9073         node->count.
9074         (handle_missing_profiles): Fix logic looking for zero profiles.
9075
9076 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
9077
9078         PR middle-end/83977
9079         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9080         on functions with #pragma omp declare simd or functions with simd
9081         attribute.
9082         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9083         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9084         Remove trailing \n from warning_at calls.
9085
9086 2018-01-25  Tom de Vries  <tom@codesourcery.com>
9087
9088         PR target/84028
9089         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9090         for neutered workers.
9091
9092 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
9093
9094         PR target/68467
9095         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9096         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9097
9098 2018-01-24  Jeff Law  <law@redhat.com>
9099
9100         PR target/83994
9101         * i386.c (get_probe_interval): Move to earlier point.
9102         (ix86_compute_frame_layout): If -fstack-clash-protection and
9103         the frame is larger than the probe interval, then use pushes
9104         to save registers rather than reg->mem moves.
9105         (ix86_expand_prologue): Remove conditional for int_registers_saved
9106         assertion.
9107
9108 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
9109
9110         PR target/84014
9111         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9112         min/max for never referenced object.
9113
9114 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
9115
9116         PR middle-end/83977
9117         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9118         here.
9119         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9120         attributes from DECL_ATTRIBUTES (decl) without affecting
9121         DECL_ATTRIBUTES (current_function_decl).
9122         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9123         functions with non-NULL DECL_ABSTRACT_ORIGIN.
9124
9125 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
9126
9127         PR tree-optimization/83979
9128         * fold-const.c (fold_comparison): Use constant_boolean_node
9129         instead of boolean_{true,false}_node.
9130
9131 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
9132
9133         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9134         with zero counts.
9135
9136 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9137
9138         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9139         Simplify the clause that sets the length attribute.
9140         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9141         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9142         clause that sets the length attribute.
9143         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9144
9145 2018-01-24  Tom de Vries  <tom@codesourcery.com>
9146
9147         PR target/83589
9148         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9149         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9150         Add strict parameter.
9151         (prevent_branch_around_nothing): Insert dummy insn between branch to
9152         label and label with no ptx insn inbetween.
9153         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9154
9155 2018-01-24  Tom de Vries  <tom@codesourcery.com>
9156
9157         PR target/81352
9158         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9159         for neutered threads in warp.
9160         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9161
9162 2018-01-24  Richard Biener  <rguenther@suse.de>
9163
9164         PR tree-optimization/83176
9165         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9166         operands.
9167
9168 2018-01-24  Richard Biener  <rguenther@suse.de>
9169
9170         PR tree-optimization/82819
9171         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9172         code generating pluses that are no-ops in the target precision.
9173
9174 2018-01-24  Richard Biener  <rguenther@suse.de>
9175
9176         PR middle-end/84000
9177         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9178
9179 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
9180
9181         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9182         to merge probabilities.
9183         * predict.c (probably_never_executed): Also mark as cold functions
9184         with global 0 profile and guessed local profile.
9185         * profile-count.c (profile_probability::combine_with_count): New
9186         member function.
9187         * profile-count.h (profile_probability::operator*,
9188         profile_probability::operator*=, profile_probability::operator/,
9189         profile_probability::operator/=): Reduce precision to adjusted
9190         and set value to guessed on contradictory divisions.
9191         (profile_probability::combine_with_freq): Remove.
9192         (profile_probability::combine_wiht_count): Declare.
9193         (profile_count::force_nonzero):: Set to adjusted.
9194         (profile_count::probability_in):: Set quality to adjusted.
9195         * tree-ssa-tail-merge.c (replace_block_by): Use
9196         combine_with_count.
9197
9198 2018-01-23  Andrew Waterman  <andrew@sifive.com>
9199             Jim Wilson  <jimw@sifive.com>
9200
9201         * config/riscv/riscv.c (riscv_stack_boundary): New.
9202         (riscv_option_override): Set riscv_stack_boundary.  Handle
9203         riscv_preferred_stack_boundary_arg.
9204         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9205         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9206         (STACK_BOUNDARY): Set to riscv_stack_boundary.
9207         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9208         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9209         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9210
9211 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
9212
9213         PR target/83905
9214         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9215         of struct ix86_frame.
9216         (ix86_expand_epilogue): Likewise.  Add a local variable for
9217         the reg_save_offset field in struct ix86_frame.
9218
9219 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
9220
9221         PR tree-optimization/82604
9222         * tree-loop-distribution.c (enum partition_kind): New enum item
9223         PKIND_PARTIAL_MEMSET.
9224         (partition_builtin_p): Support above new enum item.
9225         (generate_code_for_partition): Ditto.
9226         (compute_access_range): Differentiate cases that equality can be
9227         proven at all loops, the innermost loops or no loops.
9228         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9229         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
9230         (finalize_partitions, distribute_loop): Don't fuse partition of
9231         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9232         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9233         parloop is enabled.
9234
9235 2018-01-23  Martin Liska  <mliska@suse.cz>
9236
9237         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9238         order to ignore the predictor.
9239         (PRED_POLYMORPHIC_CALL): Likewise.
9240         (PRED_RECURSIVE_CALL): Likewise.
9241
9242 2018-01-23  Martin Liska  <mliska@suse.cz>
9243
9244         * tree-profile.c (tree_profiling): Print function header to
9245         aware reader which function we are working on.
9246         * value-prof.c (gimple_find_values_to_profile): Do not print
9247         not interesting value histograms.
9248
9249 2018-01-23  Martin Liska  <mliska@suse.cz>
9250
9251         * profile-count.h (enum profile_quality): Add
9252         profile_uninitialized as the first value. Do not number values
9253         as they are zero based.
9254         (profile_count::verify): Update sanity check.
9255         (profile_probability::verify): Likewise.
9256
9257 2018-01-23  Nathan Sidwell  <nathan@acm.org>
9258
9259         * doc/invoke.texi (ffor-scope): Deprecate.
9260
9261 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
9262
9263         PR tree-optimization/83510
9264         * domwalk.c (set_all_edges_as_executable): New function.
9265         (dom_walker::dom_walker): Convert bool param
9266         "skip_unreachable_blocks" to enum reachability.  Move setup of
9267         edge flags to set_all_edges_as_executable and only do it when
9268         reachability is REACHABLE_BLOCKS.
9269         * domwalk.h (enum dom_walker::reachability): New enum.
9270         (dom_walker::dom_walker): Convert bool param
9271         "skip_unreachable_blocks" to enum reachability.
9272         (set_all_edges_as_executable): New decl.
9273         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
9274         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9275         "reachability".
9276         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9277         but converting true to REACHABLE_BLOCKS.
9278         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9279         * tree-vrp.c
9280         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9281         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9282         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9283         REACHABLE_BLOCKS.
9284         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9285         if check_all_array_refs will be called.
9286
9287 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
9288
9289         * tree.c (selftest::test_location_wrappers): Add more test
9290         coverage.
9291
9292 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
9293
9294         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9295         (selftest::test_bit_in_range): Likewise.
9296
9297 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
9298
9299         PR testsuite/83888
9300         * doc/sourcebuild.texi (vect_float): Say that the selector
9301         only describes the situation when -funsafe-math-optimizations is on.
9302         (vect_float_strict): Document.
9303
9304 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
9305
9306         PR tree-optimization/83965
9307         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9308         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9309         instead of checking only for a reduction.
9310         (vect_recog_widen_sum_pattern): Likewise.
9311
9312 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
9313
9314         * predict.c (probably_never_executed): Only use precise profile info.
9315         (compute_function_frequency): Skip after inlining hack since we now
9316         have quality checking.
9317
9318 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
9319
9320         * profile-count.h (profile_probability::very_unlikely,
9321         profile_probability::unlikely, profile_probability::even): Set
9322         precision to guessed.
9323
9324 2018-01-23  Richard Biener  <rguenther@suse.de>
9325
9326         PR tree-optimization/83963
9327         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9328         Properly terminate dominator walk when crossing the exit edge not
9329         when visiting its source block.
9330
9331 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
9332
9333         PR c++/83918
9334         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9335         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9336
9337 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
9338
9339         PR tree-optimization/83957
9340         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
9341         semicolon after for body surrounded by braces.
9342
9343         PR tree-optimization/83081
9344         * profile-count.h (profile_probability::split): New method.
9345         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9346         Use profile_probability::split.
9347         (do_compare_rtx_and_jump): Fix adjustment of probabilities
9348         when splitting a single conditional jump into 2.
9349
9350 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
9351
9352         PR tree-optimization/69452
9353         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9354         decl.
9355
9356 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
9357
9358         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9359         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9360         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9361
9362 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
9363
9364         * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9365         * config/rl78/rl78.md (movdi): New define_expand.
9366         * config/rl78/rl78.c (rl78_split_movdi): New function.
9367
9368 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
9369
9370         PR target/83862
9371         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9372         no longer used.
9373         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9374         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9375         128-bit to produce an UNSPEC move to get the double word with the
9376         signbit and then a shift directly to do signbit.
9377         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9378         implementation with a new version that just does either a direct
9379         move or a regular move.  Move memory interface to separate insns.
9380         Move insns so they are next to the expander.
9381         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9382         with signbit move.  Split big and little endian case.
9383         (signbit<mode>2_dm_mem_le): Likewise.
9384         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9385         (signbit<mode>2_dm2): Likewise.
9386
9387 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
9388
9389         * config/rl78/rl78.md (anddi3): New define_expand.
9390
9391 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
9392
9393         * config/rl78/rl78.md (umindi3): New define_expand.
9394
9395 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
9396
9397         * config/rl78/rl78.md (smindi3): New define_expand.
9398
9399 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
9400
9401         * config/rl78/rl78.md (smaxdi3): New define_expand.
9402
9403 2018-01-22  Carl Love  <cel@us.ibm.com>
9404
9405         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9406         LVX_V1TI): Add macro expansion.
9407         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9408         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9409         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9410         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9411         Change check to determine if the instruction is a byte reversing
9412         entry.  Fix typo in comment.
9413         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9414         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9415         Add def_builtin calls for new builtins.
9416         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9417         Add define_insn expansion.
9418
9419 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
9420
9421         * config/rl78/rl78.md (umaxdi3): New define_expand.
9422
9423 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
9424
9425         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9426         for non-QImode registers.
9427
9428 2018-01-22  Richard Biener  <rguenther@suse.de>
9429
9430         PR tree-optimization/83963
9431         * graphite-scop-detection.c (scop_detection::get_sese): Delay
9432         including the loop exit block.
9433         (scop_detection::merge_sese): Likewise.
9434         (scop_detection::add_scop): Do it here instead.
9435
9436 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9437
9438         * doc/sourcebuild.texi (arm_softfloat): Document.
9439
9440 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
9441
9442         PR gcc/77734
9443         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9444         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9445         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9446
9447 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9448             David Edelsohn  <dje.gcc@gmail.com>
9449
9450         PR target/83946
9451         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9452         Change "crset eq" to "crset 2".
9453         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9454         (*call_indirect_aix<mode>_nospec): Likewise.
9455         (*call_value_indirect_aix<mode>_nospec): Likewise.
9456         (*call_indirect_elfv2<mode>_nospec): Likewise.
9457         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9458         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9459         change assembly output from . to $.
9460         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9461         (indirect_jump<mode>_nospec): Change assembly output from . to $.
9462         (*tablejump<mode>_internal1_nospec): Likewise.
9463
9464 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
9465
9466         PR target/80870
9467         * config/sh/sh_optimize_sett_clrt.cc:
9468         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9469
9470 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
9471
9472         PR tree-optimization/83940
9473         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9474         offset_dt to vect_constant_def rather than vect_unknown_def_type.
9475         (vect_check_load_store_mask): Add a mask_dt_out parameter and
9476         use it to pass back the definition type.
9477         (vect_check_store_rhs): Likewise rhs_dt_out.
9478         (vect_build_gather_load_calls): Add a mask_dt argument and use
9479         it instead of a call to vect_is_simple_use.
9480         (vectorizable_store): Update calls to vect_check_load_store_mask
9481         and vect_check_store_rhs.  Use the dt returned by the latter instead
9482         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
9483         instead of calls to vect_is_simple_use.  Pass the scalar rather
9484         than the vector operand to vect_is_simple_use when handling
9485         second and subsequent copies of an rhs value.
9486         (vectorizable_load): Update calls to vect_check_load_store_mask
9487         and vect_build_gather_load_calls.  Use the cached mask_dt and
9488         gs_info.offset_dt instead of calls to vect_is_simple_use.
9489
9490 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
9491
9492         PR middle-end/83945
9493         * tree-emutls.c: Include gimplify.h.
9494         (lower_emutls_2): New function.
9495         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9496         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9497         it before further processing.
9498
9499         PR target/83930
9500         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9501         UINTVAL (trueop1) instead of INTVAL (op1).
9502
9503 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
9504
9505         PR debug/81570
9506         PR debug/83728
9507         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9508         INCOMING_FRAME_SP_OFFSET if not defined.
9509         (scan_trace): Add ENTRY argument.  If true and
9510         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9511         emit a note to adjust the CFA offset.
9512         (create_cfi_notes): Adjust scan_trace callers.
9513         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9514         INCOMING_FRAME_SP_OFFSET in the CIE.
9515         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9516         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9517         Likewise.
9518         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9519         * doc/tm.texi: Regenerated.
9520
9521 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
9522
9523         PR rtl-optimization/83147
9524         * lra-constraints.c (remove_inheritance_pseudos): Use
9525         lra_substitute_pseudo_within_insn.
9526
9527 2018-01-19  Tom de Vries  <tom@codesourcery.com>
9528             Cesar Philippidis  <cesar@codesourcery.com>
9529
9530         PR target/83920
9531         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9532
9533 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
9534
9535         PR target/83790
9536         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9537         spaces for function labels.
9538
9539 2018-01-19  Martin Liska  <mliska@suse.cz>
9540
9541         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9542         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9543         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9544         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9545         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9546         (PRED_CONST_RETURN): Change from 69 to 65.
9547         (PRED_NULL_RETURN): Change from 91 to 71.
9548         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9549         (PRED_LOOP_GUARD): Change from 66 to 73.
9550
9551 2018-01-19  Martin Liska  <mliska@suse.cz>
9552
9553         * predict.c (predict_insn_def): Add new assert.
9554         (struct branch_predictor): Change type to signed integer.
9555         (test_prediction_value_range): Amend test to cover
9556         PROB_UNINITIALIZED.
9557         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9558         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9559         (PRED_LOOP_ITERATIONS_MAX): Likewise.
9560         (PRED_LOOP_IV_COMPARE): Likewise.
9561         * predict.h (PROB_UNINITIALIZED): Define new constant.
9562
9563 2018-01-19  Martin Liska  <mliska@suse.cz>
9564
9565         * predict.c (dump_prediction): Add new format for
9566         analyze_brprob.py script which is enabled with -details
9567         suboption.
9568         * profile-count.h (precise_p): New function.
9569
9570 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
9571
9572         PR tree-optimization/83922
9573         * tree-vect-loop.c (vect_verify_full_masking): Return false if
9574         there are no statements that need masking.
9575         (vect_active_double_reduction_p): New function.
9576         (vect_analyze_loop_operations): Use it when handling phis that
9577         are not in the loop header.
9578
9579 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
9580
9581         PR tree-optimization/83914
9582         * tree-vect-loop.c (vectorizable_induction): Don't convert
9583         init_expr or apply the peeling adjustment for inductions
9584         that are nested within the vectorized loop.
9585
9586 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9587
9588         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9589         instead of NEG.
9590
9591 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
9592
9593         PR sanitizer/81715
9594         PR testsuite/83882
9595         * function.h (gimplify_parameters): Add gimple_seq * argument.
9596         * function.c: Include gimple.h and options.h.
9597         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9598         for the added local temporaries if needed.
9599         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9600         if there are any parameter cleanups, wrap whole body into a
9601         try/finally with the cleanups.
9602
9603 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
9604
9605         PR target/82964
9606         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9607         Use GET_MODE_CLASS for scalar floating point.
9608
9609 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
9610
9611         PR ipa/82256
9612         patch by PaX Team
9613         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9614         Fix call of call_cgraph_insertion_hooks.
9615
9616 2018-01-18  Martin Sebor  <msebor@redhat.com>
9617
9618         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9619
9620 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
9621
9622         PR ipa/83619
9623         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9624         frequencies.
9625
9626 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
9627
9628         PR other/70268
9629         * common.opt: (-ffile-prefix-map): New option.
9630         * opts.c (common_handle_option): Defer it.
9631         * opts-global.c (handle_common_deferred_options): Handle it.
9632         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9633         * file-prefix-map.h: New file.
9634         (remap_debug_filename, add_debug_prefix_map): ...here.
9635         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9636         * final.c (debug_prefix_map, add_debug_prefix_map
9637         remap_debug_filename): Move to...
9638         * file-prefix-map.c: New file.
9639         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9640         generalize, get rid of alloca(), use strrchr() instead of strchr().
9641         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9642         Implement in terms of add_prefix_map().
9643         (remap_macro_filename, remap_debug_filename): Implement in term of
9644         remap_filename().
9645         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9646         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9647         * dbxout.c: Include file-prefix-map.h.
9648         * varasm.c: Likewise.
9649         * vmsdbgout.c: Likewise.
9650         * xcoffout.c: Likewise.
9651         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9652         * doc/cppopts.texi (-fmacro-prefix-map): Document.
9653         * doc/invoke.texi (-ffile-prefix-map): Document.
9654         (-fdebug-prefix-map): Update description.
9655
9656 2018-01-18  Martin Liska  <mliska@suse.cz>
9657
9658         * config/i386/i386.c (indirect_thunk_name): Document that also
9659         lfence is emitted.
9660         (output_indirect_thunk): Document why both instructions
9661         (pause and lfence) are generated.
9662
9663 2018-01-18  Richard Biener  <rguenther@suse.de>
9664
9665         PR tree-optimization/83887
9666         * graphite-scop-detection.c
9667         (scop_detection::get_nearest_dom_with_single_entry): Remove.
9668         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9669         (scop_detection::merge_sese): Re-implement with a flood-fill
9670         algorithm that properly finds a SESE region if it exists.
9671
9672 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
9673
9674         PR c/61240
9675         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9676         pointer_diff optimizations use view_convert instead of convert.
9677
9678 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9679
9680         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9681         Generate different code for -mno-speculate-indirect-jumps.
9682         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9683         (*call_indirect_aix<mode>): Disable for
9684         -mno-speculate-indirect-jumps.
9685         (*call_indirect_aix<mode>_nospec): New define_insn.
9686         (*call_value_indirect_aix<mode>): Disable for
9687         -mno-speculate-indirect-jumps.
9688         (*call_value_indirect_aix<mode>_nospec): New define_insn.
9689         (*sibcall_nonlocal_sysv<mode>): Generate different code for
9690         -mno-speculate-indirect-jumps.
9691         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9692
9693 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
9694
9695         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9696         long double type, set the flags for noting the default long double
9697         type, even if we don't pass or return a long double type.
9698
9699 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
9700
9701         PR ipa/83051
9702         * ipa-inline.c (flatten_function): Do not overwrite final inlining
9703         failure.
9704
9705 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
9706
9707         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9708         support for merge[hl].
9709         (fold_mergehl_helper): New helper function.
9710         (tree-vector-builder.h): New #include for tree_vector_builder usage.
9711         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9712         (altivec_vmrglw_direct): Add xxmrglw insn.
9713
9714 2018-01-17  Andrew Waterman  <andrew@sifive.com>
9715
9716         * config/riscv/riscv.c (riscv_conditional_register_usage): If
9717         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9718
9719 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
9720
9721         PR lto/83121
9722         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9723         call the lto_location_cache before reading the
9724         DECL_SOURCE_LOCATION of the types.
9725
9726 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
9727             Richard Sandiford  <richard.sandiford@linaro.org>
9728
9729         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9730         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9731         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9732         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9733         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9734         Add declaration.
9735         * config/aarch64/constraints.md (aarch64_movti_operand):
9736         Limit immediates.
9737         * config/aarch64/predicates.md (Uti): Add new constraint.
9738
9739 2018-01-17 Carl Love  <cel@us.ibm.com>
9740
9741         * config/rs6000/vsx.md (define_expand xl_len_r,
9742         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9743         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9744         lxvll.
9745         (define_expand, define_insn): Move the shift left from  the
9746         define_insn to the define_expand for lxvl and stxvl instructions.
9747         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9748         and XL_LEN_R definitions to PURE.
9749
9750 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
9751
9752         * config/i386/i386.c (indirect_thunk_name): Declare regno
9753         as unsigned int.  Compare regno with INVALID_REGNUM.
9754         (output_indirect_thunk): Ditto.
9755         (output_indirect_thunk_function): Ditto.
9756         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
9757         in the call to output_indirect_thunk_function.
9758
9759 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
9760
9761         PR middle-end/83884
9762         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9763         rather than the size of inner_type to determine the stack slot size
9764         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9765
9766 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
9767
9768         PR target/83546
9769         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9770         to PTA_SILVERMONT.
9771
9772 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
9773
9774         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9775         endian Linux systems to optionally enable multilibs for selecting
9776         the long double type if the user configured an explicit type.
9777         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9778         have no long double multilibs if not defined.
9779         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9780         warn if the user used -mabi={ieee,ibm}longdouble and we built
9781         multilibs for long double.
9782         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9783         appropriate multilib option.
9784         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9785         multilib options.
9786         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9787         for building long double multilibs.
9788         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9789
9790 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
9791
9792         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9793         copies.
9794
9795         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9796         64 bits.
9797         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9798         128 bits.
9799
9800         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9801         variables.
9802
9803         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9804         return value.
9805
9806 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
9807
9808         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9809         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9810
9811 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9812
9813         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9814         different rtl trees depending on TARGET_64BIT.
9815         (rs6000_gen_lvx): Likewise.
9816
9817 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
9818
9819         * config/visium/visium.md (nop): Tweak comment.
9820         (hazard_nop): Likewise.
9821
9822 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9823
9824         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9825         -mspeculate-indirect-jumps.
9826         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9827         for -mno-speculate-indirect-jumps.
9828         (*call_indirect_elfv2<mode>_nospec): New define_insn.
9829         (*call_value_indirect_elfv2<mode>): Disable for
9830         -mno-speculate-indirect-jumps.
9831         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9832         (indirect_jump): Emit different RTL for
9833         -mno-speculate-indirect-jumps.
9834         (*indirect_jump<mode>): Disable for
9835         -mno-speculate-indirect-jumps.
9836         (*indirect_jump<mode>_nospec): New define_insn.
9837         (tablejump): Emit different RTL for
9838         -mno-speculate-indirect-jumps.
9839         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9840         (tablejumpsi_nospec): New define_expand.
9841         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9842         (tablejumpdi_nospec): New define_expand.
