1 2018-04-03 Martin Liska <mliska@suse.cz>
3 PR tree-optimization/82491
4 * rtl.h (strip_offset_and_add): Replace += suboffset with
5 poly_uint64 () + suboffset.
7 2018-03-29 Martin Liska <mliska@suse.cz>
8 Martin Jambor <mjambor@suse.cz>
11 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
12 param_type is not an integral or pointer type.
14 2018-04-03 Richard Biener <rguenther@suse.de>
16 * sese.h (recompute_all_dominators): Remove.
18 2018-04-02 Martin Sebor <msebor@redhat.com>
20 * doc/invoke.texi (-Wrestrict): Fix typos.
22 2018-04-02 Jim Wilson <jimw@sifive.com>
24 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
25 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
26 (<optab>di3, <optab>si3_extend): Likewise.
27 (<optab>si3_mask, <optab>si3_mask_1): New.
28 (<optab>di3_mask, <optab>di3_mask_1): New.
29 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
30 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
31 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
33 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
35 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
38 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
40 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
41 (nds32_canonicalize_comparison): New function.
43 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
44 Kito Cheng <kito.cheng@gmail.com>
45 Kuan-Lin Chen <kuanlinchentw@gmail.com>
47 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
48 * config/nds32/constants.md (unspec_volatile_element): Add
49 UNSPEC_VOLATILE_RELAX_GROUP.
50 * config/nds32/nds32-relax-opt.c: New file.
51 * config/nds32/nds32-predicates.c
52 (nds32_symbol_load_store_p): New function.
53 * config/nds32/nds32-protos.h
54 (nds32_symbol_load_store_p): Declare function.
55 (make_pass_nds32_relax_opt): Declare new rtl pass function.
56 * config/nds32/nds32.c
57 (nds32_register_pass): New function to register pass.
58 (nds32_register_passes): New function to register passes.
59 * config/nds32/nds32.md (relax_group): New pattern.
60 * config/nds32/nds32.opt (mrelax-hint): New option.
61 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
63 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
65 * config/nds32/t-nds32: Modify files dependency.
67 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
69 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
70 (PROFILE_HOOK): Define its implementation.
72 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
74 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
77 2018-04-01 Jakub Jelinek <jakub@redhat.com>
80 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
81 (V_128_256): New mode iterator.
82 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
83 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
84 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
86 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
87 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
89 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
92 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
95 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
98 * config/rs6000/vsx.md (??r): New mode attribute.
99 (*vsx_mov<mode>_64bit): Use it.
100 (*vsx_mov<mode>_32bit): Likewise.
102 2018-03-30 Martin Sebor <msebor@redhat.com>
104 PR tree-optimization/84818
105 * builtins.c (check_access): Use warning_n.
107 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
110 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
112 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
115 2018-03-30 Julia Koval <julia.koval@intel.com>
118 * x86-tune.def (movx, partial_reg_dependency): Enable for
121 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
124 * lra-constraints.c (process_alt_operands): Move setting
125 this_alternative_matches below.
127 2018-03-29 Martin Liska <mliska@suse.cz>
130 * doc/invoke.texi: Document how LTO works with debug info.
131 Describe auto-load support of binutils. Mention 'x86-64'
132 as valid option value of -march option.
134 2018-03-29 Jakub Jelinek <jakub@redhat.com>
136 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
139 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
140 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
141 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
144 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
147 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
148 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
149 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
150 for RS6000_BTM_POWERPC64.
151 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
152 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
153 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
158 2018-03-28 Carl Love <cel@us.ibm.com>
161 2017-09-27 Carl Love <cel@us.ibm.com>
163 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
164 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
165 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
168 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
170 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
171 instead of __vector bool.
172 (_mm_max_pu8): Likewise.
173 (_mm_min_pi16): Likewise.
175 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
178 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
182 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
184 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
185 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
186 (div_extend): Likewise.
187 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
189 (__builtin_divweuo): Likewise.
190 (__builtin_divdeo): Likewise.
191 (__builtin_divdeuo): Likewise.
193 2018-03-28 Jakub Jelinek <jakub@redhat.com>
196 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
197 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
199 PR tree-optimization/82004
200 * gimple-match-head.c (optimize_pow_to_exp): New function.
201 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
202 Don't fold to exp if optimize_pow_to_exp is false.
204 2018-03-28 Martin Liska <mliska@suse.cz>
207 * calls.c (initialize_argument_information): Fix trailing space.
208 * common.opt: Fix typo and provide better explanation for
209 -fsanitize-coverage option.
210 * config/i386/i386.opt: Fix typo.
212 2018-03-28 Jakub Jelinek <jakub@redhat.com>
213 Martin Liska <mliska@suse.cz>
216 * gimplify.c (asan_poison_variable): Don't do the check for
217 gimplify_omp_ctxp here.
218 (gimplify_decl_expr): Do it here.
219 (gimplify_target_expr): Likewise.
221 2018-03-28 Martin Liska <mliska@suse.cz>
224 * config/i386/i386.c (ix86_function_arg_advance): Do not call
225 chkp_type_bounds_count if MPX is not enabled.
227 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
229 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
231 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
234 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
235 function to create the function decl for complex long double
236 multiply and divide for -mabi=ieeelongdouble.
237 (init_float128_ieee): Call it.
239 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
242 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
243 -fcf-protection=branch -mibt.
244 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
246 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
249 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
251 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
254 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
255 extern array declarations.
257 2018-03-27 Richard Biener <rguenther@suse.de>
260 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
261 explicit single_use checks.
263 2018-03-27 Richard Biener <rguenther@suse.de>
265 PR tree-optimization/85082
266 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
269 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
271 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
272 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
273 Turn on fasynchronous-unwind-tables and funwind-tables.
275 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
278 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
279 (*bmi_blsr_<mode>_ccz): Ditto.
281 2018-03-26 Tom de Vries <tom@codesourcery.com>
283 PR tree-optimization/85063
284 * omp-general.c (offloading_function_p): New function. Factor out
286 * omp-offload.c (pass_omp_target_link::gate): ... here.
287 * omp-general.h (offloading_function_p): Declare.
288 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
289 with attribute omp declare target for offloading functions.
291 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
293 PR tree-optimization/84005
294 * tree-data-ref.h (get_base_for_alignment): Declare.
295 * tree-data-ref.c (get_base_for_alignment_1): New function.
296 (get_base_for_alignment): Likewise.
297 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
298 get_base_for_alignment to find a suitable base object, instead
299 of always using drb->base_address.
301 2018-03-23 Jakub Jelinek <jakub@redhat.com>
304 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
305 known size by default.
307 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
310 * lra-constraints.c (process_alt_operands): Don't match BLKmode
311 and non BLKmode operands.
313 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
316 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
319 2018-03-23 Richard Biener <rguenther@suse.de>
322 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
323 we are going to emit early debug for LTO.
325 2018-03-23 Jakub Jelinek <jakub@redhat.com>
328 * function.c (match_asm_constraints_1): Don't optimize if input
329 doesn't satisfy general_operand predicate for output's mode.
332 * alias.c (write_dependence_p): Don't require for x_canonicalized
333 non-VOIDmode if x has VOIDmode.
336 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
337 just don't try to optimize it rather than assert it never happens.
339 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
341 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
342 macro expansions for definition of ST_INTERNAL_<mode> and
343 LD_INTERNAL_<mode> builtins.
344 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
346 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
348 (altivec_expand_st_builtin): Likewise.
349 (altivec_expand_builtin): Remove calls to deleted functions.
350 (rs6000_address_for_altivec): Delete this function.
351 * config/rs6000/vector.md: Remove expands for
352 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
354 2018-03-22 Sudakshina Das <sudi.das@arm.com>
357 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
358 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
359 re-computing once computed.
360 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
361 (arm_init_machine_status): Initialize
362 machine->static_chain_stack_bytes.
364 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
367 * doc/extend.texi: Add four new prototypes for vec_ld.
368 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
369 definitions for more logical presentation.
370 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
371 entries for V1TI variants of __builtin_altivec_ld builtin.
372 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
373 handling of V1TI variant of LVX icode pattern.
374 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
375 (rs6000_gimple_fold_builtin): Likewise.
376 (altivec_init_builtins): Add code to define
377 __builtin_altivec_lvx_v1ti function.
379 2018-03-22 Jakub Jelinek <jakub@redhat.com>
382 * function.c (match_asm_constraints_1): Don't do the optimization
383 if input isn't a REG, SUBREG, MEM or constant.
385 2018-03-22 Tom de Vries <tom@codesourcery.com>
387 PR tree-optimization/84956
388 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
389 bb_has_abnormal_pred.