9843         (*tablejump<mode>_internal1): Disable for
9844         -mno-speculate-indirect-jumps.
9845         (*tablejump<mode>_internal1_nospec): New define_insn.
9846         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9847         option.
9848
9849 2018-01-16  Artyom Skrobov tyomitch@gmail.com
9850
9851         * caller-save.c (insert_save): Drop unnecessary parameter.  All
9852         callers updated.
9853
9854 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
9855             Richard Biener  <rguenth@suse.de>
9856
9857         PR libgomp/83590
9858         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9859         return early, inline manually is_gimple_sizepos.  Make sure if we
9860         call gimplify_expr we don't end up with a gimple constant.
9861         * tree.c (variably_modified_type_p): Don't return true for
9862         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
9863         * gimplify.h (is_gimple_sizepos): Remove.
9864
9865 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
9866
9867         PR tree-optimization/83857
9868         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9869         vectorizable_live_operation for pure SLP statements.
9870         (vectorizable_live_operation): Handle PHIs.
9871
9872 2018-01-16  Richard Biener  <rguenther@suse.de>
9873
9874         PR tree-optimization/83867
9875         * tree-vect-stmts.c (vect_transform_stmt): Precompute
9876         nested_in_vect_loop_p since the scalar stmt may get invalidated.
9877
9878 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
9879
9880         PR c/83844
9881         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9882         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9883         If off is not INTEGER_CST, issue a may not be aligned warning
9884         rather than isn't aligned.  Use isn%'t rather than isn't.
9885         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9886         into MULT_EXPR.
9887         <case MULT_EXPR>: Improve the case when bottom and one of the
9888         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9889         operand, in that case check if the other operand is multiple of
9890         bottom divided by the INTEGER_CST operand.
9891
9892 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
9893
9894         PR target/83858
9895         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9896         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9897         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9898         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9899         * config/pa/pa.c (pa_function_arg_advance): Likewise.
9900         (pa_function_arg, pa_arg_partial_bytes): Likewise.
9901         (pa_function_arg_size): New function.
9902
9903 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
9904
9905         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9906         in a separate statement.
9907
9908 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
9909
9910         PR tree-optimization/83847
9911         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9912         group gathers and scatters.
9913
9914 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
9915
9916         PR rtl-optimization/86620
9917         * params.def (max-sched-ready-insns): Bump minimum value to 1.
9918
9919         PR rtl-optimization/83213
9920         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9921         to last if both are JUMP_INSNs.
9922
9923         PR tree-optimization/83843
9924         * gimple-ssa-store-merging.c
9925         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9926         store_immediate_info for bswap/nop orig_stores.
9927
9928 2018-01-15  Andrew Waterman  <andrew@sifive.com>
9929
9930         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9931         !TARGET_MUL.
9932         <UDIV>: Increase cost if !TARGET_DIV.
9933
9934 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
9935
9936         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9937         (define_attr "cr_logical_3op"): New.
9938         (cceq_ior_compare): Adjust.
9939         (cceq_ior_compare_complement): Adjust.
9940         (*cceq_rev_compare): Adjust.
9941         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9942         (is_cracked_insn): Adjust.
9943         (insn_must_be_first_in_group): Adjust.
9944         * config/rs6000/40x.md: Adjust.
9945         * config/rs6000/440.md: Adjust.
9946         * config/rs6000/476.md: Adjust.
9947         * config/rs6000/601.md: Adjust.
9948         * config/rs6000/603.md: Adjust.
9949         * config/rs6000/6xx.md: Adjust.
9950         * config/rs6000/7450.md: Adjust.
9951         * config/rs6000/7xx.md: Adjust.
9952         * config/rs6000/8540.md: Adjust.
9953         * config/rs6000/cell.md: Adjust.
9954         * config/rs6000/e300c2c3.md: Adjust.
9955         * config/rs6000/e500mc.md: Adjust.
9956         * config/rs6000/e500mc64.md: Adjust.
9957         * config/rs6000/e5500.md: Adjust.
9958         * config/rs6000/e6500.md: Adjust.
9959         * config/rs6000/mpc.md: Adjust.
9960         * config/rs6000/power4.md: Adjust.
9961         * config/rs6000/power5.md: Adjust.
9962         * config/rs6000/power6.md: Adjust.
9963         * config/rs6000/power7.md: Adjust.
9964         * config/rs6000/power8.md: Adjust.
9965         * config/rs6000/power9.md: Adjust.
9966         * config/rs6000/rs64.md: Adjust.
9967         * config/rs6000/titan.md: Adjust.
9968
9969 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
9970
9971         * config/i386/predicates.md (indirect_branch_operand): Rewrite
9972         ix86_indirect_branch_register logic.
9973
9974 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
9975
9976         * config/i386/constraints.md (Bs): Update
9977         ix86_indirect_branch_register check.  Don't check
9978         ix86_indirect_branch_register with GOT_memory_operand.
9979         (Bw): Likewise.
9980         * config/i386/predicates.md (GOT_memory_operand): Don't check
9981         ix86_indirect_branch_register here.
9982         (GOT32_symbol_operand): Likewise.
9983
9984 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
9985
9986         * config/i386/predicates.md (constant_call_address_operand):
9987         Rewrite ix86_indirect_branch_register logic.
9988         (sibcall_insn_operand): Likewise.
9989
9990 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
9991
9992         * config/i386/constraints.md (Bs): Replace
9993         ix86_indirect_branch_thunk_register with
9994         ix86_indirect_branch_register.
9995         (Bw): Likewise.
9996         * config/i386/i386.md (indirect_jump): Likewise.
9997         (tablejump): Likewise.
9998         (*sibcall_memory): Likewise.
9999         (*sibcall_value_memory): Likewise.
10000         Peepholes of indirect call and jump via memory: Likewise.
10001         * config/i386/i386.opt: Likewise.
10002         * config/i386/predicates.md (indirect_branch_operand): Likewise.
10003         (GOT_memory_operand): Likewise.
10004         (call_insn_operand): Likewise.
10005         (sibcall_insn_operand): Likewise.
10006         (GOT32_symbol_operand): Likewise.
10007
10008 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
10009
10010         PR middle-end/83837
10011         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10012         type rather than type addr's type points to.
10013         (expand_omp_atomic_mutex): Likewise.
10014         (expand_omp_atomic): Likewise.
10015
10016 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
10017
10018         PR target/83839
10019         * config/i386/i386.c (output_indirect_thunk_function): Use
10020         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10021         for  __x86_return_thunk.
10022
10023 2018-01-15  Richard Biener  <rguenther@suse.de>
10024
10025         PR middle-end/83850
10026         * expmed.c (extract_bit_field_1): Fix typo.
10027
10028 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10029
10030         PR target/83687
10031         * config/arm/iterators.md (VF): New mode iterator.
10032         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10033         Remove integer-related logic from pattern.
10034         (neon_vabd<mode>_3): Likewise.
10035
10036 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
10037
10038         PR middle-end/82694
10039         * common.opt (fstrict-overflow): No longer an alias.
10040         (fwrapv-pointer): New option.
10041         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10042         also for pointer types based on flag_wrapv_pointer.
10043         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10044         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10045         opts->x_flag_wrapv got set.
10046         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10047         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10048         POINTER_TYPE_OVERFLOW_UNDEFINED.
10049         * match.pd: Likewise in address comparison pattern.
10050         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10051
10052 2018-01-15  Richard Biener  <rguenther@suse.de>
10053
10054         PR lto/83804
10055         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10056         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
10057         Reset type names to their identifier if their TYPE_DECL doesn't
10058         have linkage (and thus is used for ODR and devirt).
10059         (save_debug_info_for_decl): Remove.
10060         (save_debug_info_for_type): Likewise.
10061         (add_tree_to_fld_list): Adjust.
10062         * tree-pretty-print.c (dump_generic_node): Make dumping of
10063         type names more robust.
10064
10065 2018-01-15  Richard Biener  <rguenther@suse.de>
10066
10067         * BASE-VER: Bump to 8.0.1.
10068
10069 2018-01-14  Martin Sebor  <msebor@redhat.com>
10070
10071         PR other/83508
10072         * builtins.c (check_access): Avoid warning when the no-warning bit
10073         is set.
10074
10075 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
10076
10077         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10078         * ira-color (allocno_hard_regs_compare): Likewise.
10079
10080 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
10081
10082         PR target/83013
10083         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10084         Use .pushsection/.popsection.
10085
10086 2018-01-14  Martin Sebor  <msebor@redhat.com>
10087
10088         PR c++/81327
10089         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10090
10091 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
10092
10093         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10094         entry from extra_headers.
10095         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10096         extra_headers, make the list bitwise identical to the i?86-*-* one.
10097
10098 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
10099
10100         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10101         -mcmodel=large with -mindirect-branch=thunk,
10102         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10103         -mfunction-return=thunk-extern.
10104         * doc/invoke.texi: Document -mcmodel=large is incompatible with
10105         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10106         -mfunction-return=thunk and -mfunction-return=thunk-extern.
10107
10108 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
10109
10110         * config/i386/i386.c (print_reg): Print the name of the full
10111         integer register without '%'.
10112         (ix86_print_operand): Handle 'V'.
10113          * doc/extend.texi: Document 'V' modifier.
10114
10115 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
10116
10117         * config/i386/constraints.md (Bs): Disallow memory operand for
10118         -mindirect-branch-register.
10119         (Bw): Likewise.
10120         * config/i386/predicates.md (indirect_branch_operand): Likewise.
10121         (GOT_memory_operand): Likewise.
10122         (call_insn_operand): Likewise.
10123         (sibcall_insn_operand): Likewise.
10124         (GOT32_symbol_operand): Likewise.
10125         * config/i386/i386.md (indirect_jump): Call convert_memory_address
10126         for -mindirect-branch-register.
10127         (tablejump): Likewise.
10128         (*sibcall_memory): Likewise.
10129         (*sibcall_value_memory): Likewise.
10130         Disallow peepholes of indirect call and jump via memory for
10131         -mindirect-branch-register.
10132         (*call_pop): Replace m with Bw.
10133         (*call_value_pop): Likewise.
10134         (*sibcall_pop_memory): Replace m with Bs.
10135         * config/i386/i386.opt (mindirect-branch-register): New option.
10136         * doc/invoke.texi: Document -mindirect-branch-register option.
10137
10138 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
10139
10140         * config/i386/i386-protos.h (ix86_output_function_return): New.
10141         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10142         set function_return_type.
10143         (indirect_thunk_name): Add ret_p to indicate thunk for function
10144         return.
10145         (output_indirect_thunk_function): Pass false to
10146         indirect_thunk_name.
10147         (ix86_output_indirect_branch_via_reg): Likewise.
10148         (ix86_output_indirect_branch_via_push): Likewise.
10149         (output_indirect_thunk_function): Create alias for function
10150         return thunk if regno < 0.
10151         (ix86_output_function_return): New function.
10152         (ix86_handle_fndecl_attribute): Handle function_return.
10153         (ix86_attribute_table): Add function_return.
10154         * config/i386/i386.h (machine_function): Add
10155         function_return_type.
10156         * config/i386/i386.md (simple_return_internal): Use
10157         ix86_output_function_return.
10158         (simple_return_internal_long): Likewise.
10159         * config/i386/i386.opt (mfunction-return=): New option.
10160         (indirect_branch): Mention -mfunction-return=.
10161         * doc/extend.texi: Document function_return function attribute.
10162         * doc/invoke.texi: Document -mfunction-return= option.
10163
10164 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
10165
10166         * config/i386/i386-opts.h (indirect_branch): New.
10167         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10168         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10169         with local indirect jump when converting indirect call and jump.
10170         (ix86_set_indirect_branch_type): New.
10171         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10172         (indirectlabelno): New.
10173         (indirect_thunk_needed): Likewise.
10174         (indirect_thunk_bnd_needed): Likewise.
10175         (indirect_thunks_used): Likewise.
10176         (indirect_thunks_bnd_used): Likewise.
10177         (INDIRECT_LABEL): Likewise.
10178         (indirect_thunk_name): Likewise.
10179         (output_indirect_thunk): Likewise.
10180         (output_indirect_thunk_function): Likewise.
10181         (ix86_output_indirect_branch_via_reg): Likewise.
10182         (ix86_output_indirect_branch_via_push): Likewise.
10183         (ix86_output_indirect_branch): Likewise.
10184         (ix86_output_indirect_jmp): Likewise.
10185         (ix86_code_end): Call output_indirect_thunk_function if needed.
10186         (ix86_output_call_insn): Call ix86_output_indirect_branch if
10187         needed.
10188         (ix86_handle_fndecl_attribute): Handle indirect_branch.
10189         (ix86_attribute_table): Add indirect_branch.
10190         * config/i386/i386.h (machine_function): Add indirect_branch_type
10191         and has_local_indirect_jump.
10192         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10193         to true.
10194         (tablejump): Likewise.
10195         (*indirect_jump): Use ix86_output_indirect_jmp.
10196         (*tablejump_1): Likewise.
10197         (simple_return_indirect_internal): Likewise.
10198         * config/i386/i386.opt (mindirect-branch=): New option.
10199         (indirect_branch): New.
10200         (keep): Likewise.
10201         (thunk): Likewise.
10202         (thunk-inline): Likewise.
10203         (thunk-extern): Likewise.
10204         * doc/extend.texi: Document indirect_branch function attribute.
10205         * doc/invoke.texi: Document -mindirect-branch= option.
10206
10207 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
10208
10209         PR ipa/83051
10210         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10211
10212 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
10213
10214         * ipa-inline.c (want_inline_small_function_p): Return false if
10215         inlining has already failed with CIF_FINAL_ERROR.
10216         (update_caller_keys): Call want_inline_small_function_p before
10217         can_inline_edge_p.
10218         (update_callee_keys): Likewise.
10219
10220 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10221
10222         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10223         New function.
10224         (rs6000_quadword_masked_address_p): Likewise.
10225         (quad_aligned_load_p): Likewise.
10226         (quad_aligned_store_p): Likewise.
10227         (const_load_sequence_p): Add comment to describe the outer-most loop.
10228         (mimic_memory_attributes_and_flags): New function.
10229         (rs6000_gen_stvx): Likewise.
10230         (replace_swapped_aligned_store): Likewise.
10231         (rs6000_gen_lvx): Likewise.
10232         (replace_swapped_aligned_load): Likewise.
10233         (replace_swapped_load_constant): Capitalize argument name in
10234         comment describing this function.
10235         (rs6000_analyze_swaps): Add a third pass to search for vector loads
10236         and stores that access quad-word aligned addresses and replace
10237         with stvx or lvx instructions when appropriate.
10238         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10239         New function prototype.
10240         (rs6000_quadword_masked_address_p): Likewise.
10241         (rs6000_gen_lvx): Likewise.
10242         (rs6000_gen_stvx): Likewise.
10243         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10244         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10245         when memory address is aligned.
10246         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10247         this split to select lvx instruction when memory address is aligned.
10248         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10249         instruction when memory address is aligned.
10250         (*vsx_le_perm_load_v16qi): Likewise.
10251         (four unnamed splitters): Modify to select the stvx instruction
10252         when memory is aligned.
10253
10254 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
10255
10256         * predict.c (determine_unlikely_bbs): Handle correctly BBs
10257         which appears in the queue multiple times.
10258
10259 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10260             Alan Hayward  <alan.hayward@arm.com>
10261             David Sherwood  <david.sherwood@arm.com>
10262
10263         * tree-vectorizer.h (vec_lower_bound): New structure.
10264         (_loop_vec_info): Add check_nonzero and lower_bounds.
10265         (LOOP_VINFO_CHECK_NONZERO): New macro.
10266         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10267         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10268         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10269         fields.  Make seg_len the distance travelled, not including the
10270         access size.
10271         (dr_direction_indicator): Declare.
10272         (dr_zero_step_indicator): Likewise.
10273         (dr_known_forward_stride_p): Likewise.
10274         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10275         tree-ssanames.h.
10276         (runtime_alias_check_p): Allow runtime alias checks with
10277         variable strides.
10278         (operator ==): Compare access_size and align.
10279         (prune_runtime_alias_test_list): Rework for new distinction between
10280         the access_size and seg_len.
10281         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
10282         segment lengths.
10283         (get_segment_min_max): New function.
10284         (create_intersect_range_checks): Use it.
10285         (dr_step_indicator): New function.
10286         (dr_direction_indicator): Likewise.
10287         (dr_zero_step_indicator): Likewise.
10288         (dr_known_forward_stride_p): Likewise.
10289         * tree-loop-distribution.c (data_ref_segment_size): Return
10290         DR_STEP * (niters - 1).
10291         (compute_alias_check_pairs): Update call to the dr_with_seg_len
10292         constructor.
10293         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10294         (vect_preserves_scalar_order_p): New function, split out from...
10295         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
10296         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10297         (vect_vfa_access_size): New function.
10298         (vect_vfa_align): Likewise.
10299         (vect_compile_time_alias): Take access_size_a and access_b arguments.
10300         (dump_lower_bound): New function.
10301         (vect_check_lower_bound): Likewise.
10302         (vect_small_gap_p): Likewise.
10303         (vectorizable_with_step_bound_p): Likewise.
10304         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10305         depencies if the vectorization factor is 1.  Convert the checks
10306         for nonzero steps into checks on the bounds of DR_STEP.  Try using
10307         a bunds check for variable steps if the minimum required step is
10308         relatively small. Update calls to the dr_with_seg_len
10309         constructor and to vect_compile_time_alias.
10310         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10311         function.
10312         (vect_loop_versioning): Call it.
10313         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10314         when retrying.
10315         (vect_estimate_min_profitable_iters): Account for any bounds checks.
10316
10317 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10318             Alan Hayward  <alan.hayward@arm.com>
10319             David Sherwood  <david.sherwood@arm.com>
10320
10321         * doc/sourcebuild.texi (vect_scatter_store): Document.
10322         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10323         optabs.
10324         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10325         Document.
10326         * genopinit.c (main): Add supports_vec_scatter_store and
10327         supports_vec_scatter_store_cached to target_optabs.
10328         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10329         IFN_MASK_SCATTER_STORE.
10330         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10331         functions.
10332         * internal-fn.h (internal_store_fn_p): Declare.
10333         (internal_fn_stored_value_index): Likewise.
10334         * internal-fn.c (scatter_store_direct): New macro.
10335         (expand_scatter_store_optab_fn): New function.
10336         (direct_scatter_store_optab_supported_p): New macro.
10337         (internal_store_fn_p): New function.
10338         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10339         IFN_MASK_SCATTER_STORE.
10340         (internal_fn_mask_index): Likewise.
10341         (internal_fn_stored_value_index): New function.
10342         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10343         for scatter stores.
10344         * optabs-query.h (supports_vec_scatter_store_p): Declare.
10345         * optabs-query.c (supports_vec_scatter_store_p): New function.
10346         * tree-vectorizer.h (vect_get_store_rhs): Declare.
10347         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10348         true for scatter stores.
10349         (vect_gather_scatter_fn_p): Handle scatter stores too.
10350         (vect_check_gather_scatter): Consider using scatter stores if
10351         supports_vec_scatter_store_p.
10352         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10353         scatter stores too.
10354         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10355         internal_fn_stored_value_index.
10356         (check_load_store_masking): Handle scatter stores too.
10357         (vect_get_store_rhs): Make public.
10358         (vectorizable_call): Use internal_store_fn_p.
10359         (vectorizable_store): Handle scatter store internal functions.
10360         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10361         when deciding whether the end of the group has been reached.
10362         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10363         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10364         (mask_scatter_store<mode>): New insns.
10365
10366 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10367             Alan Hayward  <alan.hayward@arm.com>
10368             David Sherwood  <david.sherwood@arm.com>
10369
10370         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10371         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10372         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10373         function.
10374         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10375         Use vect_truncate_gather_scatter_offset if we can't treat the
10376         operation as a normal gather load or scatter store.
10377         (get_group_load_store_type): Take the gather_scatter_info
10378         as argument.  Try using a gather load or scatter store for
10379         single-element groups.
10380         (get_load_store_type): Update calls to get_group_load_store_type
10381         and vect_use_strided_gather_scatters_p.
10382
10383 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10384             Alan Hayward  <alan.hayward@arm.com>
10385             David Sherwood  <david.sherwood@arm.com>
10386
10387         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10388         optional tree argument.
10389         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10390         null target hooks.
10391         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10392         but continue to use the current value as a fallback.
10393         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10394         to compare the updates.
10395         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10396         (get_load_store_type): Use it when handling a strided access.
10397         (vect_get_strided_load_store_ops): New function.
10398         (vect_get_data_ptr_increment): Likewise.
10399         (vectorizable_load): Handle strided gather loads.  Always pass
10400         a step to vect_create_data_ref_ptr and bump_vector_ptr.
10401
10402 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10403             Alan Hayward  <alan.hayward@arm.com>
10404             David Sherwood  <david.sherwood@arm.com>
10405
10406         * doc/md.texi (gather_load@var{m}): Document.
10407         (mask_gather_load@var{m}): Likewise.
10408         * genopinit.c (main): Add supports_vec_gather_load and
10409         supports_vec_gather_load_cached to target_optabs.
10410         * optabs-tree.c (init_tree_optimization_optabs): Use
10411         ggc_cleared_alloc to allocate target_optabs.
10412         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10413         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10414         functions.
10415         * internal-fn.h (internal_load_fn_p): Declare.
10416         (internal_gather_scatter_fn_p): Likewise.
10417         (internal_fn_mask_index): Likewise.
10418         (internal_gather_scatter_fn_supported_p): Likewise.
10419         * internal-fn.c (gather_load_direct): New macro.
10420         (expand_gather_load_optab_fn): New function.
10421         (direct_gather_load_optab_supported_p): New macro.
10422         (direct_internal_fn_optab): New function.
10423         (internal_load_fn_p): Likewise.
10424         (internal_gather_scatter_fn_p): Likewise.
10425         (internal_fn_mask_index): Likewise.
10426         (internal_gather_scatter_fn_supported_p): Likewise.
10427         * optabs-query.c (supports_at_least_one_mode_p): New function.
10428         (supports_vec_gather_load_p): Likewise.
10429         * optabs-query.h (supports_vec_gather_load_p): Declare.
10430         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10431         and memory_type field.
10432         (NUM_PATTERNS): Bump to 15.
10433         * tree-vect-data-refs.c: Include internal-fn.h.
10434         (vect_gather_scatter_fn_p): New function.
10435         (vect_describe_gather_scatter_call): Likewise.
10436         (vect_check_gather_scatter): Try using internal functions for
10437         gather loads.  Recognize existing calls to a gather load function.
10438         (vect_analyze_data_refs): Consider using gather loads if
10439         supports_vec_gather_load_p.
10440         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10441         (vect_get_gather_scatter_offset_type): Likewise.
10442         (vect_convert_mask_for_vectype): Likewise.
10443         (vect_add_conversion_to_patterm): Likewise.
10444         (vect_try_gather_scatter_pattern): Likewise.
10445         (vect_recog_gather_scatter_pattern): New pattern recognizer.
10446         (vect_vect_recog_func_ptrs): Add it.
10447         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10448         internal_fn_mask_index and internal_gather_scatter_fn_p.
10449         (check_load_store_masking): Take the gather_scatter_info as an
10450         argument and handle gather loads.
10451         (vect_get_gather_scatter_ops): New function.
10452         (vectorizable_call): Check internal_load_fn_p.
10453         (vectorizable_load): Likewise.  Handle gather load internal
10454         functions.
10455         (vectorizable_store): Update call to check_load_store_masking.
10456         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10457         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10458         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10459         (aarch64_gather_scale_operand_d): New predicates.
10460         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10461         (mask_gather_load<mode>): New insns.
10462
10463 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10464             Alan Hayward  <alan.hayward@arm.com>
10465             David Sherwood  <david.sherwood@arm.com>
10466
10467         * optabs.def (fold_left_plus_optab): New optab.
10468         * doc/md.texi (fold_left_plus_@var{m}): Document.
10469         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10470         * internal-fn.c (fold_left_direct): Define.
10471         (expand_fold_left_optab_fn): Likewise.
10472         (direct_fold_left_optab_supported_p): Likewise.
10473         * fold-const-call.c (fold_const_fold_left): New function.
10474         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10475         * tree-parloops.c (valid_reduction_p): New function.
10476         (gather_scalar_reductions): Use it.
10477         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10478         (vect_finish_replace_stmt): Declare.
10479         * tree-vect-loop.c (fold_left_reduction_fn): New function.
10480         (needs_fold_left_reduction_p): New function, split out from...
10481         (vect_is_simple_reduction): ...here.  Accept reductions that
10482         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10483         (vect_force_simple_reduction): Also store the reduction type in
10484         the assignment's STMT_VINFO_REDUC_TYPE.
10485         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10486         (merge_with_identity): New function.
10487         (vect_expand_fold_left): Likewise.
10488         (vectorize_fold_left_reduction): Likewise.
10489         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
10490         scalar phi in place for it.  Check for target support and reject
10491         cases that would reassociate the operation.  Defer the transform
10492         phase to vectorize_fold_left_reduction.
10493         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10494         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10495         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10496
10497 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10498
10499         * tree-if-conv.c (predicate_mem_writes): Remove redundant
10500         call to ifc_temp_var.
10501
10502 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10503             Alan Hayward  <alan.hayward@arm.com>
10504             David Sherwood  <david.sherwood@arm.com>
10505
10506         * target.def (legitimize_address_displacement): Take the original
10507         offset as a poly_int.
10508         * targhooks.h (default_legitimize_address_displacement): Update
10509         accordingly.
10510         * targhooks.c (default_legitimize_address_displacement): Likewise.
10511         * doc/tm.texi: Regenerate.
10512         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10513         as an argument, moving assert of ad->disp == ad->disp_term to...
10514         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
10515         Try calling targetm.legitimize_address_displacement before expanding
10516         the address rather than afterwards, and adjust for the new interface.
10517         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10518         Match the new hook interface.  Handle SVE addresses.
10519         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10520         new hook interface.
10521
10522 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10523
10524         * Makefile.in (OBJS): Add early-remat.o.
10525         * target.def (select_early_remat_modes): New hook.
10526         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10527         * doc/tm.texi: Regenerate.
10528         * targhooks.h (default_select_early_remat_modes): Declare.
10529         * targhooks.c (default_select_early_remat_modes): New function.
10530         * timevar.def (TV_EARLY_REMAT): New timevar.
10531         * passes.def (pass_early_remat): New pass.
10532         * tree-pass.h (make_pass_early_remat): Declare.
10533         * early-remat.c: New file.
10534         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10535         function.
10536         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10537
10538 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10539             Alan Hayward  <alan.hayward@arm.com>
10540             David Sherwood  <david.sherwood@arm.com>
10541
10542         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10543         vfm1 with a bound_epilog parameter.
10544         (vect_do_peeling): Update calls accordingly, and move the prologue
10545         call earlier in the function.  Treat the base bound_epilog as 0 for
10546         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
10547         this base when peeling for gaps.
10548         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10549         with fully-masked loops.
10550         (vect_estimate_min_profitable_iters): Handle the single peeled
10551         iteration in that case.
10552
10553 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10554             Alan Hayward  <alan.hayward@arm.com>
10555             David Sherwood  <david.sherwood@arm.com>
10556
10557         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10558         single-element interleaving even if the size is not a power of 2.
10559         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10560         accesses for single-element interleaving if the group size is
10561         not a power of 2.
10562
10563 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10564             Alan Hayward  <alan.hayward@arm.com>
10565             David Sherwood  <david.sherwood@arm.com>
10566
10567         * doc/md.texi (fold_extract_last_@var{m}): Document.
10568         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10569         * optabs.def (fold_extract_last_optab): New optab.
10570         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10571         * internal-fn.c (fold_extract_direct): New macro.
10572         (expand_fold_extract_optab_fn): Likewise.
10573         (direct_fold_extract_optab_supported_p): Likewise.
10574         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10575         * tree-vect-loop.c (vect_model_reduction_cost): Handle
10576         EXTRACT_LAST_REDUCTION.
10577         (get_initial_def_for_reduction): Do not create an initial vector
10578         for EXTRACT_LAST_REDUCTION reductions.
10579         (vectorizable_reduction): Leave the scalar phi in place for
10580         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
10581         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
10582         epilogue code for EXTRACT_LAST_REDUCTION and defer the
10583         transform phase to vectorizable_condition.
10584         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10585         split out from...
10586         (vect_finish_stmt_generation): ...here.
10587         (vect_finish_replace_stmt): New function.
10588         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10589         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10590         pattern.
10591         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10592
10593 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10594             Alan Hayward  <alan.hayward@arm.com>
10595             David Sherwood  <david.sherwood@arm.com>
10596
10597         * doc/md.texi (extract_last_@var{m}): Document.
10598         * optabs.def (extract_last_optab): New optab.
10599         * internal-fn.def (EXTRACT_LAST): New internal function.
10600         * internal-fn.c (cond_unary_direct): New macro.
10601         (expand_cond_unary_optab_fn): Likewise.
10602         (direct_cond_unary_optab_supported_p): Likewise.
10603         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10604         loops using EXTRACT_LAST.
10605         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10606         (extract_last_<mode>): ...this optab.
10607         (vec_extract<mode><Vel>): Update accordingly.
10608
10609 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10610             Alan Hayward  <alan.hayward@arm.com>
10611             David Sherwood  <david.sherwood@arm.com>
10612
10613         * target.def (empty_mask_is_expensive): New hook.
10614         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10615         * doc/tm.texi: Regenerate.
10616         * targhooks.h (default_empty_mask_is_expensive): Declare.
10617         * targhooks.c (default_empty_mask_is_expensive): New function.
10618         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10619         if the target says that empty masks are expensive.
10620         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10621         New function.
10622         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10623
10624 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10625             Alan Hayward  <alan.hayward@arm.com>
10626             David Sherwood  <david.sherwood@arm.com>
10627
10628         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10629         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10630         (vect_use_loop_mask_for_alignment_p): New function.
10631         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10632         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10633         niters_skip argument.  Make sure that the first niters_skip elements
10634         of the first iteration are inactive.
10635         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10636         Update call to vect_set_loop_masks_directly.
10637         (get_misalign_in_elems): New function, split out from...
10638         (vect_gen_prolog_loop_niters): ...here.
10639         (vect_update_init_of_dr): Take a code argument that specifies whether
10640         the adjustment should be added or subtracted.
10641         (vect_update_init_of_drs): Likewise.
10642         (vect_prepare_for_masked_peels): New function.
10643         (vect_do_peeling): Skip prologue peeling if we're using a mask
10644         instead.  Update call to vect_update_inits_of_drs.
10645         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10646         mask_skip_niters.
10647         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10648         alignment.  Do not include the number of peeled iterations in
10649         the minimum threshold in that case.
10650         (vectorizable_induction): Adjust the start value down by
10651         LOOP_VINFO_MASK_SKIP_NITERS iterations.
10652         (vect_transform_loop): Call vect_prepare_for_masked_peels.
10653         Take the number of skipped iterations into account when calculating
10654         the loop bounds.
10655         * tree-vect-stmts.c (vect_gen_while_not): New function.
10656
10657 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10658             Alan Hayward  <alan.hayward@arm.com>
10659             David Sherwood  <david.sherwood@arm.com>
10660
10661         * doc/sourcebuild.texi (vect_fully_masked): Document.
10662         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10663         default value to 0.
10664         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10665         split out from...
10666         (vect_analyze_loop_2): ...here. Don't check the vectorization
10667         factor against the number of loop iterations if the loop is
10668         fully-masked.
10669
10670 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10671             Alan Hayward  <alan.hayward@arm.com>
10672             David Sherwood  <david.sherwood@arm.com>
10673
10674         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10675         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10676         (dump_groups): Update accordingly.
10677         (iv_use::mem_type): New member variable.
10678         (address_p): New function.
10679         (record_use): Add a mem_type argument and initialize the new
10680         mem_type field.
10681         (record_group_use): Add a mem_type argument.  Use address_p.
10682         Remove obsolete null checks of base_object.  Update call to record_use.
10683         (find_interesting_uses_op): Update call to record_group_use.
10684         (find_interesting_uses_cond): Likewise.
10685         (find_interesting_uses_address): Likewise.
10686         (get_mem_type_for_internal_fn): New function.
10687         (find_address_like_use): Likewise.
10688         (find_interesting_uses_stmt): Try find_address_like_use before
10689         calling find_interesting_uses_op.
10690         (addr_offset_valid_p): Use the iv mem_type field as the type
10691         of the addressed memory.
10692         (add_autoinc_candidates): Likewise.
10693         (get_address_cost): Likewise.
10694         (split_small_address_groups_p): Use address_p.
10695         (split_address_groups): Likewise.
10696         (add_iv_candidate_for_use): Likewise.
10697         (autoinc_possible_for_pair): Likewise.
10698         (rewrite_groups): Likewise.
10699         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10700         (determine_group_iv_cost): Update after split of USE_ADDRESS.
10701         (get_alias_ptr_type_for_ptr_address): New function.
10702         (rewrite_use_address): Rewrite address uses in calls that were
10703         identified by find_address_like_use.
10704
10705 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10706             Alan Hayward  <alan.hayward@arm.com>
10707             David Sherwood  <david.sherwood@arm.com>
10708
10709         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10710         TARGET_MEM_REFs.
10711         * gimple-expr.h (is_gimple_addressable: Likewise.
10712         * gimple-expr.c (is_gimple_address): Likewise.
10713         * internal-fn.c (expand_call_mem_ref): New function.
10714         (expand_mask_load_optab_fn): Use it.
10715         (expand_mask_store_optab_fn): Likewise.
10716
10717 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10718             Alan Hayward  <alan.hayward@arm.com>
10719             David Sherwood  <david.sherwood@arm.com>
10720
10721         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10722         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10723         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10724         (cond_umax@var{mode}): Document.
10725         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10726         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10727         (cond_umin_optab, cond_umax_optab): New optabs.
10728         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10729         (COND_IOR, COND_XOR): New internal functions.
10730         * internal-fn.h (get_conditional_internal_fn): Declare.
10731         * internal-fn.c (cond_binary_direct): New macro.
10732         (expand_cond_binary_optab_fn): Likewise.
10733         (direct_cond_binary_optab_supported_p): Likewise.
10734         (get_conditional_internal_fn): New function.
10735         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10736         Cope with reduction statements that are vectorized as calls rather
10737         than assignments.
10738         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10739         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10740         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10741         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10742         (UNSPEC_COND_EOR): New unspecs.
10743         (optab): Add mappings for them.
10744         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10745         (sve_int_op, sve_fp_op): New int attributes.
10746
10747 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10748             Alan Hayward  <alan.hayward@arm.com>
10749             David Sherwood  <david.sherwood@arm.com>
10750
10751         * optabs.def (while_ult_optab): New optab.
10752         * doc/md.texi (while_ult@var{m}@var{n}): Document.
10753         * internal-fn.def (WHILE_ULT): New internal function.
10754         * internal-fn.h (direct_internal_fn_supported_p): New override
10755         that takes two types as argument.
10756         * internal-fn.c (while_direct): New macro.
10757         (expand_while_optab_fn): New function.
10758         (convert_optab_supported_p): Likewise.
10759         (direct_while_optab_supported_p): New macro.
10760         * wide-int.h (wi::udiv_ceil): New function.
10761         * tree-vectorizer.h (rgroup_masks): New structure.
10762         (vec_loop_masks): New typedef.
10763         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10764         and fully_masked_p.
10765         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10766         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10767         (vect_max_vf): New function.
10768         (slpeel_make_loop_iterate_ntimes): Delete.
10769         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10770         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10771         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10772         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10773         internal-fn.h, stor-layout.h and optabs-query.h.
10774         (vect_set_loop_mask): New function.
10775         (add_preheader_seq): Likewise.
10776         (add_header_seq): Likewise.
10777         (interleave_supported_p): Likewise.
10778         (vect_maybe_permute_loop_masks): Likewise.
10779         (vect_set_loop_masks_directly): Likewise.
10780         (vect_set_loop_condition_masked): Likewise.
10781         (vect_set_loop_condition_unmasked): New function, split out from
10782         slpeel_make_loop_iterate_ntimes.
10783         (slpeel_make_loop_iterate_ntimes): Rename to..
10784         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
10785         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10786         (vect_do_peeling): Update call accordingly.
10787         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10788         loops.
10789         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10790         mask_compare_type, can_fully_mask_p and fully_masked_p.
10791         (release_vec_loop_masks): New function.
10792         (_loop_vec_info): Use it to free the loop masks.
10793         (can_produce_all_loop_masks_p): New function.
10794         (vect_get_max_nscalars_per_iter): Likewise.
10795         (vect_verify_full_masking): Likewise.
10796         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10797         retries, and free the mask rgroups before retrying.  Check loop-wide
10798         reasons for disallowing fully-masked loops.  Make the final decision
10799         about whether use a fully-masked loop or not.
10800         (vect_estimate_min_profitable_iters): Do not assume that peeling
10801         for the number of iterations will be needed for fully-masked loops.
10802         (vectorizable_reduction): Disable fully-masked loops.
10803         (vectorizable_live_operation): Likewise.
10804         (vect_halve_mask_nunits): New function.
10805         (vect_double_mask_nunits): Likewise.
10806         (vect_record_loop_mask): Likewise.
10807         (vect_get_loop_mask): Likewise.
10808         (vect_transform_loop): Handle the case in which the final loop
10809         iteration might handle a partial vector.  Call vect_set_loop_condition
10810         instead of slpeel_make_loop_iterate_ntimes.
10811         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10812         (check_load_store_masking): New function.
10813         (prepare_load_store_mask): Likewise.
10814         (vectorizable_store): Handle fully-masked loops.
10815         (vectorizable_load): Likewise.
10816         (supportable_widening_operation): Use vect_halve_mask_nunits for
10817         booleans.
10818         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10819         (vect_gen_while): New function.
10820         * config/aarch64/aarch64.md (umax<mode>3): New expander.
10821         (aarch64_uqdec<mode>): New insn.
10822
10823 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10824             Alan Hayward  <alan.hayward@arm.com>
10825             David Sherwood  <david.sherwood@arm.com>
10826
10827         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10828         (reduc_xor_scal_optab): New optabs.
10829         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10830         (reduc_xor_scal_@var{m}): Document.
10831         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10832         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10833         internal functions.
10834         * fold-const-call.c (fold_const_call): Handle them.
10835         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10836         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10837         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10838         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10839         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10840         (UNSPEC_XORV): New unspecs.
10841         (optab): Add entries for them.
10842         (BITWISEV): New int iterator.
10843         (bit_reduc_op): New int attributes.
10844
10845 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10846             Alan Hayward  <alan.hayward@arm.com>
10847             David Sherwood  <david.sherwood@arm.com>
10848
10849         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10850         * internal-fn.def (VEC_SHL_INSERT): New internal function.
10851         * optabs.def (vec_shl_insert_optab): New optab.
10852         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10853         (duplicate_and_interleave): Likewise.
10854         * tree-vect-loop.c: Include internal-fn.h.
10855         (neutral_op_for_slp_reduction): New function, split out from
10856         get_initial_defs_for_reduction.
10857         (get_initial_def_for_reduction): Handle option 2 for variable-length
10858         vectors by loading the neutral value into a vector and then shifting
10859         the initial value into element 0.
10860         (get_initial_defs_for_reduction): Replace the code argument with
10861         the neutral value calculated by neutral_op_for_slp_reduction.
10862         Use gimple_build_vector for constant-length vectors.
10863         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10864         but the first group_size elements have a neutral value.
10865         Use duplicate_and_interleave otherwise.
10866         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10867         Update call to get_initial_defs_for_reduction.  Handle SLP
10868         reductions for variable-length vectors by creating one vector
10869         result for each scalar result, with the elements associated
10870         with other scalar results stubbed out with the neutral value.
10871         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10872         Require IFN_VEC_SHL_INSERT for double reductions on
10873         variable-length vectors, or SLP reductions that have
10874         a neutral value.  Require can_duplicate_and_interleave_p
10875         support for variable-length unchained SLP reductions if there
10876         is no neutral value, such as for MIN/MAX reductions.  Also require
10877         the number of vector elements to be a multiple of the number of
10878         SLP statements when doing variable-length unchained SLP reductions.
10879         Update call to vect_create_epilog_for_reduction.
10880         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10881         and remove initial values.
10882         (duplicate_and_interleave): Make public.
10883         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10884         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10885
10886 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10887             Alan Hayward  <alan.hayward@arm.com>
10888             David Sherwood  <david.sherwood@arm.com>
10889
10890         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10891         (can_duplicate_and_interleave_p): New function.
10892         (vect_get_and_check_slp_defs): Take the vector of statements
10893         rather than just the current one.  Remove excess parentheses.
10894         Restriction rejectinon of vect_constant_def and vect_external_def
10895         for variable-length vectors to boolean types, or types for which
10896         can_duplicate_and_interleave_p is false.
10897         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10898         (duplicate_and_interleave): New function.
10899         (vect_get_constant_vectors): Use gimple_build_vector for
10900         constant-length vectors and suitable variable-length constant
10901         vectors.  Use duplicate_and_interleave for other variable-length
10902         vectors.  Don't defer the update when inserting new statements.
10903
10904 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10905             Alan Hayward  <alan.hayward@arm.com>
10906             David Sherwood  <david.sherwood@arm.com>
10907
10908         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10909         min_profitable_iters doesn't go negative.
10910
10911 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10912             Alan Hayward  <alan.hayward@arm.com>
10913             David Sherwood  <david.sherwood@arm.com>
10914
10915         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10916         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10917         * optabs.def (vec_mask_load_lanes_optab): New optab.
10918         (vec_mask_store_lanes_optab): Likewise.
10919         * internal-fn.def (MASK_LOAD_LANES): New internal function.
10920         (MASK_STORE_LANES): Likewise.
10921         * internal-fn.c (mask_load_lanes_direct): New macro.
10922         (mask_store_lanes_direct): Likewise.
10923         (expand_mask_load_optab_fn): Handle masked operations.
10924         (expand_mask_load_lanes_optab_fn): New macro.
10925         (expand_mask_store_optab_fn): Handle masked operations.
10926         (expand_mask_store_lanes_optab_fn): New macro.
10927         (direct_mask_load_lanes_optab_supported_p): Likewise.
10928         (direct_mask_store_lanes_optab_supported_p): Likewise.
10929         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10930         parameter.
10931         (vect_load_lanes_supported): Likewise.
10932         * tree-vect-data-refs.c (strip_conversion): New function.
10933         (can_group_stmts_p): Likewise.
10934         (vect_analyze_data_ref_accesses): Use it instead of checking
10935         for a pair of assignments.
10936         (vect_store_lanes_supported): Take a masked_p parameter.
10937         (vect_load_lanes_supported): Likewise.
10938         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10939         vect_store_lanes_supported and vect_load_lanes_supported.
10940         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10941         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10942         parameter.  Don't allow gaps for masked accesses.
10943         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
10944         and vect_load_lanes_supported.
10945         (get_load_store_type): Take a masked_p parameter and update
10946         call to get_group_load_store_type.
10947         (vectorizable_store): Update call to get_load_store_type.
10948         Handle IFN_MASK_STORE_LANES.
10949         (vectorizable_load): Update call to get_load_store_type.
10950         Handle IFN_MASK_LOAD_LANES.
10951
10952 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10953             Alan Hayward  <alan.hayward@arm.com>
10954             David Sherwood  <david.sherwood@arm.com>
10955
10956         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10957         modes for SVE.
10958         * config/aarch64/aarch64-protos.h
10959         (aarch64_sve_struct_memory_operand_p): Declare.
10960         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10961         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10962         (VPRED, vpred): Handle SVE structure modes.
10963         * config/aarch64/constraints.md (Utx): New constraint.
10964         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10965         (aarch64_sve_struct_nonimmediate_operand): New predicates.
10966         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10967         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10968         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10969         structure modes.  Split into pieces after RA.
10970         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10971         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10972         New patterns.
10973         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10974         SVE structure modes.
10975         (aarch64_classify_address): Likewise.
10976         (sizetochar): Move earlier in file.
10977         (aarch64_print_operand): Handle SVE register lists.
10978         (aarch64_array_mode): New function.
10979         (aarch64_sve_struct_memory_operand_p): Likewise.
10980         (TARGET_ARRAY_MODE): Redefine.
10981
10982 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10983             Alan Hayward  <alan.hayward@arm.com>
10984             David Sherwood  <david.sherwood@arm.com>
10985
10986         * target.def (array_mode): New target hook.
10987         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10988         * doc/tm.texi: Regenerate.
10989         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10990         * hooks.c (hook_optmode_mode_uhwi_none): New function.
10991         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10992         targetm.array_mode.
10993         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
10994         type sizes.
10995
10996 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
10997             Alan Hayward  <alan.hayward@arm.com>
10998             David Sherwood  <david.sherwood@arm.com>
10999
11000         * fold-const.c (fold_binary_loc): Check the argument types
11001         rather than the result type when testing for a vector operation.
11002
11003 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11004
11005         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11006         * doc/tm.texi: Regenerate.
11007
11008 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11009             Alan Hayward  <alan.hayward@arm.com>
11010             David Sherwood  <david.sherwood@arm.com>
11011
11012         * doc/invoke.texi (-msve-vector-bits=): Document new option.
11013         (sve): Document new AArch64 extension.
11014         * doc/md.texi (w): Extend the description of the AArch64
11015         constraint to include SVE vectors.