391 2018-03-22 Jakub Jelinek <jakub@redhat.com>
394 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
395 DECL_INITIAL (decl) to decl at the end.
396 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
399 2018-03-21 Joseph Myers <joseph@codesourcery.com>
401 * doc/extend.texi (__builtin_tgmath): Document when complex
402 integer types are treated as _Complex _Float64.
404 2018-03-21 Tom de Vries <tom@codesourcery.com>
406 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
408 2018-03-21 Jakub Jelinek <jakub@redhat.com>
410 PR tree-optimization/84960
411 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
412 if it is ENTRY block, move them into single succ of ENTRY in that case.
414 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
416 PR tree-optimization/84811
417 * poly-int.h (poly_span_traits): Remove the T3 parameter and
418 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
419 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
420 (known_subrange_p): Update accordingly. Cast each value involved
421 in the size comparison, rather than casting the result of the
424 2018-03-21 Jakub Jelinek <jakub@redhat.com>
426 PR tree-optimization/84982
427 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
428 by flipping the least significant bit rather than all bits from
429 bitpos to bitpos + bitsize - 1.
431 2018-03-21 Nathan Sidwell <nathan@acm.org>
433 * doc/extend.texi (Deprecated Features): Remove mention of
434 long-deleted deprecations.
436 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
439 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
440 * configure: Regenerate.
442 2018-03-21 Tom de Vries <tom@codesourcery.com>
444 PR tree-optimization/83126
445 * tree-parloops.c (num_phis): New function.
446 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
448 2018-03-21 Nathan Sidwell <nathan@acm.org>
450 * doc/extend.texi (Deprecated Features): Update deprecated flags,
451 mention anon-struct/union members and trailing attributes.
453 2018-03-21 Bin Cheng <bin.cheng@arm.com>
455 PR tree-optimization/84969
456 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
457 builtin memset partitions if they set different rhs values.
459 2018-03-21 Jakub Jelinek <jakub@redhat.com>
461 PR rtl-optimization/84989
462 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
463 VEC_DUPLICATE with scalar result mode.
465 2018-03-21 Martin Liska <mliska@suse.cz>
468 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
469 not intended return statement.
471 2018-03-21 Martin Liska <mliska@suse.cz>
474 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
475 (chkp_find_bound_slots_1): Limit number of iterations.
477 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
480 * Minor grammar fixes for x86 options.
482 2018-03-20 Jakub Jelinek <jakub@redhat.com>
485 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
486 holding REG_CFA_RESTORE notes, instead turn them into a USE.
488 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
491 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
492 (altivec_lvx_<mode>_1op): Likewise.
493 (altivec_stvx_<mode>_2op): Likewise.
494 (altivec_stvx_<mode>_1op): Likewise.
495 (altivec_lvx_<VM2:mode>): New define_expand.
496 (altivec_stvx_<VM2:mode>): Likewise.
497 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
498 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
499 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
500 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
501 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
502 (rs6000_gen_lvx): Likewise.
503 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
504 (altivec_expand_stv_builtin): Likewise.
505 (altivec_expand_builtin): Likewise.
506 * config/rs6000/vector.md: Likewise.
508 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
511 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
514 2018-03-20 Richard Biener <rguenther@suse.de>
517 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
518 sign-conversions as zero, fall back to standard scalar_stmt
521 2018-03-20 Martin Liska <mliska@suse.cz>
524 * predict.c (rebuild_frequencies): Handle case when we have
525 PROFILE_ABSENT, but flag_guess_branch_prob is false.
527 2018-03-20 Jakub Jelinek <jakub@redhat.com>
530 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
531 flag_section_anchors.
532 * varasm.c (use_blocks_for_decl_p): Remove hack for
536 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
538 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
539 be created, use lowpart_subreg of operands[0] rather than operands[0]
541 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
542 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
543 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
544 and n constraint instead of aarch64_shift_imm_di and Usd.
545 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
546 (*aarch64_<optab>_reg_minus<mode>3): ... this.
548 2018-03-20 Sudakshina Das <sudi.das@arm.com>
551 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
552 to favor GPR over NEON registers.
553 (<shift>di3_neon): Likewise.
555 2018-03-20 Tom de Vries <tom@codesourcery.com>
558 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
559 (nvptx_process_pars): Emit bar.sync asap and alap.
561 2018-03-20 Tom de Vries <tom@codesourcery.com>
564 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
565 seen_label if seen_label is already set.
567 2018-03-20 Jakub Jelinek <jakub@redhat.com>
570 * config/i386/i386.c (fold_builtin_cpu): For features above 31
571 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
572 Use 1U instead of 1. Formatting fixes.
575 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
576 instead of TREE_TYPE (s1) for the return value.
578 2018-03-19 Jakub Jelinek <jakub@redhat.com>
580 PR tree-optimization/84946
581 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
582 bitsize + bitsize in poly_uint64 rather than poly_int64.
585 * dwarf2asm.c: Include fold-const.c.
586 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
587 of decl rather than decl itself.
589 PR rtl-optimization/84643
590 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
592 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
595 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
596 calling assemble_variable.
598 2018-03-19 Sudakshina Das <sudi.das@arm.com>
601 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
602 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
604 2018-03-19 Jim Wilson <jimw@sifive.com>
607 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
608 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
609 (riscv_first_stack_step): Likewise.
610 (riscv_option_override): Use STACK_BOUNDARY instead of
612 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
614 (BIGGEST_ALIGNMENT): Set to 128.
615 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
616 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
619 2018-03-19 Richard Biener <rguenther@suse.de>
621 PR tree-optimization/84933
622 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
623 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
625 2018-03-19 Richard Biener <rguenther@suse.de>
627 PR tree-optimization/84859
628 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
629 (cond_if_else_store_replacement): Perform sinking operation on
630 single-store BBs regardless of MAX_STORES_TO_SINK setting.
631 Generalize what a BB with a single eligible store is.
633 2018-03-19 Richard Biener <rguenther@suse.de>
635 PR tree-optimization/84929
636 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
637 chrec_is_positive against non-chrec arg.
639 2018-03-19 Tamar Christina <tamar.christina@arm.com>
642 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
644 2018-03-18 Martin Liska <mliska@suse.cz>
646 PR rtl-optimization/84635
647 * regrename.c (build_def_use): Use matches_mode only when
650 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
652 PR tree-optimization/84913
653 * tree-vect-loop.c (vectorizable_reduction): Don't try to
654 vectorize chains of COND_EXPRs.
656 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
658 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
660 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
662 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
664 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
666 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
668 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
669 Kito Cheng <kito.cheng@gmail.com>
671 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
672 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
673 (nds32_adjust_reg_alloc_order): New function.
674 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
676 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
678 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
679 nds32_print_operand, nds32_print_operand_address): Use
680 HOST_WIDE_INT_PRINT_DEC instead.
682 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
684 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
686 2018-03-17 Jakub Jelinek <jakub@redhat.com>
689 * config/i386/i386.c (initial_ix86_tune_features,
690 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
692 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
693 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
694 rather than 1u << ix86_tune. Formatting fix.
695 (ix86_option_override_internal): Change ix86_arch_mask from
696 unsigned int to unsigned HOST_WIDE_INT, initialize to
697 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
698 (ix86_function_specific_restore): Likewise.
700 2018-03-16 Jakub Jelinek <jakub@redhat.com>
703 * postreload.c (reload_combine_recognize_pattern): Perform
704 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
705 truncate_int_for_mode the result for the destination's mode.
708 * hsa-gen.c (mem_type_for_type): Fix comment typo.
709 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
711 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
714 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
717 * lra-assigns.c (lra_split_hard_reg_for): Don't use
718 regno_allocno_class_array and sorted_pseudos.
719 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
720 insns where regno is used.
722 2018-03-16 Martin Liska <mliska@suse.cz>
725 * multiple_target.c (create_dispatcher_calls): Redirect
726 reference in the symbol table.
728 2018-03-16 Martin Liska <mliska@suse.cz>
731 * multiple_target.c (create_dispatcher_calls): Redirect also
734 2018-03-16 Jakub Jelinek <jakub@redhat.com>
738 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
739 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
740 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
742 2018-03-16 Julia Koval <julia.koval@intel.com>
744 * doc/invoke.texi (Skylake Server): Add CLWB.
745 Cannonlake): Remove CLWB.
747 2018-03-16 Jakub Jelinek <jakub@redhat.com>
749 PR tree-optimization/84841
750 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
752 (FLOAT_ONE_CONST_TYPE): Define.
753 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
754 (sort_by_operand_rank): Put entries with higher constant_type last
755 rather than first to match comments.
757 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
759 * config/nios2/nios2.md (movsi_internal): Fix thinko in
762 2018-03-15 Jakub Jelinek <jakub@redhat.com>
765 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
766 check and use address of target always.