11016         (Upl, Upa): Document new AArch64 predicate constraints.
11017         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11018         enum.
11019         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11020         (msve-vector-bits=): New option.
11021         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11022         SVE when these are disabled.
11023         (sve): New extension.
11024         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11025         modes.  Adjust their number of units based on aarch64_sve_vg.
11026         (MAX_BITSIZE_MODE_ANY_MODE): Define.
11027         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11028         aarch64_addr_query_type.
11029         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11030         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11031         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11032         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11033         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11034         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11035         (aarch64_simd_imm_zero_p): Delete.
11036         (aarch64_check_zero_based_sve_index_immediate): Declare.
11037         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11038         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11039         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11040         (aarch64_sve_float_mul_immediate_p): Likewise.
11041         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11042         rather than an rtx.
11043         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11044         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11045         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11046         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11047         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11048         (aarch64_regmode_natural_size): Likewise.
11049         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11050         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11051         left one place.
11052         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11053         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11054         for VG and the SVE predicate registers.
11055         (V_ALIASES): Add a "z"-prefixed alias.
11056         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11057         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11058         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11059         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11060         (REG_CLASS_NAMES): Add entries for them.
11061         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
11062         and the predicate registers.
11063         (aarch64_sve_vg): Declare.
11064         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11065         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11066         (REGMODE_NATURAL_SIZE): Define.
11067         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11068         SVE macros.
11069         * config/aarch64/aarch64.c: Include cfgrtl.h.
11070         (simd_immediate_info): Add a constructor for series vectors,
11071         and an associated step field.
11072         (aarch64_sve_vg): New variable.
11073         (aarch64_dbx_register_number): Handle VG and the predicate registers.
11074         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11075         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11076         (VEC_ANY_DATA, VEC_STRUCT): New constants.
11077         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11078         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11079         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11080         (aarch64_get_mask_mode): New functions.
11081         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11082         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11083         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
11084         predicate modes and predicate registers.  Explicitly restrict
11085         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
11086         to store a vector mode if it is recognized by
11087         aarch64_classify_vector_mode.
11088         (aarch64_regmode_natural_size): New function.
11089         (aarch64_hard_regno_caller_save_mode): Return the original mode
11090         for predicates.
11091         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11092         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11093         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11094         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11095         functions.
11096         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
11097         does not overlap dest if the function is frame-related.  Handle
11098         SVE constants.
11099         (aarch64_split_add_offset): New function.
11100         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11101         them aarch64_add_offset.
11102         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11103         and update call to aarch64_sub_sp.
11104         (aarch64_add_cfa_expression): New function.
11105         (aarch64_expand_prologue): Pass extra temporary registers to the
11106         functions above.  Handle the case in which we need to emit new
11107         DW_CFA_expressions for registers that were originally saved
11108         relative to the stack pointer, but now have to be expressed
11109         relative to the frame pointer.
11110         (aarch64_output_mi_thunk): Pass extra temporary registers to the
11111         functions above.
11112         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
11113         IP0 and IP1 values for SVE frames.
11114         (aarch64_expand_vec_series): New function.
11115         (aarch64_expand_sve_widened_duplicate): Likewise.
11116         (aarch64_expand_sve_const_vector): Likewise.
11117         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11118         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
11119         into the register, rather than emitting a SET directly.
11120         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11121         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11122         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11123         (offset_9bit_signed_scaled_p): New functions.
11124         (aarch64_replicate_bitmask_imm): New function.
11125         (aarch64_bitmask_imm): Use it.
11126         (aarch64_cannot_force_const_mem): Reject expressions involving
11127         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
11128         (aarch64_classify_index): Handle SVE indices, by requiring
11129         a plain register index with a scale that matches the element size.
11130         (aarch64_classify_address): Handle SVE addresses.  Assert that
11131         the mode of the address is VOIDmode or an integer mode.
11132         Update call to aarch64_classify_symbol.
11133         (aarch64_classify_symbolic_expression): Update call to
11134         aarch64_classify_symbol.
11135         (aarch64_const_vec_all_in_range_p): New function.
11136         (aarch64_print_vector_float_operand): Likewise.
11137         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
11138         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
11139         and the FP immediates 1.0 and 0.5.
11140         (aarch64_print_address_internal): Handle SVE addresses.
11141         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11142         (aarch64_regno_regclass): Handle predicate registers.
11143         (aarch64_secondary_reload): Handle big-endian reloads of SVE
11144         data modes.
11145         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11146         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11147         (aarch64_convert_sve_vector_bits): New function.
11148         (aarch64_override_options): Use it to handle -msve-vector-bits=.
11149         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11150         rather than an rtx.
11151         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11152         Handle SVE vector and predicate modes.  Accept VL-based constants
11153         that need only one temporary register, and VL offsets that require
11154         no temporary registers.
11155         (aarch64_conditional_register_usage): Mark the predicate registers
11156         as fixed if SVE isn't available.
11157         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11158         Return true for SVE vector and predicate modes.
11159         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11160         rather than an unsigned int.  Handle SVE modes.
11161         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
11162         SVE modes.
11163         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11164         if SVE is enabled.
11165         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11166         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11167         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11168         (aarch64_sve_float_mul_immediate_p): New functions.
11169         (aarch64_sve_valid_immediate): New function.
11170         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11171         Explicitly reject structure modes.  Check for INDEX constants.
11172         Handle PTRUE and PFALSE constants.
11173         (aarch64_check_zero_based_sve_index_immediate): New function.
11174         (aarch64_simd_imm_zero_p): Delete.
11175         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11176         vector modes.  Accept constants in the range of CNT[BHWD].
11177         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11178         ask for an Advanced SIMD mode.
11179         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11180         (aarch64_simd_vector_alignment): Handle SVE predicates.
11181         (aarch64_vectorize_preferred_vector_alignment): New function.
11182         (aarch64_simd_vector_alignment_reachable): Use it instead of
11183         the vector size.
11184         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11185         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11186         functions.
11187         (MAX_VECT_LEN): Delete.
11188         (expand_vec_perm_d): Add a vec_flags field.
11189         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11190         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11191         (aarch64_evpc_ext): Don't apply a big-endian lane correction
11192         for SVE modes.
11193         (aarch64_evpc_rev): Rename to...
11194         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
11195         (aarch64_evpc_rev_global): New function.
11196         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11197         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11198         MAX_VECT_LEN.
11199         (aarch64_evpc_sve_tbl): New function.
11200         (aarch64_expand_vec_perm_const_1): Update after rename of
11201         aarch64_evpc_rev.  Handle SVE permutes too, trying
11202         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11203         than aarch64_evpc_tbl.
11204         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11205         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11206         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11207         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11208         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11209         (aarch64_expand_sve_vcond): New functions.
11210         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11211         of aarch64_vector_mode_p.
11212         (aarch64_dwarf_poly_indeterminate_value): New function.
11213         (aarch64_compute_pressure_classes): Likewise.
11214         (aarch64_can_change_mode_class): Likewise.
11215         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11216         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11217         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11218         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11219         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11220         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11221         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11222         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11223         constraints.
11224         (Dn, Dl, Dr): Accept const as well as const_vector.
11225         (Dz): Likewise.  Compare against CONST0_RTX.
11226         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11227         of "vector" where appropriate.
11228         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11229         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11230         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11231         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11232         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11233         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11234         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11235         (v_int_equiv): Extend to SVE modes.
11236         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11237         mode attributes.
11238         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11239         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11240         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11241         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11242         (SVE_COND_FP_CMP): New int iterators.
11243         (perm_hilo): Handle the new unpack unspecs.
11244         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11245         attributes.
11246         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11247         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11248         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11249         (aarch64_equality_operator, aarch64_constant_vector_operand)
11250         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11251         (aarch64_sve_nonimmediate_operand): Likewise.
11252         (aarch64_sve_general_operand): Likewise.
11253         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11254         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11255         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11256         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11257         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11258         (aarch64_sve_float_arith_immediate): Likewise.
11259         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11260         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11261         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11262         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11263         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11264         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11265         (aarch64_sve_float_arith_operand): Likewise.
11266         (aarch64_sve_float_arith_with_sub_operand): Likewise.
11267         (aarch64_sve_float_mul_operand): Likewise.
11268         (aarch64_sve_vec_perm_operand): Likewise.
11269         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11270         (aarch64_mov_operand): Accept const_poly_int and const_vector.
11271         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11272         as well as const_vector.
11273         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11274         in file.  Use CONST0_RTX and CONSTM1_RTX.
11275         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
11276         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11277         Use aarch64_simd_imm_zero.
11278         * config/aarch64/aarch64-sve.md: New file.
11279         * config/aarch64/aarch64.md: Include it.
11280         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11281         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11282         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11283         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11284         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11285         (sve): New attribute.
11286         (enabled): Disable instructions with the sve attribute unless
11287         TARGET_SVE.
11288         (movqi, movhi): Pass CONST_POLY_INT operaneds through
11289         aarch64_expand_mov_immediate.
11290         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11291         CNT[BHSD] immediates.
11292         (movti): Split CONST_POLY_INT moves into two halves.
11293         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11294         Split additions that need a temporary here if the destination
11295         is the stack pointer.
11296         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11297         (*add<mode>3_poly_1): New instruction.
11298         (set_clobber_cc): New expander.
11299
11300 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11301
11302         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11303         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
11304         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11305         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11306         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
11307         Change innermode from fixed_mode_size to machine_mode.
11308         (simplify_subreg): Update call accordingly.  Handle a constant-sized
11309         subreg of a variable-length CONST_VECTOR.
11310
11311 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
11312             Alan Hayward  <alan.hayward@arm.com>
11313             David Sherwood  <david.sherwood@arm.com>
11314
11315         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11316         (add_offset_to_base): New function, split out from...
11317         (create_mem_ref): ...here.  When handling a scale other than 1,
11318         check first whether the address is valid without the offset.
11319         Add it into the base if so, leaving the index and scale as-is.
11320
11321 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
11322
11323         PR c++/83778
11324         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11325         fold_for_warn before checking if arg2 is INTEGER_CST.
11326
11327 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
11328
11329         * config/rs6000/predicates.md (load_multiple_operation): Delete.
11330         (store_multiple_operation): Delete.
11331         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11332         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11333         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11334         guarded by TARGET_STRING.
11335         (rs6000_output_load_multiple): Delete.
11336         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11337         OPTION_MASK_STRING / TARGET_STRING handling.
11338         (print_operand) <'N', 'O'>: Add comment that these are unused now.
11339         (const rs6000_opt_masks) <"string">: Change mask to 0.
11340         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11341         (MASK_STRING): Delete.
11342         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11343         parts.  Simplify.
11344         (load_multiple): Delete.
11345         (*ldmsi8): Delete.
11346         (*ldmsi7): Delete.
11347         (*ldmsi6): Delete.
11348         (*ldmsi5): Delete.
11349         (*ldmsi4): Delete.
11350         (*ldmsi3): Delete.
11351         (store_multiple): Delete.
11352         (*stmsi8): Delete.
11353         (*stmsi7): Delete.
11354         (*stmsi6): Delete.
11355         (*stmsi5): Delete.
11356         (*stmsi4): Delete.
11357         (*stmsi3): Delete.
11358         (movmemsi_8reg): Delete.
11359         (corresponding unnamed define_insn): Delete.
11360         (movmemsi_6reg): Delete.
11361         (corresponding unnamed define_insn): Delete.
11362         (movmemsi_4reg): Delete.
11363         (corresponding unnamed define_insn): Delete.
11364         (movmemsi_2reg): Delete.
11365         (corresponding unnamed define_insn): Delete.
11366         (movmemsi_1reg): Delete.
11367         (corresponding unnamed define_insn): Delete.
11368         * config/rs6000/rs6000.opt (mno-string): New.
11369         (mstring): Replace by deprecation warning stub.
11370         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11371
11372 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
11373
11374         * regrename.c (regrename_do_replace): If replacing the same
11375         reg multiple times, try to reuse last created gen_raw_REG.
11376
11377         PR debug/81155
11378         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11379         main to workaround a bug in GDB.
11380
11381 2018-01-12  Tom de Vries  <tom@codesourcery.com>
11382
11383         PR target/83737
11384         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11385
11386 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
11387
11388         PR rtl-optimization/80481
11389         * ira-color.c (get_cap_member): New function.
11390         (allocnos_conflict_by_live_ranges_p): Use it.
11391         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11392         (setup_slot_coalesced_allocno_live_ranges): Ditto.
11393
11394 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
11395
11396         PR target/83628
11397         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11398         (*saddl_se_1): Ditto.
11399         (*ssubsi_1): Ditto.
11400         (*ssubl_se_1): Ditto.
11401
11402 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
11403
11404         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11405         rather than wi::to_widest for DR_INITs.
11406         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11407         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11408         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11409         INTEGER_CSTs.
11410         (vect_analyze_group_access_1): Note that here.
11411
11412 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
11413
11414         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11415         polynomial type sizes.
11416
11417 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
11418
11419         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11420         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11421         (gimple_add_tmp_var): Likewise.
11422
11423 2018-01-12  Martin Liska  <mliska@suse.cz>
11424
11425         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11426         (gimple_alloc_sizes): Likewise.
11427         (dump_gimple_statistics): Use PRIu64 in printf format.
11428         * gimple.h: Change uint64_t to int.
11429
11430 2018-01-12  Martin Liska  <mliska@suse.cz>
11431
11432         * tree-core.h: Use uint64_t instead of int.
11433         * tree.c (tree_node_counts): Likewise.
11434         (tree_node_sizes): Likewise.
11435         (dump_tree_statistics): Use PRIu64 in printf format.
11436
11437 2018-01-12  Martin Liska  <mliska@suse.cz>
11438
11439         * Makefile.in: As qsort_chk is implemented in vec.c, add
11440         vec.o to linkage of gencfn-macros.
11441         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11442         passing the info to record_node_allocation_statistics.
11443         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11444         and pass the info.
11445         * ggc-common.c (struct ggc_usage): Add operator== and use
11446         it in operator< and compare function.
11447         * mem-stats.h (struct mem_usage): Likewise.
11448         * vec.c (struct vec_usage): Remove operator< and compare
11449         function. Can be simply inherited.
11450
11451 2018-01-12  Martin Jambor  <mjambor@suse.cz>
11452
11453         PR target/81616
11454         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11455         * tree-ssa-math-opts.c: Include domwalk.h.
11456         (convert_mult_to_fma_1): New function.
11457         (fma_transformation_info): New type.
11458         (fma_deferring_state): Likewise.
11459         (cancel_fma_deferring): New function.
11460         (result_of_phi): Likewise.
11461         (last_fma_candidate_feeds_initial_phi): Likewise.
11462         (convert_mult_to_fma): Added deferring logic, split actual
11463         transformation to convert_mult_to_fma_1.
11464         (math_opts_dom_walker): New type.
11465         (math_opts_dom_walker::after_dom_children): New method, body moved
11466         here from pass_optimize_widening_mul::execute, added deferring logic
11467         bits.
11468         (pass_optimize_widening_mul::execute): Moved most of code to
11469         math_opts_dom_walker::after_dom_children.
11470         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11471         * config/i386/i386.c (ix86_option_override_internal): Added
11472         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11473
11474 2018-01-12  Richard Biener  <rguenther@suse.de>
11475
11476         PR debug/83157
11477         * dwarf2out.c (gen_variable_die): Do not reset old_die for
11478         inline instance vars.
11479
11480 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
11481
11482         PR target/81819
11483         * config/rx/rx.c (rx_is_restricted_memory_address):
11484         Handle SUBREG case.
11485
11486 2018-01-12  Richard Biener  <rguenther@suse.de>
11487
11488         PR tree-optimization/80846
11489         * target.def (split_reduction): New target hook.
11490         * targhooks.c (default_split_reduction): New function.
11491         * targhooks.h (default_split_reduction): Declare.
11492         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11493         target requests first reduce vectors by combining low and high
11494         parts.
11495         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11496         (get_vectype_for_scalar_type_and_size): Export.
11497         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11498         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11499         * doc/tm.texi: Regenerate.
11500         * config/i386/i386.c (ix86_split_reduction): Implement
11501         TARGET_VECTORIZE_SPLIT_REDUCTION.
11502
11503 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
11504
11505         PR target/83368
11506         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11507         in PIC mode except for TARGET_VXWORKS_RTP.
11508         * config/sparc/sparc.c: Include cfgrtl.h.
11509         (TARGET_INIT_PIC_REG): Define.
11510         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11511         (sparc_pic_register_p): New predicate.
11512         (sparc_legitimate_address_p): Use it.
11513         (sparc_legitimize_pic_address): Likewise.
11514         (sparc_delegitimize_address): Likewise.
11515         (sparc_mode_dependent_address_p): Likewise.
11516         (gen_load_pcrel_sym): Remove 4th parameter.
11517         (load_got_register): Adjust call to above.  Remove obsolete stuff.
11518         (sparc_expand_prologue): Do not call load_got_register here.
11519         (sparc_flat_expand_prologue): Likewise.
11520         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11521         (sparc_use_pseudo_pic_reg): New function.
11522         (sparc_init_pic_reg): Likewise.
11523         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11524         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11525
11526 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
11527
11528         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11529         Add item for branch_cost.
11530
11531 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
11532
11533         PR rtl-optimization/83565
11534         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11535         not extend the result to a larger mode for rotate operations.
11536         (num_sign_bit_copies1): Likewise.
11537
11538 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
11539
11540         PR target/40411
11541         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11542         -symbolic.
11543         Use values-Xc.o for -pedantic.
11544         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11545
11546 2018-01-12  Martin Liska  <mliska@suse.cz>
11547
11548         PR ipa/83054
11549         * ipa-devirt.c (final_warning_record::grow_type_warnings):
11550         New function.
11551         (possible_polymorphic_call_targets): Use it.
11552         (ipa_devirt): Likewise.
11553
11554 2018-01-12  Martin Liska  <mliska@suse.cz>
11555
11556         * profile-count.h (enum profile_quality): Use 0 as invalid
11557         enum value of profile_quality.
11558
11559 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
11560
11561         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11562         -mext-string options.
11563
11564 2018-01-12  Richard Biener  <rguenther@suse.de>
11565
11566         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11567         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11568         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11569         Likewise.
11570         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11571
11572 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
11573
11574         * configure.ac (--with-long-double-format): Add support for the
11575         configuration option to change the default long double format on
11576         PowerPC systems.
11577         * config.gcc (powerpc*-linux*-*): Likewise.
11578         * configure: Regenerate.
11579         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11580         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11581         used without modification.
11582
11583 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
11584
11585         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11586         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11587         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11588         MISC_BUILTIN_SPEC_BARRIER.
11589         (rs6000_init_builtins): Likewise.
11590         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11591         enum value.
11592         (speculation_barrier): New define_insn.
11593         * doc/extend.texi: Document __builtin_speculation_barrier.
11594
11595 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
11596
11597         PR target/83203
11598         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11599         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11600         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11601         iterators.
11602         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
11603         integral modes instead of "ss" and "sd".
11604         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11605         vectors with 32-bit and 64-bit elements.
11606         (vecdupssescalarmodesuffix): New mode attribute.
11607         (vec_dup<mode>): Use it.
11608
11609 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
11610
11611         PR target/83330
11612         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11613         frame if argument is passed on stack.
11614
11615 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
11616
11617         PR target/82682
11618         * ree.c (combine_reaching_defs): Optimize also
11619         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11620         reg2=any_extend(exp); reg1=reg2;, formatting fix.
11621
11622 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
11623
11624         PR middle-end/83189
11625         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11626
11627 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
11628
11629         PR middle-end/83718
11630         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11631         after they are computed.
11632
11633 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
11634
11635         PR tree-optimization/83695
11636         * gimple-loop-linterchange.cc
11637         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11638         reset cached scev information after interchange.
11639         (pass_linterchange::execute): Remove call to scev_reset_htab.
11640
11641 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11642
11643         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11644         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11645         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11646         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11647         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11648         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11649         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11650         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11651         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11652         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11653         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11654         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11655         (V_lane_reg): Likewise.
11656         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11657         New define_expand.
11658         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11659         (vfmal_lane_low<mode>_intrinsic,
11660         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11661         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11662         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11663         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11664         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11665         vfmsl_lane_high<mode>_intrinsic): New define_insns.
11666
11667 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11668
11669         * config/arm/arm-cpus.in (fp16fml): New feature.
11670         (ALL_SIMD): Add fp16fml.
11671         (armv8.2-a): Add fp16fml as an option.
11672         (armv8.3-a): Likewise.
11673         (armv8.4-a): Add fp16fml as part of fp16.
11674         * config/arm/arm.h (TARGET_FP16FML): Define.
11675         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11676         when appropriate.
11677         * config/arm/arm-modes.def (V2HF): Define.
11678         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11679         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11680         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11681         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11682         vfmsl_low, vfmsl_high): New set of builtins.
11683         * config/arm/iterators.md (PLUSMINUS): New code iterator.
11684         (vfml_op): New code attribute.
11685         (VFMLHALVES): New int iterator.
11686         (VFML, VFMLSEL): New mode attributes.
11687         (V_reg): Define mapping for V2HF.
11688         (V_hi, V_lo): New mode attributes.
11689         (VF_constraint): Likewise.
11690         (vfml_half, vfml_half_selector): New int attributes.
11691         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11692         define_expand.
11693         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11694         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11695         New define_insn.
11696         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11697         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11698         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11699         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
11700         documentation.
11701         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11702         Document new effective target and option set.
11703
11704 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11705
11706         * config/arm/arm-cpus.in (armv8_4): New feature.
11707         (ARMv8_4a): New fgroup.
11708         (armv8.4-a): New arch.
11709         * config/arm/arm-tables.opt: Regenerate.
11710         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11711         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11712         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11713         Add matching rules for -march=armv8.4-a and extensions.
11714         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11715
11716 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
11717
11718         PR target/81821
11719         * config/rx/rx.md (BW): New mode attribute.
11720         (sync_lock_test_and_setsi): Add mode suffix to insn output.
11721
11722 2018-01-11  Richard Biener  <rguenther@suse.de>
11723
11724         PR tree-optimization/83435
11725         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11726         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11727         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11728
11729 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
11730             Alan Hayward  <alan.hayward@arm.com>
11731             David Sherwood  <david.sherwood@arm.com>
11732
11733         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11734         field.
11735         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
11736         (aarch64_print_address_internal): Use it to check for a zero offset.
11737
11738 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
11739             Alan Hayward  <alan.hayward@arm.com>
11740             David Sherwood  <david.sherwood@arm.com>
11741
11742         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11743         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11744         Return a poly_int64 rather than a HOST_WIDE_INT.
11745         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11746         rather than a HOST_WIDE_INT.