768 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
771 * config/i386/i386.c (indirect_thunk_needed): Update comments.
772 (indirect_thunk_bnd_needed): Likewise.
773 (indirect_thunks_used): Likewise.
774 (indirect_thunks_bnd_used): Likewise.
775 (indirect_return_needed): New.
776 (indirect_return_bnd_needed): Likewise.
777 (output_indirect_thunk_function): Add a bool argument for
779 (output_indirect_thunk_function): Don't generate alias for
780 function return thunk.
781 (ix86_code_end): Call output_indirect_thunk_function to generate
782 function return thunks.
783 (ix86_output_function_return): Set indirect_return_bnd_needed
784 and indirect_return_needed instead of indirect_thunk_bnd_needed
785 and indirect_thunk_needed.
787 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
789 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
790 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
791 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
793 2018-03-15 David Malcolm <dmalcolm@redhat.com>
794 Paul Hua <paul.hua.gm@gmail.com>
797 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
799 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
801 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
802 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
805 2018-03-15 Tamar Christina <tamar.christina@arm.com>
808 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
809 instead of GET_MODE_SIZE when comparing Units.
811 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
814 * varasm.c (hash_section): Return an unchangeble hash value
815 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
816 Return !aarch64_can_use_per_function_literal_pools_p ().
818 2018-03-15 Jakub Jelinek <jakub@redhat.com>
821 * optabs.c (emit_conditional_move): Pass address of cmode's copy
822 rather than address of cmode as last argument to prepare_cmp_insn.
824 2018-03-15 Julia Koval <julia.koval@intel.com>
826 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
827 F_AVX512VNNI, F_AVX512BITALG): New.
829 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
832 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
833 insn for floating-point loads and stores.
835 2018-03-14 Carl Love <cel@us.ibm.com>
837 * config/rs6000/rs6000-c.c: Add macro definitions for
838 ALTIVEC_BUILTIN_VEC_PERMXOR.
839 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
840 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
841 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
842 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
844 * config/doc/extend.texi: Add prototypes for vec_permxor.
846 2018-03-14 David Malcolm <dmalcolm@redhat.com>
849 * diagnostic-show-locus.c (class layout_point): Convert m_line
850 from int to linenum_type.
851 (line_span::comparator): Use linenum "compare" function when
852 comparing line numbers.
853 (test_line_span): New function.
854 (layout_range::contains_point): Convert param "row" from int to
856 (layout_range::intersects_line_p): Likewise.
857 (layout::will_show_line_p): Likewise.
858 (layout::print_source_line): Likewise.
859 (layout::should_print_annotation_line_p): Likewise.
860 (layout::print_annotation_line): Likewise.
861 (layout::print_leading_fixits): Likewise.
862 (layout::annotation_line_showed_range_p): Likewise.
863 (struct line_corrections): Likewise for field m_row.
864 (line_corrections::line_corrections): Likewise for param "row".
865 (layout::print_trailing_fixits): Likewise.
866 (layout::get_state_at_point): Likewise.
867 (layout::get_x_bound_for_row): Likewise.
868 (layout::print_line): Likewise.
869 (diagnostic_show_locus): Likewise for locals "last_line" and
871 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
872 * input.c (selftest::test_linenum_comparisons): New function.
873 (selftest::input_c_tests): Call it.
874 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
875 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
876 * selftest.h (ASSERT_GT): New macro.
877 (ASSERT_GT_AT): New macro.
878 (ASSERT_LT): New macro.
879 (ASSERT_LT_AT): New macro.
881 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
883 PR rtl-optimization/84780
884 * combine.c (distribute_links): Don't make a link based on pc_rtx.
886 2018-03-14 Martin Liska <mliska@suse.cz>
888 * tree.c (record_node_allocation_statistics): Use
890 (get_stats_node_kind): New function extracted from
891 record_node_allocation_statistics.
892 (free_node): Use get_stats_node_kind.
894 2018-03-14 Richard Biener <rguenther@suse.de>
896 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
897 that the value-set of ANTIC_IN doesn't grow.
900 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
902 (BB_VISITED_WITH_VISITED_SUCCS): New define.
903 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
905 2018-03-14 Julia Koval <julia.koval@intel.com>
907 * config.gcc (icelake-client, icelake-server): New.
909 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
910 (initial_ix86_arch_features): Ditto.
911 (PTA_SKYLAKE): Add SGX.
912 (PTA_ICELAKE): Remove.
913 (PTA_ICELAKE_CLIENT): New.
914 (PTA_ICELAKE_SERVER): New.
915 (ix86_option_override_internal): Split up icelake on icelake client and
917 (get_builtin_code_for_version): Ditto.
918 (fold_builtin_cpu): Ditto.
919 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
920 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
921 * config/i386/i386.h (processor_type): Ditto.
922 * doc/invoke.texi: Ditto.
924 2018-03-14 Jakub Jelinek <jakub@redhat.com>
927 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
928 INTEGER_CST offset, add it together with bitpos / 8 and
929 sign extend based on POINTER_SIZE.
933 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
936 * config/i386/constraints.md (Yc): New register constraint.
937 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
938 Use Yc constraint for alternative 2 of operand 0. Remove
939 preferred_for_speed attribute.
941 2018-03-14 Richard Biener <rguenther@suse.de>
943 PR tree-optimization/84830
944 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
945 with the old one to avoid oscillations.
947 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
950 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
952 (assign_by_spills): Return a flag of reload assignment failure.
953 Do not process the reload assignment failures. Do not spill other
954 reload pseudos if they has the same reg class. Update n if
956 (lra_assign): Add a return arg. Set up from the result of
957 assign_by_spills call.
958 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
959 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
960 usage_insns if it is not NULL.
961 (spill_hard_reg_in_range): New function.
962 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
963 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
965 (lra_assign): Change prototype.
966 * lra.c (lra): Add code to deal with fails by splitting hard reg
969 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
971 * config/riscv/riscv.opt (mrelax): New option.
972 * config/riscv/riscv.c (riscv_file_start): Emit ".option
973 "norelax" when riscv_mrelax is disabled.
974 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
976 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
979 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
980 reassociation for int modes.
982 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
984 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
985 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
987 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
988 * config/aarch64/aarch64-sve.md
989 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
990 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
991 (*extend<mode><Vwide>2): Rename to...
992 (aarch64_sve_extend<mode><Vwide>2): ...this.
993 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
994 renaming the old pattern to...
995 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
997 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
998 define_expand, renaming the old pattern to...
999 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
1000 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
1001 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
1002 account when deciding which SVE instruction the optab should use.
1003 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
1005 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1007 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
1008 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
1009 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
1010 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
1011 (tlsdesc_small_<mode>): Turn a define_expand and use
1012 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
1013 (tlsdesc_small_advsimd_<mode>): ...this.
1014 (tlsdesc_small_sve_<mode>): New pattern.
1016 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1018 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
1019 (UNSPEC_UMUL_HIGHPART): New constants.
1020 (MUL_HIGHPART): New int iteraor.
1021 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
1022 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
1024 (*<su>mul<mode>3_highpart): New define_insn.
1026 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1029 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1032 2018-03-13 Martin Liska <mliska@suse.cz>
1035 * (sem_item_optimizer::sem_item_optimizer): Initialize new
1037 (sem_item_optimizer::~sem_item_optimizer): Release it.
1038 (sem_item_optimizer::merge_classes): Register variable aliases.
1039 (sem_item_optimizer::fixup_pt_set): New function.
1040 (sem_item_optimizer::fixup_points_to_sets): Likewise.
1041 * ipa-icf.h: Declare new variables and functions.
1043 2018-03-13 Jakub Jelinek <jakub@redhat.com>
1046 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
1047 integer_pow2p@2 and test integer_pow2p in condition.
1048 (A < 0 ? C : 0): Similarly for @1.
1051 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
1052 characters starting at p contain '\0' character, don't look beyond
1056 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
1057 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
1060 * reg-stack.c (change_stack): Change update_end var from int to
1061 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
1062 also call set_block_for_insn on the newly added insns and rescan.
1065 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
1066 on the last operand.
1069 * tree.c (stabilize_reference_1): Return save_expr (e) for
1070 STATEMENT_LIST even if it doesn't have side-effects.
1072 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
1074 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
1076 2018-03-12 Renlin Li <renlin.li@arm.com>
1078 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
1079 aarch64_output_scalar_simd_mov_immediate.
1081 2018-03-12 Martin Sebor <msebor@redhat.com>
1083 PR tree-optimization/83456
1084 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
1085 for perfectly overlapping calls to memcpy.
1086 (gimple_fold_builtin_memory_chk): Same.
1087 (gimple_fold_builtin_strcpy): Handle no-warning.
1088 (gimple_fold_builtin_stxcpy_chk): Same.
1089 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
1091 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1093 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
1094 parameter. Use it for SFmode.
1095 (rs6000_function_arg_advance_1): Adjust.
1096 (rs6000_function_arg): Adjust.
1097 (rs6000_gimplify_va_arg): Pass false for that new parameter.
1099 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1101 PR rtl-optimization/84169
1102 PR rtl-optimization/84780
1103 * combine.c (can_combine_p): Check for a 2-insn combination whether
1104 the destination register is used between the two insns, too.
1106 2018-03-12 Richard Biener <rguenther@suse.de>
1108 PR tree-optimization/84803
1109 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
1110 for refs DR analysis didn't process.
1112 2018-03-12 Richard Biener <rguenther@suse.de>
1114 PR tree-optimization/84777
1115 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
1116 force-vectorize loops ignore whether we are optimizing for size.
1118 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
1120 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
1121 (TARGET_MD_ASM_ADJUST): Define.
1123 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
1124 Kito Cheng <kito.cheng@gmail.com>
1125 Chung-Ju Wu <jasonwucj@gmail.com>
1127 * config/nds32/nds32.c (nds32_compute_stack_frame,
1128 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
1129 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
1130 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
1131 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
1132 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
1133 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
1134 * config/nds32/nds32.md (prologue, epilogue): Use macro
1135 NDS32_V3PUSH_AVAILABLE_P to do checking.
1137 2018-03-11 Jakub Jelinek <jakub@redhat.com>
1140 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
1141 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
1142 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
1143 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
1144 addition of most attributes on !orig_type_die or the attribute not
1145 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
1147 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1148 Chung-Ju Wu <jasonwucj@gmail.com>
1150 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
1152 * config/nds32/nds32.opt (mvh): New option.
1154 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1155 Chung-Ju Wu <jasonwucj@gmail.com>
1157 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
1159 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
1160 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
1163 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1164 Chung-Ju Wu <jasonwucj@gmail.com>
1166 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
1168 * config/nds32/nds32-multiple.md (strlensi): New pattern.
1169 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
1171 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
1172 Kito Cheng <kito.cheng@gmail.com>
1173 Chung-Ju Wu <jasonwucj@gmail.com>
1175 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
1176 UNSPEC_FFMISM and UNSPEC_FLMISM.
1177 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
1178 for ffb, ffmism and flmism.
1179 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
1180 (unspec_ffmism): Ditto.
1181 (unspec_flmism): Ditto.
1182 (nds32_expand_builtin_impl): Check if string extension is available.
1183 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
1184 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
1186 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
1189 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1192 * lra-assigns.c (assign_by_spills): Return a flag of reload
1193 assignment failure. Do not process the reload assignment
1194 failures. Do not spill other reload pseudos if they has the same
1196 (lra_assign): Add a return arg. Set up from the result of
1197 assign_by_spills call.
1198 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1199 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1200 usage_insns if it is not NULL.
1201 (spill_hard_reg_in_range): New function.
1202 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1203 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1204 function prototypes.
1205 (lra_assign): Change prototype.
1206 * lra.c (lra): Add code to deal with fails by splitting hard reg
1209 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
1212 * config/i386/i386.opt: Replace Enforcment with Enforcement.
1214 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
1217 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
1218 (dw_val_node): Add val_symbolic_view.
1219 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
1220 (symview_upper_bound): New.
1221 (new_line_info_table): Initialize symviews_since_reset.
1222 (dwarf2out_source_line): Count symviews_since_reset and set
1223 symview_upper_bound.
1224 (dw_val_equal_p): Handle symview.
1225 (add_AT_symview): New.
1226 (print_dw_val): Handle symview.
1227 (attr_checksum, attr_checksum_ordered): Likewise.
1228 (same_dw_val_p, size_of_die): Likewise.
1229 (value_format, output_die): Likewise.
1230 (add_high_low_attributes): Use add_AT_symview for entry_view.
1231 (dwarf2out_finish): Reset symview_upper_bound, clear
1234 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
1237 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
1238 Add strict argument and use it.
1239 (rs6000_split_multireg_move): Update for new strict argument.
1240 (mem_operand_gpr): Disallow all non-offsettable addresses.
1241 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
1243 2018-03-09 Jakub Jelinek <jakub@redhat.com>
1246 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
1247 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
1248 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
1251 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
1252 decl, use remap_type if we want to use the type.
1254 2018-03-09 Martin Sebor <msebor@redhat.com>
1256 PR tree-optimization/84526
1257 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1259 (builtin_access::generic_overlap): Be prepared to handle non-array
1262 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
1264 PR rtl-optimization/84682
1265 * lra-constraints.c (process_address_1): Check is_address flag
1266 for address constraints.
1267 (process_alt_operands): Likewise.
1268 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
1269 preprocess_constraints.
1270 * recog.h (preprocess_constraints): Add oploc parameter.
1272 * recog.c (preprocess_constraints): Test address_operand for
1273 CT_ADDRESS constraints.
1275 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1278 * lra-assigns.c (assign_by_spills): Return a flag of reload
1279 assignment failure. Do not process the reload assignment
1280 failures. Do not spill other reload pseudos if they has the same
1282 (lra_assign): Add a return arg. Set up from the result of
1283 assign_by_spills call.
1284 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1285 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1286 usage_insns if it is not NULL.
1287 (spill_hard_reg_in_range): New function.
1288 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1289 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1290 function prototypes.
1291 (lra_assign): Change prototype.
1292 * lra.c (lra): Add code to deal with fails by splitting hard reg
1295 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1298 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
1299 Accept complain bool parameter. Only emit errors if it is true.
1300 (arm_parse_cpu_option_name): Likewise.
1301 (arm_target_thumb_only): Adjust callers of the above.
1302 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
1303 prototype to take a default true bool parameter.
1304 (arm_parse_arch_option_name): Likewise.
1306 2018-03-09 David Malcolm <dmalcolm@redhat.com>
1307 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
1311 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
1312 * configure: Regenerate.
1313 * configure.ac ("linker --version-script option"): New.
1314 ("linker soname option"): New.
1316 2018-03-09 Richard Biener <rguenther@suse.de>
1318 PR tree-optimization/84775
1319 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
1320 immediate uses of predicate stmts and mark them modified.
1323 PR tree-optimization/84178
1324 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1326 (version_loop_for_if_conversion): Delay update_ssa call.
1327 (tree_if_conversion): Delay update_ssa until after predicate
1330 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
1333 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
1334 when the function accesses prior frames.
1336 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1339 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
1340 gen_llsym, otherwise call maybe_gen_llsym.
1343 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
1344 has ',' character inside of it.
1346 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1349 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
1350 as clobbering CC_REGNUM.
1352 2018-03-08 Richard Biener <rguenther@suse.de>
1355 * tree-scalar-evolution.c: Include tree-into-ssa.h.
1356 (follow_copies_to_constant): Do not follow SSA names registered
1359 2018-03-08 Richard Biener <rguenther@suse.de>
1361 PR tree-optimization/84178
1362 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1364 (version_loop_for_if_conversion): Delay update_ssa call.
1365 (tree_if_conversion): Delay update_ssa until after predicate
1368 2018-03-08 David Malcolm <dmalcolm@redhat.com>
1370 PR tree-optimization/84178
1371 * tree-if-conv.c (release_bb_predicate): Remove the
1372 the assertion that the stmts have NULL use_ops.
1373 Discard the statements, asserting that they haven't
1374 yet been added to a BB.
1376 2018-03-08 Richard Biener <rguenther@suse.de>
1378 PR tree-optimization/84746
1379 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
1380 (phi_translate): Pass in destination ANTIC_OUT set.
1381 (phi_translate_1): Likewise. For a simplified result lookup
1382 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
1383 (phi_translate_set): Adjust.
1384 (do_pre_regular_insertion): Likewise.
1385 (do_pre_partial_partial_insertion): Likewise.
1387 2018-03-08 Martin Liska <mliska@suse.cz>
1389 PR gcov-profile/84735
1390 * doc/gcov.texi: Document usage of profile files.
1391 * gcov-io.h: Document changes in the format.
1393 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
1397 * dwarf2out.c (struct dw_line_info_table): Update comments for
1399 (FORCE_RESET_NEXT_VIEW): New.
1400 (FORCE_RESETTING_VIEW_P): New.
1401 (RESETTING_VIEW_P): Check for -1 too.
1402 (ZERO_VIEW_P): Likewise.
1403 (new_line_info_table): Force-reset next view.