11747         * config/aarch64/aarch64.h (aarch64_frame): Protect with
11748         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
11749         hard_fp_offset, frame_size, initial_adjust, callee_offset and
11750         final_offset from HOST_WIDE_INT to poly_int64.
11751         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11752         to_constant when getting the number of units in an Advanced SIMD
11753         mode.
11754         (aarch64_builtin_vectorized_function): Check for a constant number
11755         of units.
11756         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11757         GET_MODE_SIZE.
11758         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11759         attribute instead of GET_MODE_NUNITS.
11760         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11761         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11762         GET_MODE_SIZE for fixed-size registers.
11763         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11764         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11765         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11766         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11767         (aarch64_print_operand, aarch64_print_address_internal)
11768         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11769         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11770         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11771         Handle polynomial GET_MODE_SIZE.
11772         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
11773         wider than SImode without modification.
11774         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11775         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11776         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11777         passing and returning SVE modes.
11778         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11779         rather than GEN_INT.
11780         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11781         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11782         (aarch64_allocate_and_probe_stack_space): Likewise.
11783         (aarch64_layout_frame): Cope with polynomial offsets.
11784         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11785         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
11786         polynomial offsets.
11787         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11788         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11789         poly_int64 rather than a HOST_WIDE_INT.
11790         (aarch64_get_separate_components, aarch64_process_components)
11791         (aarch64_expand_prologue, aarch64_expand_epilogue)
11792         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11793         (aarch64_anchor_offset): New function, split out from...
11794         (aarch64_legitimize_address): ...here.
11795         (aarch64_builtin_vectorization_cost): Handle polynomial
11796         TYPE_VECTOR_SUBPARTS.
11797         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11798         GET_MODE_NUNITS.
11799         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11800         number of elements from the PARALLEL rather than the mode.
11801         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11802         rather than GET_MODE_BITSIZE.
11803         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11804         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11805         (aarch64_expand_vec_perm_const_1): Handle polynomial
11806         d->perm.length () and d->perm elements.
11807         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
11808         Apply to_constant to d->perm elements.
11809         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11810         polynomial CONST_VECTOR_NUNITS.
11811         (aarch64_move_pointer): Take amount as a poly_int64 rather
11812         than an int.
11813         (aarch64_progress_pointer): Avoid temporary variable.
11814         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11815         the mode attribute instead of GET_MODE.
11816
11817 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
11818             Alan Hayward  <alan.hayward@arm.com>
11819             David Sherwood  <david.sherwood@arm.com>
11820
11821         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11822         x exists before using it.
11823         (aarch64_add_constant_internal): Rename to...
11824         (aarch64_add_offset_1): ...this.  Replace regnum with separate
11825         src and dest rtxes.  Handle the case in which they're different,
11826         including when the offset is zero.  Replace scratchreg with an rtx.
11827         Use 2 additions if there is no spare register into which we can
11828         move a 16-bit constant.
11829         (aarch64_add_constant): Delete.
11830         (aarch64_add_offset): Replace reg with separate src and dest
11831         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
11832         Use aarch64_add_offset_1.
11833         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11834         an rtx rather than an int.  Take the delta as a poly_int64
11835         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
11836         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11837         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11838         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11839         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11840         and aarch64_add_sp.
11841         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11842         aarch64_add_constant.
11843
11844 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
11845
11846         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11847         Use scalar_float_mode.
11848
11849 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
11850
11851         * config/aarch64/aarch64-simd.md
11852         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11853         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11854         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11855         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11856         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11857         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11858         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11859         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11860         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11861         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11862
11863 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
11864
11865         PR target/83514
11866         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11867         targ_options->x_arm_arch_string is non NULL.
11868
11869 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
11870
11871         * config/aarch64/aarch64.h
11872         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
11873
11874 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
11875
11876         PR target/82096
11877         * expmed.c (emit_store_flag_force): Swap if const op0
11878         and change VOIDmode to mode of op0.
11879
11880 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
11881
11882         PR rtl-optimization/83761
11883         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11884         than bytes to mode_for_size.
11885
11886 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
11887
11888         PR middle-end/83189
11889         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11890         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11891         profile.
11892
11893 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
11894
11895         PR middle-end/83575
11896         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11897         when in layout mode.
11898         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11899         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11900         partition fixup.
11901
11902 2018-01-10  Michael Collison  <michael.collison@arm.com>
11903
11904         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11905         * config/aarch64/aarch64-option-extension.def: Add
11906         AARCH64_OPT_EXTENSION of 'fp16fml'.
11907         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11908         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11909         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11910         * config/aarch64/constraints.md (Ui7): New constraint.
11911         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11912         (VFMLA_SEL_W): Ditto.
11913         (f16quad): Ditto.
11914         (f16mac1): Ditto.
11915         (VFMLA16_LOW): New int iterator.
11916         (VFMLA16_HIGH): Ditto.
11917         (UNSPEC_FMLAL): New unspec.
11918         (UNSPEC_FMLSL): Ditto.
11919         (UNSPEC_FMLAL2): Ditto.
11920         (UNSPEC_FMLSL2): Ditto.
11921         (f16mac): New code attribute.
11922         * config/aarch64/aarch64-simd-builtins.def
11923         (aarch64_fmlal_lowv2sf): Ditto.
11924         (aarch64_fmlsl_lowv2sf): Ditto.
11925         (aarch64_fmlalq_lowv4sf): Ditto.
11926         (aarch64_fmlslq_lowv4sf): Ditto.
11927         (aarch64_fmlal_highv2sf): Ditto.
11928         (aarch64_fmlsl_highv2sf): Ditto.
11929         (aarch64_fmlalq_highv4sf): Ditto.
11930         (aarch64_fmlslq_highv4sf): Ditto.
11931         (aarch64_fmlal_lane_lowv2sf): Ditto.
11932         (aarch64_fmlsl_lane_lowv2sf): Ditto.
11933         (aarch64_fmlal_laneq_lowv2sf): Ditto.
11934         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11935         (aarch64_fmlalq_lane_lowv4sf): Ditto.
11936         (aarch64_fmlsl_lane_lowv4sf): Ditto.
11937         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11938         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11939         (aarch64_fmlal_lane_highv2sf): Ditto.
11940         (aarch64_fmlsl_lane_highv2sf): Ditto.
11941         (aarch64_fmlal_laneq_highv2sf): Ditto.
11942         (aarch64_fmlsl_laneq_highv2sf): Ditto.
11943         (aarch64_fmlalq_lane_highv4sf): Ditto.
11944         (aarch64_fmlsl_lane_highv4sf): Ditto.
11945         (aarch64_fmlalq_laneq_highv4sf): Ditto.
11946         (aarch64_fmlsl_laneq_highv4sf): Ditto.
11947         * config/aarch64/aarch64-simd.md:
11948         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11949         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11950         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11951         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11952         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11953         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11954         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11955         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11956         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11957         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11958         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11959         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11960         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11961         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11962         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11963         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11964         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11965         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11966         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11967         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11968         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11969         (vfmlsl_low_u32): Ditto.
11970         (vfmlalq_low_u32): Ditto.
11971         (vfmlslq_low_u32): Ditto.
11972         (vfmlal_high_u32): Ditto.
11973         (vfmlsl_high_u32): Ditto.
11974         (vfmlalq_high_u32): Ditto.
11975         (vfmlslq_high_u32): Ditto.
11976         (vfmlal_lane_low_u32): Ditto.
11977         (vfmlsl_lane_low_u32): Ditto.
11978         (vfmlal_laneq_low_u32): Ditto.
11979         (vfmlsl_laneq_low_u32): Ditto.
11980         (vfmlalq_lane_low_u32): Ditto.
11981         (vfmlslq_lane_low_u32): Ditto.
11982         (vfmlalq_laneq_low_u32): Ditto.
11983         (vfmlslq_laneq_low_u32): Ditto.
11984         (vfmlal_lane_high_u32): Ditto.
11985         (vfmlsl_lane_high_u32): Ditto.
11986         (vfmlal_laneq_high_u32): Ditto.
11987         (vfmlsl_laneq_high_u32): Ditto.
11988         (vfmlalq_lane_high_u32): Ditto.
11989         (vfmlslq_lane_high_u32): Ditto.
11990         (vfmlalq_laneq_high_u32): Ditto.
11991         (vfmlslq_laneq_high_u32): Ditto.
11992         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11993         (AARCH64_FL_FOR_ARCH8_4): New.
11994         (AARCH64_ISA_F16FML): New ISA flag.
11995         (TARGET_F16FML): New feature flag for fp16fml.
11996         (doc/invoke.texi): Document new fp16fml option.
11997
11998 2018-01-10  Michael Collison  <michael.collison@arm.com>
11999
12000         * config/aarch64/aarch64-builtins.c:
12001         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12002         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12003         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12004         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12005         (AARCH64_ISA_SHA3): New ISA flag.
12006         (TARGET_SHA3): New feature flag for sha3.
12007         * config/aarch64/iterators.md (sha512_op): New int attribute.
12008         (CRYPTO_SHA512): New int iterator.
12009         (UNSPEC_SHA512H): New unspec.
12010         (UNSPEC_SHA512H2): Ditto.
12011         (UNSPEC_SHA512SU0): Ditto.
12012         (UNSPEC_SHA512SU1): Ditto.
12013         * config/aarch64/aarch64-simd-builtins.def
12014         (aarch64_crypto_sha512hqv2di): New builtin.
12015         (aarch64_crypto_sha512h2qv2di): Ditto.
12016         (aarch64_crypto_sha512su0qv2di): Ditto.
12017         (aarch64_crypto_sha512su1qv2di): Ditto.
12018         (aarch64_eor3qv8hi): Ditto.
12019         (aarch64_rax1qv2di): Ditto.
12020         (aarch64_xarqv2di): Ditto.
12021         (aarch64_bcaxqv8hi): Ditto.
12022         * config/aarch64/aarch64-simd.md:
12023         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12024         (aarch64_crypto_sha512su0qv2di): Ditto.
12025         (aarch64_crypto_sha512su1qv2di): Ditto.
12026         (aarch64_eor3qv8hi): Ditto.
12027         (aarch64_rax1qv2di): Ditto.
12028         (aarch64_xarqv2di): Ditto.
12029         (aarch64_bcaxqv8hi): Ditto.
12030         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12031         (vsha512h2q_u64): Ditto.
12032         (vsha512su0q_u64): Ditto.
12033         (vsha512su1q_u64): Ditto.
12034         (veor3q_u16): Ditto.
12035         (vrax1q_u64): Ditto.
12036         (vxarq_u64): Ditto.
12037         (vbcaxq_u16): Ditto.
12038         * config/arm/types.md (crypto_sha512): New type attribute.
12039         (crypto_sha3): Ditto.
12040         (doc/invoke.texi): Document new sha3 option.
12041
12042 2018-01-10  Michael Collison  <michael.collison@arm.com>
12043
12044         * config/aarch64/aarch64-builtins.c:
12045         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12046         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12047         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12048         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12049         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12050         (AARCH64_ISA_SM4): New ISA flag.
12051         (TARGET_SM4): New feature flag for sm4.
12052         * config/aarch64/aarch64-simd-builtins.def
12053         (aarch64_sm3ss1qv4si): Ditto.
12054         (aarch64_sm3tt1aq4si): Ditto.
12055         (aarch64_sm3tt1bq4si): Ditto.
12056         (aarch64_sm3tt2aq4si): Ditto.
12057         (aarch64_sm3tt2bq4si): Ditto.
12058         (aarch64_sm3partw1qv4si): Ditto.
12059         (aarch64_sm3partw2qv4si): Ditto.
12060         (aarch64_sm4eqv4si): Ditto.
12061         (aarch64_sm4ekeyqv4si): Ditto.
12062         * config/aarch64/aarch64-simd.md:
12063         (aarch64_sm3ss1qv4si): Ditto.
12064         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12065         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12066         (aarch64_sm4eqv4si): Ditto.
12067         (aarch64_sm4ekeyqv4si): Ditto.
12068         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12069         (sm3part_op): Ditto.
12070         (CRYPTO_SM3TT): Ditto.
12071         (CRYPTO_SM3PART): Ditto.
12072         (UNSPEC_SM3SS1): New unspec.
12073         (UNSPEC_SM3TT1A): Ditto.
12074         (UNSPEC_SM3TT1B): Ditto.
12075         (UNSPEC_SM3TT2A): Ditto.
12076         (UNSPEC_SM3TT2B): Ditto.
12077         (UNSPEC_SM3PARTW1): Ditto.
12078         (UNSPEC_SM3PARTW2): Ditto.
12079         (UNSPEC_SM4E): Ditto.
12080         (UNSPEC_SM4EKEY): Ditto.
12081         * config/aarch64/constraints.md (Ui2): New constraint.
12082         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12083         * config/arm/types.md (crypto_sm3): New type attribute.
12084         (crypto_sm4): Ditto.
12085         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12086         (vsm3tt1aq_u32): Ditto.
12087         (vsm3tt1bq_u32): Ditto.
12088         (vsm3tt2aq_u32): Ditto.
12089         (vsm3tt2bq_u32): Ditto.
12090         (vsm3partw1q_u32): Ditto.
12091         (vsm3partw2q_u32): Ditto.
12092         (vsm4eq_u32): Ditto.
12093         (vsm4ekeyq_u32): Ditto.
12094         (doc/invoke.texi): Document new sm4 option.
12095
12096 2018-01-10  Michael Collison  <michael.collison@arm.com>
12097
12098         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12099         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12100         (AARCH64_FL_FOR_ARCH8_4): New.
12101         (AARCH64_FL_V8_4): New flag.
12102         (doc/invoke.texi): Document new armv8.4-a option.
12103
12104 2018-01-10  Michael Collison  <michael.collison@arm.com>
12105
12106         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12107         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12108         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12109         * config/aarch64/aarch64-option-extension.def: Add
12110         AARCH64_OPT_EXTENSION of 'sha2'.
12111         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12112         (crypto): Disable sha2 and aes if crypto disabled.
12113         (crypto): Enable aes and sha2 if enabled.
12114         (simd): Disable sha2 and aes if simd disabled.
12115         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12116         New flags.
12117         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12118         (TARGET_SHA2): New feature flag for sha2.
12119         (TARGET_AES): New feature flag for aes.
12120         * config/aarch64/aarch64-simd.md:
12121         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12122         conditional on TARGET_AES.
12123         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12124         (aarch64_crypto_sha1hsi): Make pattern conditional
12125         on TARGET_SHA2.
12126         (aarch64_crypto_sha1hv4si): Ditto.
12127         (aarch64_be_crypto_sha1hv4si): Ditto.
12128         (aarch64_crypto_sha1su1v4si): Ditto.
12129         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12130         (aarch64_crypto_sha1su0v4si): Ditto.
12131         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12132         (aarch64_crypto_sha256su0v4si): Ditto.
12133         (aarch64_crypto_sha256su1v4si): Ditto.
12134         (doc/invoke.texi): Document new aes and sha2 options.
12135
12136 2018-01-10  Martin Sebor  <msebor@redhat.com>
12137
12138         PR tree-optimization/83781
12139         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12140         as string arrays.
12141
12142 2018-01-11  Martin Sebor  <msebor@gmail.com>
12143             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
12144
12145         PR tree-optimization/83501
12146         PR tree-optimization/81703
12147
12148         * tree-ssa-strlen.c (get_string_cst): Rename...
12149         (get_string_len): ...to this.  Handle global constants.
12150         (handle_char_store): Adjust.
12151
12152 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
12153             Jim Wilson  <jimw@sifive.com>
12154
12155         * config/riscv/riscv-protos.h (riscv_output_return): New.
12156         * config/riscv/riscv.c (struct machine_function): New naked_p field.
12157         (riscv_attribute_table, riscv_output_return),
12158         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12159         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12160         (riscv_compute_frame_info): Only compute frame->mask if not a naked
12161         function.
12162         (riscv_expand_prologue): Add early return for naked function.
12163         (riscv_expand_epilogue): Likewise.
12164         (riscv_function_ok_for_sibcall): Return false for naked function.
12165         (riscv_set_current_function): New.
12166         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12167         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12168         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12169         * doc/extend.texi (RISC-V Function Attributes): New.
12170
12171 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
12172
12173         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12174         check for 128-bit long double before checking TCmode.
12175         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12176         128-bit long doubles before checking TFmode or TCmode.
12177         (FLOAT128_IBM_P): Likewise.
12178
12179 2018-01-10  Martin Sebor  <msebor@redhat.com>
12180
12181         PR tree-optimization/83671
12182         * builtins.c (c_strlen): Unconditionally return zero for the empty
12183         string.
12184         Use -Warray-bounds for warnings.
12185         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12186         for non-constant array indices with COMPONENT_REF, arrays of
12187         arrays, and pointers to arrays.
12188         (gimple_fold_builtin_strlen): Determine and set length range for
12189         non-constant character arrays.
12190
12191 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
12192
12193         PR middle-end/81897
12194         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12195         empty blocks.
12196
12197 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
12198
12199         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12200
12201 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
12202
12203         PR target/83399
12204         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12205         VECTOR_MEM_ALTIVEC_OR_VSX_P.
12206         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12207         indexed_or_indirect_operand predicate.
12208         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12209         (*vsx_le_perm_load_v8hi): Likewise.
12210         (*vsx_le_perm_load_v16qi): Likewise.
12211         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12212         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12213         (*vsx_le_perm_store_v8hi): Likewise.
12214         (*vsx_le_perm_store_v16qi): Likewise.
12215         (eight unnamed splitters): Likewise.
12216
12217 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
12218
12219         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12220         * config/rs6000/emmintrin.h: Likewise.
12221         * config/rs6000/mmintrin.h: Likewise.
12222         * config/rs6000/xmmintrin.h: Likewise.
12223
12224 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
12225
12226         PR c++/43486
12227         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12228         "public_flag".
12229         * tree.c (tree_nop_conversion): Return true for location wrapper
12230         nodes.
12231         (maybe_wrap_with_location): New function.
12232         (selftest::check_strip_nops): New function.
12233         (selftest::test_location_wrappers): New function.
12234         (selftest::tree_c_tests): Call it.
12235         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12236         (maybe_wrap_with_location): New decl.
12237         (EXPR_LOCATION_WRAPPER_P): New macro.
12238         (location_wrapper_p): New inline function.
12239         (tree_strip_any_location_wrapper): New inline function.
12240
12241 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
12242
12243         PR target/83735
12244         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12245         stack_realign_offset for the largest alignment of stack slot
12246         actually used.
12247         (ix86_find_max_used_stack_alignment): New function.
12248         (ix86_finalize_stack_frame_flags): Use it.  Set
12249         max_used_stack_alignment if we don't realign stack.
12250         * config/i386/i386.h (machine_function): Add
12251         max_used_stack_alignment.
12252
12253 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
12254
12255         * config/arm/arm.opt (-mbranch-cost): New option.
12256         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12257         account.
12258
12259 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
12260
12261         PR target/83629
12262         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12263         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12264
12265 2018-01-10  Richard Biener  <rguenther@suse.de>
12266
12267         PR debug/83765
12268         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12269         early out so it also covers the case where we have a non-NULL
12270         origin.
12271
12272 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
12273
12274         PR tree-optimization/83753
12275         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12276         for non-strided grouped accesses if the number of elements is 1.
12277
12278 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
12279
12280         PR target/81616
12281         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12282         * i386.h (TARGET_USE_GATHER): Define.
12283         * x86-tune.def (X86_TUNE_USE_GATHER): New.
12284
12285 2018-01-10  Martin Liska  <mliska@suse.cz>
12286
12287         PR bootstrap/82831
12288         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12289         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12290         partitioning.
12291         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12292         CLEANUP_NO_PARTITIONING is not set.
12293
12294 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
12295
12296         * doc/rtl.texi: Remove documentation of (const ...) wrappers
12297         for vectors, as a partial revert of r254296.
12298         * rtl.h (const_vec_p): Delete.
12299         (const_vec_duplicate_p): Don't test for vector CONSTs.
12300         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12301         * expmed.c (make_tree): Likewise.
12302
12303         Revert:
12304         * common.md (E, F): Use CONSTANT_P instead of checking for
12305         CONST_VECTOR.
12306         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12307         checking for CONST_VECTOR.
12308
12309 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
12310
12311         PR middle-end/83575
12312         * predict.c (force_edge_cold): Handle in more sane way edges
12313         with no prediction.
12314
12315 2018-01-09  Carl Love  <cel@us.ibm.com>
12316
12317         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12318         V4SI, V4SF types.
12319         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12320         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12321         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12322         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
12323         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12324         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
12325         * config/rs6000/rs6000-protos.h: Add extern defition for
12326         rs6000_generate_float2_double_code.
12327         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12328         function.
12329         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12330         (float2_v2df): Add define_expand.
12331
12332 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
12333
12334         PR target/83628
12335         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12336         op_mode in the force_to_mode call.
12337
12338 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
12339
12340         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12341         instead of checking each element individually.
12342         (aarch64_evpc_uzp): Likewise.
12343         (aarch64_evpc_zip): Likewise.
12344         (aarch64_evpc_ext): Likewise.
12345         (aarch64_evpc_rev): Likewise.
12346         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12347         instead of checking each element individually.  Return true without
12348         generating rtl if
12349         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12350         whether all selected elements come from the same input, instead of
12351         checking each element individually.  Remove calls to gen_rtx_REG,
12352         start_sequence and end_sequence and instead assert that no rtl is
12353         generated.
12354
12355 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
12356
12357         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12358         order of HIGH and CONST checks.
12359
12360 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
12361
12362         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12363         if the destination isn't an SSA_NAME.
12364
12365 2018-01-09  Richard Biener  <rguenther@suse.de>
12366
12367         PR tree-optimization/83668
12368         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12369         move prologue...
12370         (canonicalize_loop_form): ... here, renamed from ...
12371         (canonicalize_loop_closed_ssa_form): ... this and amended to
12372         swap successor edges for loop exit blocks to make us use
12373         the RPO order we need for initial schedule generation.
12374
12375 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
12376
12377         PR tree-optimization/64811
12378         * match.pd: When optimizing comparisons with Inf, avoid
12379         introducing or losing exceptions from comparisons with NaN.
12380
12381 2018-01-09  Martin Liska  <mliska@suse.cz>
12382
12383         PR sanitizer/82517
12384         * asan.c (shadow_mem_size): Add gcc_assert.
12385
12386 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
12387
12388         Don't save registers in main().
12389
12390         PR target/83738
12391         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12392         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12393         * config/avr/avr.c (avr_set_current_function): Don't error if
12394         naked, OS_task or OS_main are specified at the same time.
12395         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12396         OS_main.
12397         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12398         attribute.
12399         * common/config/avr/avr-common.c (avr_option_optimization_table):
12400         Switch on -mmain-is-OS_task for optimizing compilations.
12401
12402 2018-01-09  Richard Biener  <rguenther@suse.de>
12403
12404         PR tree-optimization/83572
12405         * graphite.c: Include cfganal.h.