1404 (dwarf2out_begin_function): Likewise.
1405 (dwarf2out_source_line): Simplify zero_view_p initialization.
1406 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
1407 view directly. Omit view when omitting .loc at line 0.
1409 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1411 PR tree-optimization/84740
1412 * tree-switch-conversion.c (process_switch): Call build_constructors
1413 only if info.phi_count is non-zero.
1415 PR tree-optimization/84739
1416 * tree-tailcall.c (find_tail_calls): Check call arguments against
1417 DECL_ARGUMENTS (current_function_decl) rather than
1418 DECL_ARGUMENTS (func) when checking for tail recursion.
1420 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1422 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1423 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1424 Volker Reichelt's entry and add entries for people that perform
1425 GCC fuzzy testing and report numerous bugs.
1427 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1430 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
1431 readonly data in sdata, if that is disabled.
1432 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
1433 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1434 -mreadonly-in-sdata option.
1436 2018-03-07 Martin Sebor <msebor@redhat.com>
1438 PR tree-optimization/84468
1439 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
1440 basic block when looking for nul assignment.
1442 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
1445 * except.h (output_function_exception_table): Adjust prototype.
1446 * except.c (output_function_exception_table): Remove FNNAME parameter
1447 and add SECTION parameter. Ouput one part of the table at a time.
1448 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
1449 the first part of the exception table and emit unwind directives.
1450 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
1451 (i386_pe_seh_cold_init): Likewise.
1452 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1453 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1454 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1455 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1456 * config/i386/winnt.c: Include except.h.
1457 (struct seh_frame_state): Add reg_offset, after_prologue and
1458 in_cold_section fields.
1459 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1460 (i386_pe_seh_cold_init): New function.
1461 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1462 to seh->in_cold_section.
1463 (seh_emit_push): Record the offset of the push.
1464 (seh_emit_save): Record the offet of the save.
1465 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1466 Test seh->after_prologue to disregard the epilogue.
1467 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1468 (i386_pe_end_cold_function): New function.
1470 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1473 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1474 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1477 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1479 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1480 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1483 * multiple_target.c: Include tree-inline.h and intl.h.
1484 (expand_target_clones): Diagnose and fail if node->definition and
1485 !tree_versionable_function_p (node->decl).
1487 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1489 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1491 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1492 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1493 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1495 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1498 * combine.c (try_combine): Use reg_or_subregno instead of handling
1499 just paradoxical SUBREGs and REGs.
1501 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1503 * config/arc/arc.c (arc_finalize_pic): Remove function.
1504 (arc_must_save_register): We use single base PIC register, remove
1505 checks to save/restore the PIC register.
1506 (arc_expand_prologue): Likewise.
1507 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1509 (arc_verify_short): Likewise.
1510 (arc_attr_type): Likewise.
1511 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
1512 (walk_stores): Likewise.
1513 (arc_address_cost): Make it static.
1514 (arc_verify_short): Likewise.
1515 (branch_dest): Likewise.
1516 (arc_attr_type): Likewise.
1517 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
1518 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
1519 (arc_final_prescan_insn): Remove inserting the nops due to
1520 hardware hazards. It is done in reorg step.
1521 (insn_length_variant_t): Remove.
1522 (insn_length_parameters_t): Likewise.
1523 (arc_insn_length_parameters): Likewise.
1524 (arc_get_insn_variants): Likewise.
1525 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
1527 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1530 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
1533 PR tree-optimization/84687
1534 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
1536 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
1538 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1540 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
1541 Rename to ppc_speculation_barrier.
1542 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
1543 __builtin_ppc_speculation_barrier.
1545 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1548 * combine.c (combine_simplify_rtx): Don't try to simplify if
1549 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
1552 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
1554 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
1555 to 32 bytes when compiling for POWER9.
1557 2018-03-05 Jakub Jelinek <jakub@redhat.com>
1560 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
1561 regparm >= 3 with no arg reg available also for calls with
1562 flag_force_indirect_call. Pass decl to ix86_function_regparm.
1565 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
1567 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
1569 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
1572 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
1574 2018-03-05 Richard Biener <rguenther@suse.de>
1576 PR tree-optimization/84486
1577 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
1578 When inserting a __builtin_assume_aligned call set the LHS
1579 SSA name alignment info accordingly.
1581 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
1583 PR tree-optimization/84114
1584 * config/aarch64/aarch64.c (aarch64_reassociation_width)
1585 Avoid reassociation of FLOAT_MODE addition.
1587 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
1589 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
1590 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
1591 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
1592 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
1593 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
1594 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
1595 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
1597 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
1598 __builtin_ia32_wbinvd): New builtins.
1599 (SPECIAL_ARGS2): New.
1600 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
1601 (SPECIAL_ARGS2): New.
1602 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
1603 (ix86_valid_target_attribute_inner_p): Ditto.
1604 (ix86_init_mmx_sse_builtins): Add special_args2.
1605 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
1606 TARGET_WBNOINVD_P): New.
1607 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
1608 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
1609 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
1610 * config/i386/immintrin.h (_wbinvd): New intrinsic.
1611 * config/i386/pconfigintrin.h: New file.
1612 * config/i386/wbnoinvdintrin.h: Ditto.
1613 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
1614 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
1616 2018-03-05 Richard Biener <rguenther@suse.de>
1618 PR tree-optimization/84670
1619 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1621 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1622 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1623 (compute_antic_aux): Only assert the number of values in ANTIC_IN
1624 doesn't grow if all successors (recursively) were visited at least
1627 2018-03-05 Richard Biener <rguenther@suse.de>
1629 PR tree-optimization/84650
1630 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
1631 if executed in the loop pipeline.
1633 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
1635 * doc/configfiles.texi (Configuration Files): Move info about
1636 conditionalizing $target-protos.h to...
1637 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
1638 differs from $target-protos.h.
1640 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
1641 Chung-Ju Wu <jasonwucj@gmail.com>
1643 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
1644 * config/nds32/nds32-multiple.md (setmemsi): Define.
1645 * config/nds32/nds32-memory-manipulation.c
1646 (nds32_gen_dup_4_byte_to_word_value): New.
1647 (emit_setmem_word_loop): New.
1648 (emit_setmem_byte_loop): New.
1649 (nds32_expand_setmem_loop): New.
1650 (nds32_expand_setmem_loop_v3m): New.
1651 (nds32_expand_setmem_unroll): New.
1652 (nds32_expand_setmem): New.
1654 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1655 Chung-Ju Wu <jasonwucj@gmail.com>
1657 * config/nds32/nds32-memory-manipulation.c
1658 (nds32_emit_load_store): New.
1659 (nds32_emit_post_inc_load_store): New.
1660 (nds32_emit_mem_move): New.
1661 (nds32_emit_mem_move_block): New.
1662 (nds32_expand_movmemsi_loop_unknown_size): New.
1663 (nds32_expand_movmemsi_loop_known_size): New.
1664 (nds32_expand_movmemsi_loop): New.
1665 (nds32_expand_movmemsi_unroll): New.
1666 (nds32_expand_movmemqi): Rename ...
1667 (nds32_expand_movmemsi): ... to this.
1668 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
1669 (movmemsi): ... to this.
1670 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
1671 (nds32_expand_movmemsi): ... to this.
1673 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1674 Monk Chiang <sh.chiang04@gmail.com>
1675 Chung-Ju Wu <jasonwucj@gmail.com>
1677 * config/nds32/nds32-protos.h
1678 (nds32_expand_load_multiple): New arguments.
1679 (nds32_expand_store_multiple): Ditto.
1680 (nds32_valid_multiple_load_store): Rename ...
1681 (nds32_valid_multiple_load_store_p): ... to this.
1682 * config/nds32/nds32-memory-manipulation.c
1683 (nds32_expand_load_multiple): Refine implementation.
1684 (nds32_expand_store_multiple): Ditto.
1685 * config/nds32/nds32-multiple.md
1686 (load_multiple): Update nds32_expand_load_multiple interface.
1687 (store_multiple): Update nds32_expand_store_multiple interface.
1688 * config/nds32/nds32-predicates.c
1689 (nds32_valid_multiple_load_store): Rename ...
1690 (nds32_valid_multiple_load_store_p): ... to this and refine
1692 * config/nds32/predicates.md
1693 (nds32_load_multiple_and_update_address_operation): New predicate.
1694 (nds32_store_multiple_and_update_address_operation): New predicate.
1696 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
1697 Chung-Ju Wu <jasonwucj@gmail.com>
1699 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
1700 (combo): New attribute.
1701 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
1703 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
1705 * config/nds32/nds32.opt: Change -mcmodel= default value.