12406         (graphite_transform_loops): Connect infinite loops to exit
12407         and remove fake edges at the end.
12408
12409 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
12410
12411         * ipa-inline.c (edge_badness): Revert accidental checkin.
12412
12413 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
12414
12415         PR ipa/80763
12416         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12417         symbols; not inline clones.
12418
12419 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
12420
12421         PR target/83507
12422         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12423         hard registers.  Formatting fixes.
12424
12425         PR preprocessor/83722
12426         * gcc.c (try_generate_repro): Pass
12427         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12428         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12429         do_report_bug.
12430
12431 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
12432             Kito Cheng  <kito.cheng@gmail.com>
12433
12434         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12435         (riscv_leaf_function_p): Delete.
12436         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12437
12438 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
12439
12440         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12441         function.
12442         (do_ifelse): New function.
12443         (do_isel): New function.
12444         (do_sub3): New function.
12445         (do_add3): New function.
12446         (do_load_mask_compare): New function.
12447         (do_overlap_load_compare): New function.
12448         (expand_compare_loop): New function.
12449         (expand_block_compare): Call expand_compare_loop() when appropriate.
12450         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12451         option description.
12452         (-mblock-compare-inline-loop-limit): New option.
12453
12454 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
12455
12456         PR target/83677
12457         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12458         Reverse order of second and third operands in first alternative.
12459         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12460         of first and second elements in UNSPEC_VPERMR vector.
12461         (altivec_expand_vec_perm_le): Likewise.
12462
12463 2018-01-08  Jeff Law  <law@redhat.com>
12464
12465         PR rtl-optimizatin/81308
12466         * tree-switch-conversion.c (cfg_altered): New file scoped static.
12467         (process_switch): If group_case_labels makes a change, then set
12468         cfg_altered.
12469         (pass_convert_switch::execute): If a switch is converted, then
12470         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
12471
12472         PR rtl-optimization/81308
12473         * recog.c (split_all_insns): Conditionally cleanup the CFG after
12474         splitting insns.
12475
12476 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
12477
12478         PR target/83663 - Revert r255946
12479         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12480         generation for cases where splatting a value is not useful.
12481         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12482         across a vec_duplicate and a paradoxical subreg forming a vector
12483         mode to a vec_concat.
12484
12485 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12486
12487         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12488         -march=armv8.3-a variants.
12489         * config/arm/t-multilib: Likewise.
12490         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
12491
12492 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
12493
12494         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12495         to generate rtl.
12496         (cceq_ior_compare_complement): Give it a name so I can use it, and
12497         change boolean_or_operator predicate to boolean_operator so it can
12498         be used to generate a crand.
12499         (eqne): New code iterator.
12500         (bd/bd_neg): New code_attrs.
12501         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12502         a single define_insn.
12503         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12504         decrement (bdnzt/bdnzf/bdzt/bdzf).
12505         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12506         with the new names of the branch decrement patterns, and added the
12507         names of the branch decrement conditional patterns.
12508
12509 2018-01-08  Richard Biener  <rguenther@suse.de>
12510
12511         PR tree-optimization/83563
12512         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12513         cache.
12514
12515 2018-01-08  Richard Biener  <rguenther@suse.de>
12516
12517         PR middle-end/83713
12518         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12519
12520 2018-01-08  Richard Biener  <rguenther@suse.de>
12521
12522         PR tree-optimization/83685
12523         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12524         references to abnormals.
12525
12526 2018-01-08  Richard Biener  <rguenther@suse.de>
12527
12528         PR lto/83719
12529         * dwarf2out.c (output_indirect_strings): Handle empty
12530         skeleton_debug_str_hash.
12531         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12532
12533 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
12534
12535         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12536         (emit_store_direct): Likewise.
12537         (arc_trampoline_adjust_address): Likewise.
12538         (arc_asm_trampoline_template): New function.
12539         (arc_initialize_trampoline): Use asm_trampoline_template.
12540         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12541         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12542         * config/arc/arc.md (flush_icache): Delete pattern.
12543
12544 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
12545
12546         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12547         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12548         munaligned-access.
12549
12550 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12551
12552         PR target/83681
12553         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12554         by not USED_FOR_TARGET.
12555         (make_pass_resolve_sw_modes): Likewise.
12556
12557 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12558
12559         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12560         USED_FOR_TARGET.
12561
12562 2018-01-08  Richard Biener  <rguenther@suse.de>
12563
12564         PR middle-end/83580
12565         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12566
12567 2018-01-08  Richard Biener  <rguenther@suse.de>
12568
12569         PR middle-end/83517
12570         * match.pd ((t * 2) / 2) -> t): Add missing :c.
12571
12572 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
12573
12574         PR middle-end/81897
12575         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12576         basic blocks with a small number of successors.
12577         (convert_control_dep_chain_into_preds): Improve handling of
12578         forwarder blocks.
12579         (dump_predicates): Split apart into...
12580         (dump_pred_chain): ...here...
12581         (dump_pred_info): ...and here.
12582         (can_one_predicate_be_invalidated_p): Add debugging printfs.
12583         (can_chain_union_be_invalidated_p): Improve check for invalidation
12584         of paths.
12585         (uninit_uses_cannot_happen): Avoid unnecessary if
12586         convert_control_dep_chain_into_preds yielded nothing.
12587
12588 2018-01-06  Martin Sebor  <msebor@redhat.com>
12589
12590         PR tree-optimization/83640
12591         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12592         subtracting negative offset from size.
12593         (builtin_access::overlap): Adjust offset bounds of the access to fall
12594         within the size of the object if possible.
12595
12596 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
12597
12598         PR rtl-optimization/83699
12599         * expmed.c (extract_bit_field_1): Restrict the vector usage of
12600         extract_bit_field_as_subreg to cases in which the extracted
12601         value is also a vector.
12602
12603         * lra-constraints.c (process_alt_operands): Test for the equivalence
12604         substitutions when detecting a possible reload cycle.
12605
12606 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
12607
12608         PR debug/83480
12609         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12610         by default if flag_selective_schedling{,2}.  Formatting fixes.
12611
12612         PR rtl-optimization/83682
12613         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12614         if it has non-VECTOR_MODE element mode.
12615         (vec_duplicate_p): Likewise.
12616
12617         PR middle-end/83694
12618         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12619         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12620
12621 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
12622
12623         PR target/83604
12624         * config/i386/i386-builtin.def
12625         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12626         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12627         Require also OPTION_MASK_ISA_AVX512F in addition to
12628         OPTION_MASK_ISA_GFNI.
12629         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12630         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12631         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12632         to OPTION_MASK_ISA_GFNI.
12633         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12634         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12635         OPTION_MASK_ISA_AVX512BW.
12636         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12637         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12638         addition to OPTION_MASK_ISA_GFNI.
12639         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12640         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12641         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12642         to OPTION_MASK_ISA_GFNI.
12643         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12644         a requirement for all ISAs rather than any of them with a few
12645         exceptions.
12646         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12647         processing.
12648         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12649         bitmasks to be enabled with 3 exceptions, instead of requiring any
12650         enabled ISA with lots of exceptions.
12651         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12652         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12653         Change avx512bw in isa attribute to avx512f.
12654         * config/i386/sgxintrin.h: Add license boilerplate.
12655         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
12656         to __AVX512F__ and __AVX512VL to __AVX512VL__.
12657         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12658         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12659         defined.
12660         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12661         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12662         temporarily sse2 rather than sse if not enabled already.
12663
12664         PR target/83604
12665         * config/i386/sse.md (VI248_VLBW): Rename to ...
12666         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
12667         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12668         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12669         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12670         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12671         mode iterator instead of VI248_VLBW.
12672
12673 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
12674
12675         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12676         (record_modified): Skip clobbers; add debug output.
12677         (param_change_prob): Use sreal frequencies.
12678
12679 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
12680
12681         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12682         punt for user-aligned variables.
12683
12684 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
12685
12686         * tree-chrec.c (chrec_contains_symbols): Return true for
12687         POLY_INT_CST.
12688
12689 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
12690
12691         PR target/82439
12692         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12693         of (x|y) == x for BICS pattern.
12694
12695 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
12696
12697         PR tree-optimization/83605
12698         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12699         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12700         can throw.
12701
12702 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12703
12704         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12705         * config/epiphany/rtems.h: New file.
12706
12707 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
12708             Uros Bizjak  <ubizjak@gmail.com>
12709
12710         PR target/83554
12711         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12712         QIreg_operand instead of register_operand predicate.
12713         * config/i386/i386.c (ix86_rop_should_change_byte_p,
12714         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12715         comments instead of -fmitigate[-_]rop.
12716
12717 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12718
12719         PR bootstrap/81926
12720         * cgraphunit.c (symbol_table::compile): Switch to text_section
12721         before calling assembly_start debug hook.
12722         * run-rtl-passes.c (run_rtl_passes): Likewise.
12723         Include output.h.
12724
12725 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
12726
12727         * tree-vrp.c (extract_range_from_binary_expr_1): Check
12728         range_int_cst_p rather than !symbolic_range_p before calling
12729         extract_range_from_multiplicative_op_1.
12730
12731 2018-01-04  Jeff Law  <law@redhat.com>
12732
12733         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12734         redundant test in assertion.
12735
12736 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
12737
12738         * doc/rtl.texi: Document machine_mode wrapper classes.
12739
12740 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
12741
12742         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12743         using tree_to_uhwi.
12744
12745 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
12746
12747         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12748         the VEC_PERM_EXPR fold to fail.
12749
12750 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
12751
12752         PR debug/83585
12753         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12754         to switched_sections.
12755
12756 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
12757
12758         PR target/83680
12759         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12760         test for d.testing.
12761
12762 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
12763
12764         PR target/83387
12765         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12766         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12767
12768 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
12769
12770         PR debug/83666
12771         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12772         is BLKmode and bitpos not zero or mode change is needed.
12773
12774 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
12775
12776         PR target/83675
12777         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12778         TARGET_VIS2.
12779
12780 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
12781
12782         PR target/83628
12783         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12784         instead of MULT rtx.  Update all corresponding splitters.
12785         (*saddl_se): Ditto.
12786         (*ssub<modesuffix>): Ditto.
12787         (*ssubl_se): Ditto.
12788         (*cmp_sadd_di): Update split patterns.
12789         (*cmp_sadd_si): Ditto.
12790         (*cmp_sadd_sidi): Ditto.
12791         (*cmp_ssub_di): Ditto.
12792         (*cmp_ssub_si): Ditto.
12793         (*cmp_ssub_sidi): Ditto.
12794         * config/alpha/predicates.md (const23_operand): New predicate.
12795         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12796         Look for ASHIFT, not MULT inner operand.
12797         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12798
12799 2018-01-04  Martin Liska  <mliska@suse.cz>
12800
12801         PR gcov-profile/83669
12802         * gcov.c (output_intermediate_file): Add version to intermediate
12803         gcov file.
12804         * doc/gcov.texi: Document new field 'version' in intermediate
12805         file format. Fix location of '-k' option of gcov command.
12806
12807 2018-01-04  Martin Liska  <mliska@suse.cz>
12808
12809         PR ipa/82352
12810         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12811
12812 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
12813
12814         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12815
12816 2018-01-03  Martin Sebor  <msebor@redhat.com>
12817
12818         PR tree-optimization/83655
12819         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12820         checking calls with invalid arguments.
12821
12822 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12823
12824         * tree-vect-stmts.c (vect_get_store_rhs): New function.
12825         (vectorizable_mask_load_store): Delete.
12826         (vectorizable_call): Return false for masked loads and stores.
12827         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
12828         instead of gimple_assign_rhs1.
12829         (vectorizable_load): Handle IFN_MASK_LOAD.
12830         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12831
12832 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12833
12834         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12835         split out from..,
12836         (vectorizable_mask_load_store): ...here.
12837         (vectorizable_load): ...and here.
12838
12839 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12840
12841         * tree-vect-stmts.c (vect_build_all_ones_mask)
12842         (vect_build_zero_merge_argument): New functions, split out from...
12843         (vectorizable_load): ...here.
12844
12845 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12846
12847         * tree-vect-stmts.c (vect_check_store_rhs): New function,
12848         split out from...
12849         (vectorizable_mask_load_store): ...here.
12850         (vectorizable_store): ...and here.
12851
12852 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12853
12854         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12855         split out from...
12856         (vectorizable_mask_load_store): ...here.
12857
12858 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12859
12860         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12861         (vect_model_store_cost): Take a vec_load_store_type instead of a
12862         vect_def_type.
12863         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12864         (vect_model_store_cost): Take a vec_load_store_type instead of a
12865         vect_def_type.
12866         (vectorizable_mask_load_store): Update accordingly.
12867         (vectorizable_store): Likewise.
12868         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12869
12870 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12871
12872         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12873         IFN_MASK_LOAD calls here rather than...
12874         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12875
12876 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12877             Alan Hayward  <alan.hayward@arm.com>
12878             David Sherwood  <david.sherwood@arm.com>
12879
12880         * expmed.c (extract_bit_field_1): For vector extracts,
12881         fall back to extract_bit_field_as_subreg if vec_extract
12882         isn't available.
12883
12884 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12885             Alan Hayward  <alan.hayward@arm.com>
12886             David Sherwood  <david.sherwood@arm.com>
12887
12888         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12889         they are variable or constant sized.
12890         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12891         slots for constant-sized data.
12892
12893 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12894             Alan Hayward  <alan.hayward@arm.com>
12895             David Sherwood  <david.sherwood@arm.com>
12896
12897         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12898         handling COND_EXPRs with boolean comparisons, try to find a better
12899         basis for the mask type than the boolean itself.
12900
12901 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12902
12903         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12904         is calculated and how it can be overridden.
12905         * genmodes.c (max_bitsize_mode_any_mode): New variable.
12906         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12907         if defined.
12908         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12909         if nonzero.
12910
12911 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12912             Alan Hayward  <alan.hayward@arm.com>
12913             David Sherwood  <david.sherwood@arm.com>
12914
12915         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12916         Remove the mode argument.
12917         (aarch64_simd_valid_immediate): Remove the mode and inverse
12918         arguments.
12919         * config/aarch64/iterators.md (bitsize): New iterator.
12920         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12921         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12922         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12923         aarch64_simd_valid_immediate.
12924         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12925         (aarch64_reg_or_bic_imm): Likewise.
12926         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12927         with an insn_type enum and msl with a modifier_type enum.
12928         Replace element_width with a scalar_mode.  Change the shift
12929         to unsigned int.  Add constructors for scalar_float_mode and
12930         scalar_int_mode elements.
12931         (aarch64_vect_float_const_representable_p): Delete.
12932         (aarch64_can_const_movi_rtx_p)
12933         (aarch64_simd_scalar_immediate_valid_for_move)
12934         (aarch64_simd_make_constant): Update call to
12935         aarch64_simd_valid_immediate.
12936         (aarch64_advsimd_valid_immediate_hs): New function.
12937         (aarch64_advsimd_valid_immediate): Likewise.
12938         (aarch64_simd_valid_immediate): Remove mode and inverse
12939         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
12940         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12941         and aarch64_float_const_representable_p on the result.
12942         (aarch64_output_simd_mov_immediate): Remove mode argument.
12943         Update call to aarch64_simd_valid_immediate and use of
12944         simd_immediate_info.
12945         (aarch64_output_scalar_simd_mov_immediate): Update call
12946         accordingly.
12947
12948 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12949             Alan Hayward  <alan.hayward@arm.com>
12950             David Sherwood  <david.sherwood@arm.com>
12951
12952         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12953         (mode_nunits): Likewise CONST_MODE_NUNITS.
12954         * machmode.def (ADJUST_NUNITS): Document.
12955         * genmodes.c (mode_data::need_nunits_adj): New field.
12956         (blank_mode): Update accordingly.
12957         (adj_nunits): New variable.
12958         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12959         parameter.
12960         (emit_mode_size_inline): Set need_bytesize_adj for all modes
12961         listed in adj_nunits.
12962         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12963         listed in adj_nunits.  Don't emit case statements for such modes.
12964         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12965         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
12966         nothing if adj_nunits is nonnull.
12967         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12968         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12969         (emit_mode_fbit): Update use of print_maybe_const_decl.
12970         (emit_move_size): Likewise.  Treat the array as non-const
12971         if adj_nunits.
12972         (emit_mode_adjustments): Handle adj_nunits.
12973
12974 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12975
12976         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12977         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12978         (VECTOR_MODES): Use it.
12979         (make_vector_modes): Take the prefix as an argument.
12980
12981 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
12982             Alan Hayward  <alan.hayward@arm.com>
12983             David Sherwood  <david.sherwood@arm.com>
12984
12985         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12986         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12987         for MODE_VECTOR_BOOL.
12988         * machmode.def (VECTOR_BOOL_MODE): Document.
12989         * genmodes.c (VECTOR_BOOL_MODE): New macro.
12990         (make_vector_bool_mode): New function.
12991         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12992         MODE_VECTOR_BOOL.
12993         * lto-streamer-in.c (lto_input_mode_table): Likewise.
12994         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12995         Likewise.
12996         * stor-layout.c (int_mode_for_mode): Likewise.
12997         * tree.c (build_vector_type_for_mode): Likewise.
12998         * varasm.c (output_constant_pool_2): Likewise.
12999         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13000         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
13001         for MODE_VECTOR_BOOL.
13002         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13003         of mode class checks.
13004         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13005         instead of a list of mode class checks.
13006         (expand_vector_scalar_condition): Likewise.
13007         (type_for_widest_vector_mode): Handle BImode as an inner mode.
13008
13009 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13010             Alan Hayward  <alan.hayward@arm.com>
13011             David Sherwood  <david.sherwood@arm.com>
13012
13013         * machmode.h (mode_size): Change from unsigned short to
13014         poly_uint16_pod.
13015         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13016         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13017         or if measurement_type is not polynomial.
13018         (fixed_size_mode::includes_p): Check for constant-sized modes.
13019         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13020         return a poly_uint16 rather than an unsigned short.
13021         (emit_mode_size): Change the type of mode_size from unsigned short
13022         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
13023         (emit_mode_adjustments): Cope with polynomial vector sizes.
13024         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13025         for GET_MODE_SIZE.
13026         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13027         for GET_MODE_SIZE.
13028         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13029         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13030         * caller-save.c (setup_save_areas): Likewise.
13031         (replace_reg_with_saved_mem): Likewise.
13032         * calls.c (emit_library_call_value_1): Likewise.
13033         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13034         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13035         (gen_lowpart_for_combine): Likewise.
13036         * convert.c (convert_to_integer_1): Likewise.
13037         * cse.c (equiv_constant, cse_insn): Likewise.
13038         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13039         (cselib_subst_to_values): Likewise.
13040         * dce.c (word_dce_process_block): Likewise.
13041         * df-problems.c (df_word_lr_mark_ref): Likewise.
13042         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13043         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13044         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13045         (rtl_for_decl_location): Likewise.
13046         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13047         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13048         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13049         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13050         (expand_expr_real_1): Likewise.
13051         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13052         (pad_below): Likewise.
13053         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13054         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13055         * ira.c (get_subreg_tracking_sizes): Likewise.
13056         * ira-build.c (ira_create_allocno_objects): Likewise.
13057         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13058         (ira_sort_regnos_for_alter_reg): Likewise.
13059         * ira-costs.c (record_operand_costs): Likewise.
13060         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13061         (resolve_simple_move): Likewise.
13062         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13063         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13064         (lra_constraints): Likewise.
13065         (CONST_POOL_OK_P): Reject variable-sized modes.
13066         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13067         (add_pseudo_to_slot, lra_spill): Likewise.
13068         * omp-low.c (omp_clause_aligned_alignment): Likewise.
13069         * optabs-query.c (get_best_extraction_insn): Likewise.
13070         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13071         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13072         (expand_mult_highpart, valid_multiword_target_p): Likewise.
13073         * recog.c (offsettable_address_addr_space_p): Likewise.
13074         * regcprop.c (maybe_mode_change): Likewise.
13075         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13076         * regrename.c (build_def_use): Likewise.
13077         * regstat.c (dump_reg_info): Likewise.
13078         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13079         (find_reloads, find_reloads_subreg_address): Likewise.
13080         * reload1.c (eliminate_regs_1): Likewise.
13081         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13082         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13083         (simplify_binary_operation_1, simplify_subreg): Likewise.
13084         * targhooks.c (default_function_arg_padding): Likewise.
13085         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13086         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13087         (verify_gimple_assign_ternary): Likewise.
13088         * tree-inline.c (estimate_move_cost): Likewise.
13089         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13090         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13091         (get_address_cost_ainc): Likewise.
13092         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13093         (vect_supportable_dr_alignment): Likewise.
13094         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13095         (vectorizable_reduction): Likewise.
13096         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13097         (vectorizable_operation, vectorizable_load): Likewise.
13098         * tree.c (build_same_sized_truth_vector_type): Likewise.
13099         * valtrack.c (cleanup_auto_inc_dec): Likewise.
13100         * var-tracking.c (emit_note_insn_var_location): Likewise.
13101         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13102         (ADDR_VEC_ALIGN): Likewise.
13103
13104 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13105             Alan Hayward  <alan.hayward@arm.com>
13106             David Sherwood  <david.sherwood@arm.com>
13107
13108         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13109         unsigned short.
13110         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13111         or if measurement_type is polynomial.
13112         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13113         * combine.c (make_extraction): Likewise.
13114         * dse.c (find_shift_sequence): Likewise.
13115         * dwarf2out.c (mem_loc_descriptor): Likewise.
13116         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13117         (extract_bit_field, extract_low_bits): Likewise.
13118         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13119         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13120         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13121         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13122         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13123         * reload.c (find_reloads): Likewise.
13124         * reload1.c (alter_reg): Likewise.
13125         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13126         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13127         * tree-if-conv.c (predicate_mem_writes): Likewise.
13128         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13129         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13130         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13131         * valtrack.c (dead_debug_insert_temp): Likewise.
13132         * varasm.c (mergeable_constant_section): Likewise.
13133         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13134
13135 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13136             Alan Hayward  <alan.hayward@arm.com>
13137             David Sherwood  <david.sherwood@arm.com>
13138
13139         * expr.c (expand_assignment): Cope with polynomial mode sizes
13140         when assigning to a CONCAT.
13141
13142 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13143             Alan Hayward  <alan.hayward@arm.com>
13144             David Sherwood  <david.sherwood@arm.com>
13145
13146         * machmode.h (mode_precision): Change from unsigned short to
13147         poly_uint16_pod.
13148         (mode_to_precision): Return a poly_uint16 rather than an unsigned
13149         short.
13150         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13151         or if measurement_type is not polynomial.
13152         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
13153         in which the mode is already known to be a scalar_int_mode.
13154         * genmodes.c (emit_mode_precision): Change the type of mode_precision
13155         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
13156         initializer.
13157         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13158         for GET_MODE_PRECISION.
13159         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13160         for GET_MODE_PRECISION.
13161         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13162         as polynomial.
13163         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13164         (expand_field_assignment, make_extraction): Likewise.
13165         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13166         (get_last_value): Likewise.
13167         * convert.c (convert_to_integer_1): Likewise.
13168         * cse.c (cse_insn): Likewise.
13169         * expr.c (expand_expr_real_1): Likewise.
13170         * lra-constraints.c (simplify_operand_subreg): Likewise.
13171         * optabs-query.c (can_atomic_load_p): Likewise.
13172         * optabs.c (expand_atomic_load): Likewise.
13173         (expand_atomic_store): Likewise.
13174         * ree.c (combine_reaching_defs): Likewise.
13175         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13176         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13177         * tree.h (type_has_mode_precision_p): Likewise.
13178         * ubsan.c (instrument_si_overflow): Likewise.
13179
13180 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13181             Alan Hayward  <alan.hayward@arm.com>
13182             David Sherwood  <david.sherwood@arm.com>
13183
13184         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13185         polynomial numbers of units.
13186         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13187         (valid_vector_subparts_p): New function.
13188         (build_vector_type): Remove temporary shim and take the number
13189         of units as a poly_uint64 rather than an int.
13190         (build_opaque_vector_type): Take the number of units as a
13191         poly_uint64 rather than an int.
13192         * tree.c (build_vector_from_ctor): Handle polynomial
13193         TYPE_VECTOR_SUBPARTS.
13194         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13195         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13196         (build_vector_from_val): If the number of units is variable,
13197         use build_vec_duplicate_cst for constant operands and
13198         VEC_DUPLICATE_EXPR otherwise.
13199         (make_vector_type): Remove temporary is_constant ().
13200         (build_vector_type, build_opaque_vector_type): Take the number of
13201         units as a poly_uint64 rather than an int.
13202         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13203         VECTOR_CST_NELTS.
13204         * cfgexpand.c (expand_debug_expr): Likewise.
13205         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13206         (store_constructor, expand_expr_real_1): Likewise.
13207         (const_scalar_mask_from_tree): Likewise.
13208         * fold-const-call.c (fold_const_reduction): Likewise.
13209         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13210         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13211         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13212         (fold_relational_const): Likewise.
13213         (native_interpret_vector): Likewise.  Change the size from an
13214         int to an unsigned int.
13215         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13216         TYPE_VECTOR_SUBPARTS.
13217         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13218         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13219         duplicating a non-constant operand into a variable-length vector.
13220         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13221         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13222         * ipa-icf.c (sem_variable::equals): Likewise.
13223         * match.pd: Likewise.
13224         * omp-simd-clone.c (simd_clone_subparts): Likewise.
13225         * print-tree.c (print_node): Likewise.
13226         * stor-layout.c (layout_type): Likewise.
13227         * targhooks.c (default_builtin_vectorization_cost): Likewise.
13228         * tree-cfg.c (verify_gimple_comparison): Likewise.
13229         (verify_gimple_assign_binary): Likewise.
13230         (verify_gimple_assign_ternary): Likewise.
13231         (verify_gimple_assign_single): Likewise.
13232         * tree-pretty-print.c (dump_generic_node): Likewise.
13233         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13234         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13235         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13236         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13237         (vect_shift_permute_load_chain): Likewise.
13238         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13239         (expand_vector_condition, optimize_vector_constructor): Likewise.
13240         (lower_vec_perm, get_compute_type): Likewise.
13241         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13242         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13243         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13244         (vect_recog_mask_conversion_pattern): Likewise.
13245         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13246         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13247         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13248         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13249         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13250         (vectorizable_shift, vectorizable_operation, vectorizable_store)
13251         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13252         (supportable_widening_operation): Likewise.
13253         (supportable_narrowing_operation): Likewise.
13254         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13255         Likewise.
13256         * varasm.c (output_constant): Likewise.
13257
13258 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13259             Alan Hayward  <alan.hayward@arm.com>
13260             David Sherwood  <david.sherwood@arm.com>
13261
13262         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13263         so that both the length == 3 and length != 3 cases set up their
13264         own permute vectors.  Add comments explaining why we know the
13265         number of elements is constant.
13266         (vect_permute_load_chain): Likewise.
13267
13268 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13269             Alan Hayward  <alan.hayward@arm.com>
13270             David Sherwood  <david.sherwood@arm.com>
13271
13272         * machmode.h (mode_nunits): Change from unsigned char to
13273         poly_uint16_pod.
13274         (ONLY_FIXED_SIZE_MODES): New macro.
13275         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13276         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13277         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13278         New typedefs.
13279         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13280         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13281         or if measurement_type is not polynomial.
13282         * genmodes.c (ZERO_COEFFS): New macro.
13283         (emit_mode_nunits_inline): Make mode_nunits_inline return a
13284         poly_uint16.
13285         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13286         Use ZERO_COEFFS when emitting initializers.
13287         * data-streamer.h (bp_pack_poly_value): New function.
13288         (bp_unpack_poly_value): Likewise.
13289         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13290         for GET_MODE_NUNITS.
13291         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13292         for GET_MODE_NUNITS.
13293         * tree.c (make_vector_type): Remove temporary shim and make
13294         the real function take the number of units as a poly_uint64
13295         rather than an int.
13296         (build_vector_type_for_mode): Handle polynomial nunits.
13297         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13298         * emit-rtl.c (const_vec_series_p_1): Likewise.
13299         (gen_rtx_CONST_VECTOR): Likewise.
13300         * fold-const.c (test_vec_duplicate_folding): Likewise.
13301         * genrecog.c (validate_pattern): Likewise.
13302         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13303         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13304         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13305         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13306         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13307         * rtlanal.c (subreg_get_info): Likewise.
13308         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13309         (vect_grouped_load_supported): Likewise.
13310         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13311         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13312         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13313         (simplify_const_unary_operation, simplify_binary_operation_1)
13314         (simplify_const_binary_operation, simplify_ternary_operation)
13315         (test_vector_ops_duplicate, test_vector_ops): Likewise.
13316         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13317         instead of CONST_VECTOR_NUNITS.
13318         * varasm.c (output_constant_pool_2): Likewise.
13319         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13320         explicit-encoded elements in the XVEC for variable-length vectors.
13321
13322 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13323
13324         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13325
13326 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13327             Alan Hayward  <alan.hayward@arm.com>
13328             David Sherwood  <david.sherwood@arm.com>
13329
13330         * coretypes.h (fixed_size_mode): Declare.
13331         (fixed_size_mode_pod): New typedef.
13332         * builtins.h (target_builtins::x_apply_args_mode)
13333         (target_builtins::x_apply_result_mode): Change type to
13334         fixed_size_mode_pod.
13335         * builtins.c (apply_args_size, apply_result_size, result_vector)
13336         (expand_builtin_apply_args_1, expand_builtin_apply)
13337         (expand_builtin_return): Update accordingly.
13338
13339 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13340
13341         * cse.c (hash_rtx_cb): Hash only the encoded elements.
13342         * cselib.c (cselib_hash_rtx): Likewise.
13343         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13344         CONST_VECTOR encoding.
13345
13346 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
13347             Jeff Law  <law@redhat.com>
13348
13349         PR target/83641
13350         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13351         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13352         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13353         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13354
13355         PR target/83641
13356         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13357         explicitly probe *sp in a noreturn function if there were any callee
13358         register saves or frame pointer is needed.
13359
13360 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
13361
13362         PR debug/83621
13363         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13364         BLKmode for ternary, binary or unary expressions.
13365
13366         PR debug/83645
13367         * var-tracking.c (delete_vta_debug_insn): New inline function.
13368         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13369         insns from get_insns () to NULL instead of each bb separately.
13370         Use delete_vta_debug_insn.  No longer static.
13371         (vt_debug_insns_local, variable_tracking_main_1): Adjust
13372         delete_vta_debug_insns callers.
13373         * rtl.h (delete_vta_debug_insns): Declare.
13374         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13375         instead of variable_tracking_main.
13376
13377 2018-01-03  Martin Sebor  <msebor@redhat.com>
13378
13379         PR tree-optimization/83603
13380         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13381         arguments past the endof the argument list in functions declared
13382         without a prototype.
13383         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13384         Avoid checking when arguments are null.
13385
13386 2018-01-03  Martin Sebor  <msebor@redhat.com>
13387
13388         PR c/83559
13389         * doc/extend.texi (attribute const): Fix a typo.
13390         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13391         issuing -Wsuggest-attribute for void functions.
13392
13393 2018-01-03  Martin Sebor  <msebor@redhat.com>
13394
13395         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13396         offset_int::from instead of wide_int::to_shwi.
13397         (maybe_diag_overlap): Remove assertion.
13398         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13399         * gimple-ssa-sprintf.c (format_directive): Same.
13400         (parse_directive): Same.
13401         (sprintf_dom_walker::compute_format_length): Same.
13402         (try_substitute_return_value): Same.
13403
13404 2018-01-03  Jeff Law  <law@redhat.com>
13405
13406         PR middle-end/83654
13407         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13408         non-constant residual for zero at runtime and avoid probing in
13409         that case.  Reorganize code for trailing problem to mirror handling
13410         of the residual.
13411
13412 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
13413
13414         PR tree-optimization/83501
13415         * tree-ssa-strlen.c (get_string_cst): New.
13416         (handle_char_store): Call get_string_cst.
13417
13418 2018-01-03  Martin Liska  <mliska@suse.cz>
13419
13420         PR tree-optimization/83593
13421         * tree-ssa-strlen.c: Include tree-cfg.h.
13422         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13423         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13424         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13425         to false.
13426         (strlen_dom_walker::before_dom_children): Call
13427         gimple_purge_dead_eh_edges. Dump tranformation with details
13428         dump flags.
13429         (strlen_dom_walker::before_dom_children): Update call by adding
13430         new argument cleanup_eh.
13431         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13432
13433 2018-01-03  Martin Liska  <mliska@suse.cz>
13434
13435         PR ipa/83549
13436         * cif-code.def (VARIADIC_THUNK): New enum value.
13437         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13438         thunks.
13439
13440 2018-01-03  Jan Beulich  <jbeulich@suse.com>
13441
13442         * sse.md (mov<mode>_internal): Tighten condition for when to use
13443         vmovdqu<ssescalarsize> for TI and OI modes.
13444
13445 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
13446
13447         Update copyright years.
13448
13449 2018-01-03  Martin Liska  <mliska@suse.cz>
13450
13451         PR ipa/83594
13452         * ipa-visibility.c (function_and_variable_visibility): Skip
13453         functions with noipa attribure.
13454
13455 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
13456
13457         * gcc.c (process_command): Update copyright notice dates.
13458         * gcov-dump.c (print_version): Ditto.
13459         * gcov.c (print_version): Ditto.
13460         * gcov-tool.c (print_version): Ditto.
13461         * gengtype.c (create_file): Ditto.
13462         * doc/cpp.texi: Bump @copying's copyright year.
13463         * doc/cppinternals.texi: Ditto.
13464         * doc/gcc.texi: Ditto.
13465         * doc/gccint.texi: Ditto.
13466         * doc/gcov.texi: Ditto.
13467         * doc/install.texi: Ditto.
13468         * doc/invoke.texi: Ditto.
13469
13470 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13471
13472         * vector-builder.h (vector_builder::m_full_nelts): Change from
13473         unsigned int to poly_uint64.
13474         (vector_builder::full_nelts): Update prototype accordingly.
13475         (vector_builder::new_vector): Likewise.
13476         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13477         (vector_builder::operator ==): Likewise.
13478         (vector_builder::finalize): Likewise.
13479         * int-vector-builder.h (int_vector_builder::int_vector_builder):
13480         Take the number of elements as a poly_uint64 rather than an
13481         unsigned int.
13482         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13483         from unsigned int to poly_uint64.
13484         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13485         (vec_perm_indices::new_vector): Likewise.
13486         (vec_perm_indices::length): Likewise.
13487         (vec_perm_indices::nelts_per_input): Likewise.
13488         (vec_perm_indices::input_nelts): Likewise.
13489         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13490         number of elements per input as a poly_uint64 rather than an
13491         unsigned int.  Use the original encoding for variable-length
13492         vectors, rather than clamping each individual element.
13493         For the second and subsequent elements in each pattern,
13494         clamp the step and base before clamping their sum.
13495         (vec_perm_indices::series_p): Handle polynomial element counts.
13496         (vec_perm_indices::all_in_range_p): Likewise.
13497         (vec_perm_indices_to_tree): Likewise.
13498         (vec_perm_indices_to_rtx): Likewise.
13499         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13500         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13501         (tree_vector_builder::new_binary_operation): Handle polynomial
13502         element counts.  Return false if we need to know the number
13503         of elements at compile time.
13504         * fold-const.c (fold_vec_perm): Punt if the number of elements
13505         isn't known at compile time.
13506
13507 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13508
13509         * vec-perm-indices.h (vec_perm_builder): Change element type
13510         from HOST_WIDE_INT to poly_int64.
13511         (vec_perm_indices::element_type): Update accordingly.
13512         (vec_perm_indices::clamp): Handle polynomial element_types.
13513         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13514         (vec_perm_indices::all_in_range_p): Likewise.
13515         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13516         than shwi trees.
13517         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13518         polynomial vec_perm_indices element types.
13519         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13520         * fold-const.c (fold_vec_perm): Likewise.
13521         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13522         * tree-vect-generic.c (lower_vec_perm): Likewise.
13523         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13524         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13525         element type to HOST_WIDE_INT.
13526
13527 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13528             Alan Hayward  <alan.hayward@arm.com>
13529             David Sherwood  <david.sherwood@arm.com>
13530
13531         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13532         rather than an int.  Use plus_constant.
13533         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13534         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13535
13536 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13537             Alan Hayward  <alan.hayward@arm.com>
13538             David Sherwood  <david.sherwood@arm.com>
13539
13540         * calls.c (emit_call_1, expand_call): Change struct_value_size from
13541         a HOST_WIDE_INT to a poly_int64.
13542
13543 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13544             Alan Hayward  <alan.hayward@arm.com>
13545             David Sherwood  <david.sherwood@arm.com>
13546
13547         * calls.c (load_register_parameters): Cope with polynomial
13548         mode sizes.  Require a constant size for BLKmode parameters
13549         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
13550         forces a parameter to be padded at the lsb end in order to
13551         fill a complete number of words, require the parameter size
13552         to be ordered wrt UNITS_PER_WORD.
13553
13554 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13555             Alan Hayward  <alan.hayward@arm.com>
13556             David Sherwood  <david.sherwood@arm.com>
13557
13558         * reload1.c (spill_stack_slot_width): Change element type
13559         from unsigned int to poly_uint64_pod.
13560         (alter_reg): Treat mode sizes as polynomial.
13561
13562 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13563             Alan Hayward  <alan.hayward@arm.com>
13564             David Sherwood  <david.sherwood@arm.com>
13565
13566         * reload.c (complex_word_subreg_p): New function.
13567         (reload_inner_reg_of_subreg, push_reload): Use it.
13568
13569 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13570             Alan Hayward  <alan.hayward@arm.com>
13571             David Sherwood  <david.sherwood@arm.com>
13572
13573         * lra-constraints.c (process_alt_operands): Reject matched
13574         operands whose sizes aren't ordered.
13575         (match_reload): Refer to this check here.
13576
13577 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13578             Alan Hayward  <alan.hayward@arm.com>
13579             David Sherwood  <david.sherwood@arm.com>
13580
13581         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13582         that the mode size is in the set {1, 2, 4, 8, 16}.
13583
13584 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13585             Alan Hayward  <alan.hayward@arm.com>
13586             David Sherwood  <david.sherwood@arm.com>
13587
13588         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13589         Use plus_constant instead of gen_rtx_PLUS.
13590
13591 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13592             Alan Hayward  <alan.hayward@arm.com>
13593             David Sherwood  <david.sherwood@arm.com>
13594
13595         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13596         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13597         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13598         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13599         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13600         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13601         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13602         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13603         * config/i386/i386.c (ix86_push_rounding): ...this new function.
13604         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13605         a poly_int64.
13606         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13607         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13608         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13609         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13610         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13611         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13612         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13613         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13614         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13615         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13616         function.
13617         * expr.c (emit_move_resolve_push): Treat the input and result
13618         of PUSH_ROUNDING as a poly_int64.
13619         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13620         (emit_push_insn): Likewise.
13621         * lra-eliminations.c (mark_not_eliminable): Likewise.
13622         * recog.c (push_operand): Likewise.
13623         * reload1.c (elimination_effects): Likewise.
13624         * rtlanal.c (nonzero_bits1): Likewise.
13625         * calls.c (store_one_arg): Likewise.  Require the padding to be
13626         known at compile time.
13627
13628 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13629             Alan Hayward  <alan.hayward@arm.com>
13630             David Sherwood  <david.sherwood@arm.com>
13631
13632         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13633         Use plus_constant instead of gen_rtx_PLUS.
13634
13635 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13636             Alan Hayward  <alan.hayward@arm.com>
13637             David Sherwood  <david.sherwood@arm.com>
13638
13639         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13640         rather than an int.
13641
13642 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13643             Alan Hayward  <alan.hayward@arm.com>
13644             David Sherwood  <david.sherwood@arm.com>
13645
13646         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13647         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13648         via stack temporaries.  Treat the mode size as polynomial too.
13649
13650 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13651             Alan Hayward  <alan.hayward@arm.com>
13652             David Sherwood  <david.sherwood@arm.com>
13653
13654         * expr.c (expand_expr_real_2): When handling conversions involving
13655         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13656         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
13657         as a poly_uint64 too.
13658
13659 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13660             Alan Hayward  <alan.hayward@arm.com>
13661             David Sherwood  <david.sherwood@arm.com>
13662
13663         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13664
13665 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13666             Alan Hayward  <alan.hayward@arm.com>
13667             David Sherwood  <david.sherwood@arm.com>
13668
13669         * combine.c (can_change_dest_mode): Handle polynomial
13670         REGMODE_NATURAL_SIZE.
13671         * expmed.c (store_bit_field_1): Likewise.
13672         * expr.c (store_constructor): Likewise.
13673         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13674         and polynomial REGMODE_NATURAL_SIZE.
13675         (gen_lowpart_common): Likewise.
13676         * reginfo.c (record_subregs_of_mode): Likewise.
13677         * rtlanal.c (read_modify_subreg_p): Likewise.
13678
13679 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13680             Alan Hayward  <alan.hayward@arm.com>
13681             David Sherwood  <david.sherwood@arm.com>
13682
13683         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13684         numbers of elements.
13685
13686 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13687             Alan Hayward  <alan.hayward@arm.com>
13688             David Sherwood  <david.sherwood@arm.com>
13689
13690         * match.pd: Cope with polynomial numbers of vector elements.
13691
13692 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13693             Alan Hayward  <alan.hayward@arm.com>
13694             David Sherwood  <david.sherwood@arm.com>
13695
13696         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13697         in a POINTER_PLUS_EXPR.
13698
13699 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13700             Alan Hayward  <alan.hayward@arm.com>
13701             David Sherwood  <david.sherwood@arm.com>
13702
13703         * omp-simd-clone.c (simd_clone_subparts): New function.
13704         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13705         (ipa_simd_modify_function_body): Likewise.
13706
13707 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13708             Alan Hayward  <alan.hayward@arm.com>
13709             David Sherwood  <david.sherwood@arm.com>
13710
13711         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13712         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13713         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13714         (expand_vector_condition, vector_element): Likewise.
13715         (subparts_gt): New function.
13716         (get_compute_type): Use subparts_gt.
13717         (count_type_subparts): Delete.
13718         (expand_vector_operations_1): Use subparts_gt instead of
13719         count_type_subparts.
13720
13721 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13722             Alan Hayward  <alan.hayward@arm.com>
13723             David Sherwood  <david.sherwood@arm.com>
13724
13725         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13726         (vect_compile_time_alias): ...this new function.  Do the calculation
13727         on poly_ints rather than trees.
13728         (vect_prune_runtime_alias_test_list): Update call accordingly.
13729
13730 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13731             Alan Hayward  <alan.hayward@arm.com>
13732             David Sherwood  <david.sherwood@arm.com>
13733
13734         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13735         numbers of units.
13736         (vect_schedule_slp_instance): Likewise.
13737
13738 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13739             Alan Hayward  <alan.hayward@arm.com>
13740             David Sherwood  <david.sherwood@arm.com>
13741
13742         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13743         constant and extern definitions for variable-length vectors.
13744         (vect_get_constant_vectors): Note that the number of units
13745         is known to be constant.
13746
13747 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13748             Alan Hayward  <alan.hayward@arm.com>
13749             David Sherwood  <david.sherwood@arm.com>
13750
13751         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13752         of units as polynomial.  Choose between WIDE and NARROW based
13753         on multiple_p.
13754
13755 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13756             Alan Hayward  <alan.hayward@arm.com>
13757             David Sherwood  <david.sherwood@arm.com>
13758
13759         * tree-vect-stmts.c (simd_clone_subparts): New function.
13760         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13761
13762 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13763             Alan Hayward  <alan.hayward@arm.com>
13764             David Sherwood  <david.sherwood@arm.com>
13765
13766         * tree-vect-stmts.c (vectorizable_call): Treat the number of
13767         vectors as polynomial.  Use build_index_vector for
13768         IFN_GOMP_SIMD_LANE.
13769
13770 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13771             Alan Hayward  <alan.hayward@arm.com>
13772             David Sherwood  <david.sherwood@arm.com>
13773
13774         * tree-vect-stmts.c (get_load_store_type): Treat the number of
13775         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13776         for variable-length vectors.
13777         (vectorizable_mask_load_store): Treat the number of units as
13778         polynomial, asserting that it is constant if the condition has
13779         already been enforced.
13780         (vectorizable_store, vectorizable_load): Likewise.
13781
13782 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13783             Alan Hayward  <alan.hayward@arm.com>
13784             David Sherwood  <david.sherwood@arm.com>
13785
13786         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13787         of units as polynomial.  Punt if we can't tell at compile time
13788         which vector contains the final result.
13789
13790 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13791             Alan Hayward  <alan.hayward@arm.com>
13792             David Sherwood  <david.sherwood@arm.com>
13793
13794         * tree-vect-loop.c (vectorizable_induction): Treat the number
13795         of units as polynomial.  Punt on SLP inductions.  Use an integer
13796         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
13797         cast of such a series for variable-length floating-point
13798         reductions.
13799
13800 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13801             Alan Hayward  <alan.hayward@arm.com>
13802             David Sherwood  <david.sherwood@arm.com>
13803
13804         * tree.h (build_index_vector): Declare.
13805         * tree.c (build_index_vector): New function.