1707 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
1708 Monk Chiang <sh.chiang04@gmail.com>
1709 Chung-Ju Wu <jasonwucj@gmail.com>
1711 * config/nds32/constants.md (unspec_element): New enum.
1712 * config/nds32/constraints.md (Umw): New constraint.
1713 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
1714 * config/nds32/nds32-intrinsic.md: Likewise.
1715 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
1716 (nds32_valid_smw_lwm_base_p): New.
1717 (nds32_output_smw_single_word): New.
1718 (nds32_output_lmw_single_word): New.
1719 (nds32_expand_unaligned_load): New.
1720 (nds32_expand_unaligned_store): New.
1721 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
1722 (nds32_output_smw_single_word): Declare.
1723 (nds32_output_lmw_single_word): Declare.
1724 (nds32_expand_unaligned_load): Declare.
1725 (nds32_expand_unaligned_store): Declare.
1726 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
1727 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
1728 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
1729 NDS32_BUILTIN_UASTORE_DW.
1730 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
1733 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
1734 Kito Cheng <kito.cheng@gmail.com>
1735 Chung-Ju Wu <jasonwucj@gmail.com>
1737 * config/nds32/nds32-intrinsic.c
1738 (nds32_expand_builtin_null_ftype_reg): Delete.
1739 (nds32_expand_builtin_reg_ftype_imm): Ditto.
1740 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
1741 (nds32_read_argument): New.
1742 (nds32_legitimize_target): Ditto.
1743 (nds32_legitimize_argument): Ditto.
1744 (nds32_check_constant_argument): Ditto.
1745 (nds32_expand_unop_builtin): Ditto.
1746 (nds32_expand_unopimm_builtin): Ditto.
1747 (nds32_expand_binop_builtin): Ditto.
1748 (nds32_builtin_decl_impl): Ditto.
1749 (builtin_description): Ditto.
1750 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
1751 (nds32_init_builtins_impl): Ditto.
1752 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
1753 (nds32_builtin_decl): New.
1754 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
1755 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
1757 2018-03-02 Jeff Law <law@redhat.com>
1759 * reorg.c (stop_search_p): Handle DEBUG_INSN.
1760 (redundant_insn, fill_simple_delay_slots): Likewise.
1761 (fill_slots_from_thread): Likewise.
1762 * resource.c (mark_referenced_resources): Likewise.
1763 (mark_set_resources, find_dead_or_set_registers): Likewise.
1765 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1767 * substring-locations.h (format_warning_va): Formatting fix for
1769 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
1771 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
1772 * substring-locations.c: Include intl.h.
1773 (format_warning_va): Turned into small wrapper around
1774 format_warning_n_va, renamed to ...
1775 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
1776 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
1778 (format_warning_at_substring_n): New function.
1779 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
1780 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
1781 format_warning_at_substring with just a shorter name instead of
1782 const function pointer.
1783 (fmtwarn_n): New function.
1784 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
1785 appropriate, get rid of all the fmtstr temporaries, move conditionals
1786 with G_() wrapped string literals directly into fmtwarn arguments,
1787 cast dir.len to (int), formatting fixes.
1789 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
1791 * doc/invoke.texi: Remove "Cilk Plus" references.
1793 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1794 Richard Biener <rguenther@suse.de>
1797 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
1798 for error or warning attributes if CALL_FROM_THUNK_P is set.
1801 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1804 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
1805 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
1808 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
1809 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
1811 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
1812 instead of -1U in last predictors element's probability member.
1814 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1817 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1818 arguments if they are comparable.
1820 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1822 PR tree-optimization/84634
1823 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
1824 masks and masked_loop_p with a single loop_masks, making sure it's
1825 null for bb vectorization.
1827 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
1829 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1830 (vect_analyze_data_ref_access): Use loop->safe_len rather than
1831 loop->force_vectorize to check whether there is no alias.
1833 2018-03-02 Jakub Jelinek <jakub@redhat.com>
1836 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
1838 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
1840 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
1841 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
1842 instead of a loop around prev_real_insn.
1843 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
1847 * config/i386/i386.c (ix86_print_operand): Use conditional
1848 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
1851 2018-03-02 Richard Biener <rguenther@suse.de>
1853 PR tree-optimization/84427
1854 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
1855 (bitmap_set_subtract_values): Rewrite to handle multiple
1858 (prune_clobbered_mems): Likewise.
1859 (phi_translate): Take edge instead of pred/phiblock.
1860 (phi_translate_1): Likewise.
1861 (phi_translate_set): Likewise. Insert all translated
1862 exprs for a value into the set, keeping possibly multiple
1863 expressions per value.
1864 (compute_antic_aux): Adjust for phi_translate changes.
1865 When intersecting union the expressions and prune those
1866 not in the final value set, keeping possibly multiple
1867 expressions per value. Do not use value-insertion
1868 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
1869 all expressions. Add verification that the value-sets
1870 only shrink during iteration.
1871 (compute_partial_antic_aux): Adjust for the phi_translate changes.
1872 (do_pre_regular_insertion): Likewise.
1873 (do_pre_partial_partial_insertion): Likewise.
1875 2018-03-02 Richard Biener <rguenther@suse.de>
1878 * config/darwin.c (saved_debug_info_level): New static global.
1879 (darwin_asm_lto_start): Disable debug info generation for LTO out.
1880 (darwin_asm_lto_end): Restore debug info generation settings.
1882 2018-03-01 Martin Liska <mliska@suse.cz>
1885 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
1888 2018-03-01 Richard Biener <rguenther@suse.de>
1891 * dwarf2out.c (gen_variable_die): Properly handle late VLA
1892 type annotation with LTO when debug was disabled at compile-time.
1894 2018-03-01 Matthew Fortune <mfortune@gmail.com>
1896 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
1898 (mips_final_postscan_insn): Likewise.
1900 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
1902 PR rtl-optimization/84528
1903 * alias.c (init_alias_target): Add commentary.
1904 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
1905 a unique base value if the frame pointer is not eliminated
1906 to the stack pointer.
1908 2018-03-01 Tom de Vries <tom@codesourcery.com>
1910 PR rtl-optimization/83327
1911 * lra-int.h (hard_regs_spilled_into): Declare.
1912 * lra.c (hard_regs_spilled_into): Define.
1913 (init_reg_info): Init hard_regs_spilled_into.
1914 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
1915 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
1916 (process_bb_lives): Handle hard_regs_spilled_into.
1917 (lra_create_live_ranges_1): Before doing liveness propagation, clear
1918 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
1920 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
1922 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
1923 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
1924 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
1925 * config/rs6000/aix72.h: New file.
1927 2018-02-28 Jakub Jelinek <jakub@redhat.com>
1929 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
1930 instead of warning_at with conditional singular and plural messages
1934 * stor-layout.c (update_alignment_for_field): For
1935 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
1936 && !DECL_PACKED (field), do the alignment update, just use
1937 only desired_align instead of MAX (type_align, desired_align)
1939 (place_field): Don't do known_align < desired_align handling
1940 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
1941 is non-NULL, instead do it after rli->prev_field handling and
1942 only if not within a bitfield word. For DECL_PACKED (field)
1943 use type_align of BITS_PER_UNIT.
1945 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
1947 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
1948 superfluous parentheses and trailing spaces.
1950 2018-02-28 Richard Biener <rguenther@suse.de>
1952 PR tree-optimization/84584
1953 * graphite-scop-detection.c (scop_detection::add_scop): Discard
1954 SCoPs with fake exit edge.
1956 2018-02-28 Martin Liska <mliska@suse.cz>
1959 * timevar.c (timer::print): Fix format to properly print 100%
1962 2018-02-28 Richard Biener <rguenther@suse.de>
1965 * genmatch.c (capture_info::walk_match): Do not mark
1966 captured expressions without operands as expr_p given
1967 they act more like predicates and should be subject to
1968 "lost tail" side-effect preserving.
1970 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
1972 PR rtl-optimization/81611
1973 * auto-inc-dec.c (attempt_change): Move dead note from
1974 mem_insn if it's the next use of regno
1975 (find_address): Take address use of reg holding
1976 non-incremented value. Add parm to limit search to the named
1978 (merge_in_block): Attempt to use a mem insn that is the next
1979 use of the original regno.
1981 2018-02-27 Martin Sebor <msebor@redhat.com>
1984 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
1985 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
1987 2018-02-27 Martin Sebor <msebor@redhat.com>
1989 PR translation/84207
1990 * diagnostic-core.h (warning_n, error_n, inform_n): Change
1991 n argument to unsigned HOST_WIDE_INT.
1992 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
1993 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
1994 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
1995 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
1997 2018-02-27 Richard Biener <rguenther@suse.de>
1999 PR tree-optimization/84512
2000 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2001 Do not use the estimate returned from record_stmt_cost for
2002 the scalar iteration cost but sum properly using add_stmt_cost.