13806         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13807         of units as polynomial, forcibly converting it to a constant if
13808         vectorizable_reduction has already enforced the condition.
13809         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
13810         to create a {1,2,3,...} vector.
13811         (vectorizable_reduction): Treat the number of units as polynomial.
13812         Choose vectype_in based on the largest scalar element size rather
13813         than the smallest number of units.  Enforce the restrictions
13814         relied on above.
13815
13816 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13817             Alan Hayward  <alan.hayward@arm.com>
13818             David Sherwood  <david.sherwood@arm.com>
13819
13820         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13821         number of units as polynomial.
13822
13823 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13824             Alan Hayward  <alan.hayward@arm.com>
13825             David Sherwood  <david.sherwood@arm.com>
13826
13827         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13828         * target.def (autovectorize_vector_sizes): Return the vector sizes
13829         by pointer, using vector_sizes rather than a bitmask.
13830         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13831         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13832         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13833         Likewise.
13834         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13835         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13836         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13837         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13838         * omp-general.c (omp_max_vf): Likewise.
13839         * omp-low.c (omp_clause_aligned_alignment): Likewise.
13840         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13841         * tree-vect-loop.c (vect_analyze_loop): Likewise.
13842         * tree-vect-slp.c (vect_slp_bb): Likewise.
13843         * doc/tm.texi: Regenerate.
13844         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13845         to a poly_uint64.
13846         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13847         the vector size as a poly_uint64 rather than an unsigned int.
13848         (current_vector_size): Change from an unsigned int to a poly_uint64.
13849         (get_vectype_for_scalar_type): Update accordingly.
13850         * tree.h (build_truth_vector_type): Take the size and number of
13851         units as a poly_uint64 rather than an unsigned int.
13852         (build_vector_type): Add a temporary overload that takes
13853         the number of units as a poly_uint64 rather than an unsigned int.
13854         * tree.c (make_vector_type): Likewise.
13855         (build_truth_vector_type): Take the number of units as a poly_uint64
13856         rather than an unsigned int.
13857
13858 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13859             Alan Hayward  <alan.hayward@arm.com>
13860             David Sherwood  <david.sherwood@arm.com>
13861
13862         * target.def (get_mask_mode): Take the number of units and length
13863         as poly_uint64s rather than unsigned ints.
13864         * targhooks.h (default_get_mask_mode): Update accordingly.
13865         * targhooks.c (default_get_mask_mode): Likewise.
13866         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13867         * doc/tm.texi: Regenerate.
13868
13869 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13870             Alan Hayward  <alan.hayward@arm.com>
13871             David Sherwood  <david.sherwood@arm.com>
13872
13873         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13874         * omp-general.c (omp_max_vf): Likewise.
13875         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13876         (expand_omp_simd): Handle polynomial safelen.
13877         * omp-low.c (omplow_simd_context): Add a default constructor.
13878         (omplow_simd_context::max_vf): Change from int to poly_uint64.
13879         (lower_rec_simd_input_clauses): Update accordingly.
13880         (lower_rec_input_clauses): Likewise.
13881
13882 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13883             Alan Hayward  <alan.hayward@arm.com>
13884             David Sherwood  <david.sherwood@arm.com>
13885
13886         * tree-vectorizer.h (vect_nunits_for_cost): New function.
13887         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13888         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13889         (vect_analyze_slp_cost): Likewise.
13890         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13891         (vect_model_load_cost): Likewise.
13892
13893 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13894             Alan Hayward  <alan.hayward@arm.com>
13895             David Sherwood  <david.sherwood@arm.com>
13896
13897         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13898         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13899         from an unsigned int * to a poly_uint64_pod *.
13900         (calculate_unrolling_factor): New function.
13901         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
13902
13903 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13904             Alan Hayward  <alan.hayward@arm.com>
13905             David Sherwood  <david.sherwood@arm.com>
13906
13907         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13908         from an unsigned int to a poly_uint64.
13909         (_loop_vec_info::slp_unrolling_factor): Likewise.
13910         (_loop_vec_info::vectorization_factor): Change from an int
13911         to a poly_uint64.
13912         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13913         (vect_get_num_vectors): New function.
13914         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13915         (vect_get_num_copies): Use vect_get_num_vectors.
13916         (vect_analyze_data_ref_dependences): Change max_vf from an int *
13917         to an unsigned int *.
13918         (vect_analyze_data_refs): Change min_vf from an int * to a
13919         poly_uint64 *.
13920         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13921         than an unsigned HOST_WIDE_INT.
13922         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13923         (vect_analyze_data_ref_dependence): Change max_vf from an int *
13924         to an unsigned int *.
13925         (vect_analyze_data_ref_dependences): Likewise.
13926         (vect_compute_data_ref_alignment): Handle polynomial vf.
13927         (vect_enhance_data_refs_alignment): Likewise.
13928         (vect_prune_runtime_alias_test_list): Likewise.
13929         (vect_shift_permute_load_chain): Likewise.
13930         (vect_supportable_dr_alignment): Likewise.
13931         (dependence_distance_ge_vf): Take the vectorization factor as a
13932         poly_uint64 rather than an unsigned HOST_WIDE_INT.
13933         (vect_analyze_data_refs): Change min_vf from an int * to a
13934         poly_uint64 *.
13935         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13936         vfm1 as a poly_uint64 rather than an int.  Make the same change
13937         for the returned bound_scalar.
13938         (vect_gen_vector_loop_niters): Handle polynomial vf.
13939         (vect_do_peeling): Likewise.  Update call to
13940         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13941         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13942         be constant.
13943         * tree-vect-loop.c (vect_determine_vectorization_factor)
13944         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13945         (vect_get_known_peeling_cost): Likewise.
13946         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13947         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13948         (vect_transform_loop): Likewise.  Use the lowest possible VF when
13949         updating the upper bounds of the loop.
13950         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
13951         rather than an int.
13952         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13953         polynomial unroll factors.
13954         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13955         (vect_make_slp_decision): Likewise.
13956         (vect_supported_load_permutation_p): Likewise, and polynomial
13957         vf too.
13958         (vect_analyze_slp_cost): Handle polynomial vf.
13959         (vect_slp_analyze_node_operations): Likewise.
13960         (vect_slp_analyze_bb_1): Likewise.
13961         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13962         than an unsigned HOST_WIDE_INT.
13963         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13964         (vectorizable_load): Handle polynomial vf.
13965         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13966         a poly_uint64.
13967         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13968
13969 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13970             Alan Hayward  <alan.hayward@arm.com>
13971             David Sherwood  <david.sherwood@arm.com>
13972
13973         * match.pd: Handle bit operations involving three constants
13974         and try to fold one pair.
13975
13976 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
13977
13978         * tree-vect-loop-manip.c: Include gimple-fold.h.
13979         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13980         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
13981         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13982         Add a path that uses a step of VF instead of 1, but disable it
13983         for now.
13984         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13985         and niters_no_overflow parameters.  Update calls to
13986         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13987         Create a new SSA name if the latter choses to use a ste other
13988         than zero, and return it via niters_vector_mult_vf_var.
13989         * tree-vect-loop.c (vect_transform_loop): Update calls to
13990         vect_do_peeling, vect_gen_vector_loop_niters and
13991         slpeel_make_loop_iterate_ntimes.
13992         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13993         (vect_gen_vector_loop_niters): Update declarations after above changes.
13994
13995 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
13996
13997         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13998         128-bit round to integer instructions.
13999         (ceil<mode>2): Likewise.
14000         (btrunc<mode>2): Likewise.
14001         (round<mode>2): Likewise.
14002
14003 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
14004
14005         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14006         unaligned VSX load/store on P8/P9.
14007         (expand_block_clear): Allow the use of unaligned VSX
14008         load/store on P8/P9.
14009
14010 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
14011
14012         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14013         New function.
14014         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14015         swap associated with both a load and a store.
14016
14017 2018-01-02  Andrew Waterman  <andrew@sifive.com>
14018
14019         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14020         * config/riscv/riscv.md (clear_cache): Use it.
14021
14022 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
14023
14024         * web.c: Remove out-of-date comment.
14025
14026 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14027
14028         * expr.c (fixup_args_size_notes): Check that any existing
14029         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14030         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14031         (emit_single_push_insn): ...here.
14032
14033 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14034
14035         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14036         (const_vector_encoded_nelts): New function.
14037         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14038         (const_vector_int_elt, const_vector_elt): Declare.
14039         * emit-rtl.c (const_vector_int_elt_1): New function.
14040         (const_vector_elt): Likewise.
14041         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14042         of CONST_VECTOR_ELT.
14043
14044 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14045
14046         * expr.c: Include rtx-vector-builder.h.
14047         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14048         directly on the tree encoding.
14049         (const_vector_from_tree): Likewise.
14050         * optabs.c: Include rtx-vector-builder.h.
14051         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14052         sequence of "u" values.
14053         * vec-perm-indices.c: Include rtx-vector-builder.h.
14054         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14055         directly on the vec_perm_indices encoding.
14056
14057 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14058
14059         * doc/rtl.texi (const_vector): Describe new encoding scheme.
14060         * Makefile.in (OBJS): Add rtx-vector-builder.o.
14061         * rtx-vector-builder.h: New file.
14062         * rtx-vector-builder.c: Likewise.
14063         * rtl.h (rtx_def::u2): Add a const_vector field.
14064         (CONST_VECTOR_NPATTERNS): New macro.
14065         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14066         (CONST_VECTOR_DUPLICATE_P): Likewise.
14067         (CONST_VECTOR_STEPPED_P): Likewise.
14068         (CONST_VECTOR_ENCODED_ELT): Likewise.
14069         (const_vec_duplicate_p): Check for a duplicated vector encoding.
14070         (unwrap_const_vec_duplicate): Likewise.
14071         (const_vec_series_p): Check for a non-duplicated vector encoding.
14072         Say that the function only returns true for integer vectors.
14073         * emit-rtl.c: Include rtx-vector-builder.h.
14074         (gen_const_vec_duplicate_1): Delete.
14075         (gen_const_vector): Call gen_const_vec_duplicate instead of
14076         gen_const_vec_duplicate_1.
14077         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14078         (gen_const_vec_duplicate): Use rtx_vector_builder.
14079         (gen_const_vec_series): Likewise.
14080         (gen_rtx_CONST_VECTOR): Likewise.
14081         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14082         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14083         Build a new vector rather than modifying a CONST_VECTOR in-place.
14084         (handle_special_swappables): Update call accordingly.
14085         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14086         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14087         Build a new vector rather than modifying a CONST_VECTOR in-place.
14088         (handle_special_swappables): Update call accordingly.
14089
14090 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14091
14092         * simplify-rtx.c (simplify_const_binary_operation): Use
14093         CONST_VECTOR_ELT instead of XVECEXP.
14094
14095 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14096
14097         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14098         the selector elements to be different from the data elements
14099         if the selector is a VECTOR_CST.
14100         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14101         ssizetype for the selector.
14102
14103 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14104
14105         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14106         before testing each element individually.
14107         * tree-vect-generic.c (lower_vec_perm): Likewise.
14108
14109 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14110
14111         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14112         * selftest-run-tests.c (selftest::run_tests): Call it.
14113         * vector-builder.h (vector_builder::operator ==): New function.
14114         (vector_builder::operator !=): Likewise.
14115         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14116         (vec_perm_indices::all_from_input_p): New function.
14117         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14118         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14119         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14120         instead of reading the VECTOR_CST directly.  Detect whether both
14121         vector inputs are the same before constructing the vec_perm_indices,
14122         and update the number of inputs argument accordingly.  Use the
14123         utility functions added above.  Only construct sel2 if we need to.
14124
14125 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14126
14127         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14128         the broadcast of the low byte.
14129         (expand_mult_highpart): Use an explicit encoding for the permutes.
14130         * optabs-query.c (can_mult_highpart_p): Likewise.
14131         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14132         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14133         (vectorizable_bswap): Likewise.
14134         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14135         explicit encoding for the power-of-2 permutes.
14136         (vect_permute_store_chain): Likewise.
14137         (vect_grouped_load_supported): Likewise.
14138         (vect_permute_load_chain): Likewise.
14139
14140 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14141
14142         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14143         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14144         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14145         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14146         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14147         (vect_gen_perm_mask_any): Likewise.
14148
14149 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14150
14151         * int-vector-builder.h: New file.
14152         * vec-perm-indices.h: Include int-vector-builder.h.
14153         (vec_perm_indices): Redefine as an int_vector_builder.
14154         (auto_vec_perm_indices): Delete.
14155         (vec_perm_builder): Redefine as a stand-alone class.
14156         (vec_perm_indices::vec_perm_indices): New function.
14157         (vec_perm_indices::clamp): Likewise.
14158         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14159         (vec_perm_indices::new_vector): New function.
14160         (vec_perm_indices::new_expanded_vector): Update for new
14161         vec_perm_indices class.
14162         (vec_perm_indices::rotate_inputs): New function.
14163         (vec_perm_indices::all_in_range_p): Operate directly on the
14164         encoded form, without computing elided elements.
14165         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14166         encoding.  Update for new vec_perm_indices class.
14167         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14168         the given vec_perm_builder.
14169         (expand_vec_perm_var): Update vec_perm_builder constructor.
14170         (expand_mult_highpart): Use vec_perm_builder instead of
14171         auto_vec_perm_indices.
14172         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14173         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
14174         or double series encoding as appropriate.
14175         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14176         vec_perm_indices instead of auto_vec_perm_indices.
14177         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14178         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14179         (vect_permute_store_chain): Likewise.
14180         (vect_grouped_load_supported): Likewise.
14181         (vect_permute_load_chain): Likewise.
14182         (vect_shift_permute_load_chain): Likewise.
14183         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14184         (vect_transform_slp_perm_load): Likewise.
14185         (vect_schedule_slp_instance): Likewise.
14186         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14187         (vectorizable_mask_load_store): Likewise.
14188         (vectorizable_bswap): Likewise.
14189         (vectorizable_store): Likewise.
14190         (vectorizable_load): Likewise.
14191         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14192         vec_perm_indices instead of auto_vec_perm_indices.  Use
14193         tree_to_vec_perm_builder to read the vector from a tree.
14194         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14195         vec_perm_builder instead of a vec_perm_indices.
14196         (have_whole_vector_shift): Use vec_perm_builder and
14197         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
14198         truncation to calc_vec_perm_mask_for_shift.
14199         (vect_create_epilog_for_reduction): Likewise.
14200         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14201         from auto_vec_perm_indices to vec_perm_indices.
14202         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14203         instead of changing individual elements.
14204         (aarch64_vectorize_vec_perm_const): Use new_vector to install
14205         the vector in d.perm.
14206         * config/arm/arm.c (expand_vec_perm_d::perm): Change
14207         from auto_vec_perm_indices to vec_perm_indices.
14208         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14209         instead of changing individual elements.
14210         (arm_vectorize_vec_perm_const): Use new_vector to install
14211         the vector in d.perm.
14212         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14213         Update vec_perm_builder constructor.
14214         (rs6000_expand_interleave): Likewise.
14215         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14216         (rs6000_expand_interleave): Likewise.
14217
14218 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14219
14220         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14221         to qimode could truncate the indices.
14222         * optabs.c (expand_vec_perm_var): Likewise.
14223
14224 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14225
14226         * Makefile.in (OBJS): Add vec-perm-indices.o.
14227         * vec-perm-indices.h: New file.
14228         * vec-perm-indices.c: Likewise.
14229         * target.h (vec_perm_indices): Replace with a forward class
14230         declaration.
14231         (auto_vec_perm_indices): Move to vec-perm-indices.h.
14232         * optabs.h: Include vec-perm-indices.h.
14233         (expand_vec_perm): Delete.
14234         (selector_fits_mode_p, expand_vec_perm_var): Declare.
14235         (expand_vec_perm_const): Declare.
14236         * target.def (vec_perm_const_ok): Replace with...
14237         (vec_perm_const): ...this new hook.
14238         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14239         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14240         * doc/tm.texi: Regenerate.
14241         * optabs.def (vec_perm_const): Delete.
14242         * doc/md.texi (vec_perm_const): Likewise.
14243         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14244         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14245         expand_vec_perm for constant permutation vectors.  Assert that
14246         the mode of variable permutation vectors is the integer equivalent
14247         of the mode that is being permuted.
14248         * optabs-query.h (selector_fits_mode_p): Declare.
14249         * optabs-query.c: Include vec-perm-indices.h.
14250         (selector_fits_mode_p): New function.
14251         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14252         is defined, instead of checking whether the vec_perm_const_optab
14253         exists.  Use targetm.vectorize.vec_perm_const instead of
14254         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
14255         fit in the vector mode before using a variable permute.
14256         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14257         vec_perm_indices instead of an rtx.
14258         (expand_vec_perm): Replace with...
14259         (expand_vec_perm_const): ...this new function.  Take the selector
14260         as a vec_perm_indices rather than an rtx.  Also take the mode of
14261         the selector.  Update call to shift_amt_for_vec_perm_mask.
14262         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14263         Use vec_perm_indices::new_expanded_vector to expand the original
14264         selector into bytes.  Check whether the indices fit in the vector
14265         mode before using a variable permute.
14266         (expand_vec_perm_var): Make global.
14267         (expand_mult_highpart): Use expand_vec_perm_const.
14268         * fold-const.c: Includes vec-perm-indices.h.
14269         * tree-ssa-forwprop.c: Likewise.
14270         * tree-vect-data-refs.c: Likewise.
14271         * tree-vect-generic.c: Likewise.
14272         * tree-vect-loop.c: Likewise.
14273         * tree-vect-slp.c: Likewise.
14274         * tree-vect-stmts.c: Likewise.
14275         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14276         Delete.
14277         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14278         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14279         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14280         (aarch64_vectorize_vec_perm_const): ...this new function.
14281         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14282         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14283         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14284         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14285         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14286         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14287         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14288         into...
14289         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
14290         check for NEON modes.
14291         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14292         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14293         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14294         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14295         into...
14296         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
14297         the old VEC_PERM_CONST conditions.
14298         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14299         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14300         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14301         (ia64_vectorize_vec_perm_const_ok): Merge into...
14302         (ia64_vectorize_vec_perm_const): ...this new function.
14303         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14304         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14305         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14306         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14307         * config/mips/mips.c (mips_expand_vec_perm_const)
14308         (mips_vectorize_vec_perm_const_ok): Merge into...
14309         (mips_vectorize_vec_perm_const): ...this new function.
14310         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14311         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14312         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14313         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14314         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14315         (rs6000_expand_vec_perm_const): Delete.
14316         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14317         Delete.
14318         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14319         (altivec_expand_vec_perm_const_le): Take each operand individually.
14320         Operate on constant selectors rather than rtxes.
14321         (altivec_expand_vec_perm_const): Likewise.  Update call to
14322         altivec_expand_vec_perm_const_le.
14323         (rs6000_expand_vec_perm_const): Delete.
14324         (rs6000_vectorize_vec_perm_const_ok): Delete.
14325         (rs6000_vectorize_vec_perm_const): New function.
14326         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14327         an element count and rtx array.
14328         (rs6000_expand_extract_even): Update call accordingly.
14329         (rs6000_expand_interleave): Likewise.
14330         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14331         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14332         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14333         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14334         (rs6000_expand_vec_perm_const): Delete.
14335         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14336         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14337         (altivec_expand_vec_perm_const_le): Take each operand individually.
14338         Operate on constant selectors rather than rtxes.
14339         (altivec_expand_vec_perm_const): Likewise.  Update call to
14340         altivec_expand_vec_perm_const_le.
14341         (rs6000_expand_vec_perm_const): Delete.
14342         (rs6000_vectorize_vec_perm_const_ok): Delete.
14343         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
14344         reference to the SPE evmerge intructions.
14345         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14346         an element count and rtx array.
14347         (rs6000_expand_extract_even): Update call accordingly.
14348         (rs6000_expand_interleave): Likewise.
14349         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14350         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14351         new function.
14352         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14353
14354 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14355
14356         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14357         vector mode and that that mode matches the mode of the data
14358         being permuted.
14359         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14360         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
14361         directly using expand_vec_perm_1 when forcing selectors into
14362         registers.
14363         (expand_vec_perm_var): New function, split out from expand_vec_perm.
14364
14365 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14366
14367         * optabs-query.h (can_vec_perm_p): Delete.
14368         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14369         * optabs-query.c (can_vec_perm_p): Split into...
14370         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14371         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14372         particular selector is valid.
14373         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14374         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14375         (vect_grouped_load_supported): Likewise.
14376         (vect_shift_permute_load_chain): Likewise.
14377         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14378         (vect_transform_slp_perm_load): Likewise.
14379         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14380         (vectorizable_bswap): Likewise.
14381         (vect_gen_perm_mask_checked): Likewise.
14382         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
14383         implementations of variable permutation vectors into account
14384         when deciding which selector to use.
14385         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14386         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14387         with a false third argument.
14388         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14389         to test whether the constant selector is valid and can_vec_perm_var_p
14390         to test whether a variable selector is valid.
14391
14392 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14393
14394         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14395         * optabs-query.c (can_vec_perm_p): Likewise.
14396         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14397         instead of vec_perm_indices.
14398         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14399         (vect_gen_perm_mask_checked): Likewise,
14400         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14401         (vect_gen_perm_mask_checked): Likewise,
14402
14403 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
14404
14405         * optabs-query.h (qimode_for_vec_perm): Declare.
14406         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14407         (qimode_for_vec_perm): ...this new function.
14408         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14409
14410 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
14411
14412         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14413         does not have a conditional at the top.
14414
14415 2018-01-02  Richard Biener  <rguenther@suse.de>
14416
14417         * ipa-inline.c (big_speedup_p): Fix expression.
14418
14419 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
14420
14421         PR target/81616
14422         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14423         for generic 4->6.
14424
14425 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
14426
14427         PR target/81616
14428         Generic tuning.
14429         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14430         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14431         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14432         cond_taken_branch_cost 3->4.
14433
14434 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
14435
14436         PR tree-optimization/83581
14437         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14438         TODO_cleanup_cfg if any changes have been made.
14439
14440         PR middle-end/83608
14441         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14442         convert_modes if target mode has the right side, but different mode
14443         class.
14444
14445         PR middle-end/83609
14446         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14447         last argument when extracting from CONCAT.  If either from_real or
14448         from_imag is NULL, use expansion through memory.  If result is not
14449         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14450         the parts directly to inner mode, if even that fails, use expansion
14451         through memory.
14452
14453         PR middle-end/83623
14454         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14455         check for bswap in mode rather than HImode and use that in expand_unop
14456         too.
14457 \f
14458 Copyright (C) 2018 Free Software Foundation, Inc.
14459
14460 Copying and distribution of this file, with or without modification,
14461 are permitted in any medium without royalty provided the copyright
14462 notice and this notice are preserved.