2004 2018-02-27 Richard Biener <rguenther@suse.de>
2006 PR tree-optimization/84466
2007 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2008 Adjust last change to less strictly validate use operands.
2010 2018-02-27 Martin Liska <mliska@suse.cz>
2012 PR gcov-profile/84548
2013 * gcov.c (process_file): Allow partial overlap and consider it
2014 also as group functions.
2015 (output_lines): Properly calculate range of lines for a group.
2017 2018-02-27 Martin Liska <mliska@suse.cz>
2019 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
2020 'ggc' suffixes. Change first column width.
2021 (timer::print): Fix formatting of the column.
2023 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
2025 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
2026 preserve inline entry blocks for the sake of debug inline
2027 entry point markers alone.
2028 (remove_unused_locals): Suggest in comments a better place to
2029 force the preservation of inline entry blocks that are
2030 otherwise unused, but do not preserve them.
2032 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2034 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
2036 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2039 * config/i386/constraints.md (Bs): Replace
2040 ix86_indirect_branch_register with
2041 TARGET_INDIRECT_BRANCH_REGISTER.
2043 * config/i386/i386.md (indirect_jump): Likewise.
2044 (tablejump): Likewise.
2045 (*sibcall_memory): Likewise.
2046 (*sibcall_value_memory): Likewise.
2047 Peepholes of indirect call and jump via memory: Likewise.
2048 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
2049 (*sibcall_value_GOT_32): Likewise.
2050 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2051 (GOT_memory_operand): Likewise.
2052 (call_insn_operand): Likewise.
2053 (sibcall_insn_operand): Likewise.
2054 (GOT32_symbol_operand): Likewise.
2055 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
2057 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2059 PR rtl-optimization/83496
2060 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
2061 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
2062 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
2063 redundant insn, if any.
2064 (relax_delay_slots): Likewise.
2065 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
2067 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
2069 PR tree-optimization/83965
2070 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
2071 that grouped statements are part of a reduction chain. Return
2072 true if the statement is not marked as a reduction itself but
2074 (vect_recog_dot_prod_pattern): Don't check whether the statement
2075 is part of a group here.
2076 (vect_recog_sad_pattern): Likewise.
2077 (vect_recog_widen_sum_pattern): Likewise.
2079 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2082 * final.c (rest_of_clean_state): Also look for calls inside sequences.
2084 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2087 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
2089 (ix86_output_indirect_function_return): New prototype.
2090 (ix86_split_simple_return_pop_internal): Likewise.
2091 * config/i386/i386.c (indirect_return_via_cx): New.
2092 (indirect_return_via_cx_bnd): Likewise.
2093 (indirect_thunk_name): Handle return va CX_REG.
2094 (output_indirect_thunk_function): Create alias for
2095 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
2096 (ix86_output_indirect_jmp): Remove the bool argument.
2097 (ix86_output_indirect_function_return): New function.
2098 (ix86_split_simple_return_pop_internal): Likewise.
2099 * config/i386/i386.md (*indirect_jump): Don't pass false
2100 to ix86_output_indirect_jmp.
2101 (*tablejump_1): Likewise.
2102 (simple_return_pop_internal): Change it to define_insn_and_split.
2103 Call ix86_split_simple_return_pop_internal to split it for
2105 (simple_return_indirect_internal): Call
2106 ix86_output_indirect_function_return instead of
2107 ix86_output_indirect_jmp.
2109 2018-02-26 Jakub Jelinek <jakub@redhat.com>
2112 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
2113 memset and value initialization afterwards.
2115 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
2117 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
2119 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2122 * common/config/aarch64/aarch64-common.c
2123 (aarch_option_optimization_table[]): Switch
2124 off fomit-frame-pointer
2126 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2127 Chung-Ju Wu <jasonwucj@gmail.com>
2129 * config/nds32/nds32-multiple.md (load_multiple): Disallow
2131 (store_multiple): Ditto.
2133 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2135 * config.gcc: Add --with-cpu support for nds32 target.
2136 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
2137 * config/nds32/nds32.opt: Add -mcpu= option.
2139 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
2141 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
2142 isel=yes): Warn for these deprecated options.
2144 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
2146 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
2147 ISA_2_5_MASKS_EMBEDDED.
2149 2018-02-23 Jakub Jelinek <jakub@redhat.com>
2151 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
2152 p->max as pointers rather than using iterative_hash_expr.
2154 2018-02-23 Carl Love <cel@us.ibm.com>
2156 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
2157 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
2159 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
2160 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
2161 P8V_BUILTIN_VEC_VUNSIGNED2.
2163 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
2166 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
2167 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
2169 * lra-constraints.c (curr_insn_transform): Set up
2170 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
2172 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
2173 * lra-eliminations.c (spill_pseudos): Ditto.
2174 (process_insn_for_elimination): Ditto.
2175 * lra-lives.c (reg_early_clobber_p): Use the new macros.
2176 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
2177 LRA_NON_CLOBBERED_ALT.
2179 2018-02-22 Martin Sebor <msebor@redhat.com>
2181 PR tree-optimization/84480
2182 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
2183 to maybe_diag_stxncpy_trunc. Call it.
2184 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
2185 from gimple_fold_builtin_strcpy. Print inlining stack.
2186 (handle_builtin_stxncpy): Print inlining stack.
2187 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
2189 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
2192 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
2193 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
2194 and -fcheck-pointer-bounds are used together.
2195 (indirect_thunk_prefix): New enum.
2196 (indirect_thunk_need_prefix): New function.
2197 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
2198 "_nt" instead of "_bnd" for NOTRACK prefix.
2199 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
2200 (output_indirect_thunk_function): Likewise.
2202 (ix86_code_end): Update output_indirect_thunk_function calls.
2203 (ix86_output_indirect_branch_via_reg): Replace
2204 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
2205 (ix86_output_indirect_branch_via_push): Likewise.
2206 (ix86_output_function_return): Likewise.
2207 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
2208 incompatible with -fcf-protection=branch and
2209 -fcheck-pointer-bounds.
2211 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2214 * config/aarch64/aarch64.c (aarch64_print_address_internal):
2215 Change gcc_assert call to output_operand_lossage.
2217 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2219 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
2221 2018-02-22 DJ Delorie <dj@redhat.com>
2222 Sebastian Perta <sebastian.perta@renesas.com>
2223 Oleg Endo <olegendo@gcc.gnu.org>
2225 * config/rx/rx.c (rx_rtx_costs): New function.
2226 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
2228 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
2230 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
2232 2018-02-22 Martin Liska <mliska@suse.cz>
2235 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2236 Add "native" as a possible value.
2238 2018-02-22 Martin Liska <mliska@suse.cz>
2241 * config/i386/i386.c (ix86_option_override_internal):
2242 Add "native" as a possible value for -march and -mtune.
2244 2018-02-22 Jakub Jelinek <jakub@redhat.com>
2247 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
2248 to all type variants.
2250 PR tree-optimization/84503
2251 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
2252 width as info->bitpos + info->bitsize - start.
2253 (merged_store_group::merge_overlapping): Simplify width computation.
2254 (check_no_overlap): New function.
2255 (imm_store_chain_info::try_coalesce_bswap): Compute expected
2256 start + width and last_order of the group, fail if check_no_overlap
2258 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
2259 to group if check_no_overlap fails.
2261 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2263 * config/rs6000/altivec.md: Delete contraint arguments to
2264 define_expand, define_split, and define_peephole2, and in
2265 define_insn_and_split if always unused.
2266 * config/rs6000/darwin.md: Ditto.
2267 * config/rs6000/dfp.md: Ditto.
2268 * config/rs6000/rs6000.md: Ditto.
2269 * config/rs6000/sync.md: Ditto.
2270 * config/rs6000/vector.md: Ditto.
2271 * config/rs6000/vsx.md: Ditto.
2273 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2275 * config/rs6000/altivec.md: Write output control strings as braced
2276 blocks instead of double-quoted strings.
2277 * config/rs6000/darwin.md: Ditto.
2278 * config/rs6000/rs6000.md: Ditto.
2279 * config/rs6000/vector.md: Ditto.
2280 * config/rs6000/vsx.md: Ditto.
2282 2018-02-21 Jason Merrill <jason@redhat.com>
2284 PR c++/84314 - ICE with templates and fastcall attribute.
2285 * attribs.c (build_type_attribute_qual_variant): Remove assert.
2287 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2289 * ipa-cp.c (determine_versionability): Fix comment typos.
2291 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2294 * ipa-cp.c (determine_versionability): Do not version functions caling
2297 2018-02-21 Martin Liska <mliska@suse.cz>
2300 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
2301 Add "native" as a possible value.
2302 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
2303 the macro when native cpu detection is available.
2305 2018-02-21 Martin Liska <mliska@suse.cz>
2308 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
2309 Add "native" as a possible value.
2310 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
2311 when native cpu detection is available.
2313 2018-02-21 Jakub Jelinek <jakub@redhat.com>
2314 Martin Sebor <msebor@redhat.com>
2316 PR tree-optimization/84478
2317 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
2319 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
2320 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
2321 support which is conservatively correct, for 2 only stay conservative
2322 for maxlen. Formatting and comment capitalization fixes. Add STRICT
2323 argument to the 2 argument get_range_strlen, adjust 6 arg
2324 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
2326 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
2327 (gimple_fold_builtin_strlen): Pass true as last argument to
2330 2018-02-20 Martin Sebor <msebor@redhat.com>
2333 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
2334 (builtin_memref::set_base_and_offset): Same. Handle inner references.
2335 (builtin_memref::builtin_memref): Factor out parts into
2336 set_base_and_offset and call it.
2338 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2341 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
2342 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
2343 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
2344 search at the associated MODE_INT.
2346 2018-02-20 Jeff Law <law@redhat.com>
2349 PR tree-optimization/81592
2351 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
2352 for range data rather than using global data.
2353 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
2354 range data rather than using global data.
2355 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
2356 pass it to children as needed.
2357 (struct directive::fmtresult): Similarly.
2358 (struct directive::set_width): Similarly.
2359 (struct directive::set_precision): Similarly.
2360 (format_integer, format_directive, parse_directive): Similarly.
2361 (format_none): Accept unnamed vr_values parameter.
2362 (format_percent, format_floating, format_character): Similarly.
2363 (format_string, format_plain): Similarly.
2364 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
2365 the EVRP range analyzer for range data rather than using global data.
2366 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
2367 gimple-ssa-evrp-analyze.h
2368 (class sprintf_dom_walker): Add after_dom_children member function.
2369 Add evrp_range_analyzer member.
2370 (sprintf_dom_walker::before_dom_children): Call into the EVRP
2371 range analyzer as needed.
2372 (sprintf_dom_walker::after_dom_children): New member function.
2373 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
2375 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
2376 (evrp_range_analyzer::pop_to_marker): Likewise.
2378 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2380 PR tree-optimization/84419
2381 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
2382 with the required type if its current type is compatible but
2385 2018-02-20 Jakub Jelinek <jakub@redhat.com>
2388 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
2389 after vectorization.
2391 2018-02-20 Martin Liska <mliska@suse.cz>
2394 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
2395 possible values if we don't have a hint.
2397 2018-02-20 Martin Liska <mliska@suse.cz>
2401 * final.c (shorten_branches): Build align_tab array with one
2403 * opts.c (finish_options): Add alignment option limit check.
2404 (MAX_CODE_ALIGN): Likewise.
2405 (MAX_CODE_ALIGN_VALUE): Likewise.
2406 * doc/invoke.texi: Document maximum allowed option value for
2407 all -falign-* options.
2409 2018-02-19 Jakub Jelinek <jakub@redhat.com>
2412 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
2413 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
2414 * var-tracking.c (emit_note_insn_var_location): Remove all references
2415 to NOTE_INSN_CALL_ARG_LOCATION.
2416 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
2417 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
2418 Use copy_rtx_if_shared.
2419 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
2420 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
2421 (dwarf2out_var_location): Remove handling of
2422 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2424 * final.c (final_scan_insn): Remove all references to
2425 NOTE_INSN_CALL_ARG_LOCATION.
2426 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2427 before dumping final insns.
2428 * except.c (emit_note_eh_region_end): Remove all references to
2429 NOTE_INSN_CALL_ARG_LOCATION.
2430 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
2431 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
2432 * config/arc/arc.c (hwloop_optimize): Likewise.
2433 * config/arm/arm.c (create_fix_barrier): Likewise.
2434 * config/s390/s390.c (s390_chunkify_start): Likewise.
2435 * config/sh/sh.c (find_barrier): Likewise.
2436 * config/i386/i386.c (rest_of_insert_endbranch,
2437 ix86_seh_fixup_eh_fallthru): Likewise.
2438 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
2439 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
2440 * config/frv/frv.c (frv_function_prologue): Likewise.
2441 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
2443 (note_outside_basic_block_p): Remove all references to
2444 NOTE_INSN_CALL_ARG_LOCATION.
2445 * gengtype.c (adjust_field_rtx_def): Likewise.
2446 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
2448 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
2449 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
2452 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2455 PR tree-optimization/84452
2456 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2457 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2460 2018-02-19 Martin Liska <mliska@suse.cz>
2463 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2465 2018-02-19 Martin Liska <mliska@suse.cz>
2466 Richard Sandiford <richard.sandiford@linaro.org>
2468 PR tree-optimization/82491
2469 * gimple-fold.c (get_base_constructor): Make earlier bail out
2472 2018-02-19 Carl Love <cel@us.ibm.com>
2474 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2475 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2477 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2478 P8V_BUILTIN_VEC_NEG.
2480 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2482 * config/rl78/rl78.md (movdf): New define expand.
2484 2018-02-19 Martin Liska <mliska@suse.cz>
2487 * doc/invoke.texi: Fix typo.
2488 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2490 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2492 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2493 handle rs6000_single_float and rs6000_double_float specially for
2496 2018-02-16 Jeff Law <law@redhat.com>
2498 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2501 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2504 * ipa-inline.c (inline_small_functions): Fix a typo.
2506 2018-02-16 Nathan Sidwell <nathan@acm.org>
2508 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
2510 2018-02-16 Carl Love <cel@us.ibm.com>
2512 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
2513 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
2514 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
2515 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
2516 expansion to P8V_BUILTIN_VEC_FLOAT2.
2518 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
2520 PR rtl-optimization/70023
2521 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
2522 src_regno into account.
2524 2018-02-16 Carl Love <cel@us.ibm.com>
2526 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
2527 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
2528 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
2529 * config/rs6000/rs6000.c: Remove case statements for
2530 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
2531 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
2532 and P9V_BUILTIN_VEC_VINSERT4B.
2533 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
2534 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
2535 * config/rs6000/vsx.md:
2536 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
2539 2018-02-16 Carl Love <cel@us.ibm.com>
2541 * config/rs6000/altivec.h: Add builtin names vec_extract4b
2543 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
2545 * config/rs6000/rs6000-c.c: Add the definitions for
2546 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
2547 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
2548 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
2549 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
2550 definition for insert4b and define insn *insert3b_internal.
2551 * doc/extend.texi: Add documentation for vec_extract4b.
2553 2018-02-16 Nathan Sidwell <nathan@acm.org>
2555 * doc/extend.texi (Backwards Compatibility): Mention friend
2556 injection. Note for-scope is deprecated.
2557 * doc/invoke.texi (-ffriend-injection): Deprecate.
2559 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
2561 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
2562 that moved to I2, also allow destinations that are a paradoxical
2563 subreg (instead of a normal reg).
2565 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
2568 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
2571 2018-02-16 Richard Biener <rguenther@suse.de>
2573 PR tree-optimization/84037
2574 PR tree-optimization/84016
2576 * config/i386/i386.c (ix86_builtin_vectorization_cost):
2577 Adjust vec_construct for the fact we need additional higher latency
2578 128bit inserts for AVX256 and AVX512 vector builds.
2579 (ix86_add_stmt_cost): Scale vector construction cost for
2582 2018-02-16 Richard Biener <rguenther@suse.de>
2584 PR tree-optimization/84417
2585 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
2586 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
2587 (non_rewritable_lvalue_p): Likewise, use poly-ints.
2589 2018-02-16 Martin Liska <mliska@suse.cz>
2592 * internal-fn.def (ASAN_CHECK): Set proper flags.
2593 (ASAN_MARK): Likewise.
2595 2018-02-16 Julia Koval <julia.koval@intel.com>
2597 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
2598 from PTA_CANNONLAKE.
2600 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2603 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
2604 Use ++iter rather than iter++ for std::list iterators.
2605 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
2606 defer deleting them until all nodes in the forest are processed. Do
2607 free even leaf nodes. Change to_process into auto_vec.
2610 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
2611 * vec.h (vec_default_construct): Use memset instead of placement new
2612 if BROKEN_VALUE_INITIALIZATION is defined.
2613 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
2614 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
2617 PR rtl-optimization/83723
2618 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
2619 * lra.c (lra_substitute_pseudo): Likewise. If true, use
2620 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
2622 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
2624 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
2626 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
2628 PR rtl-optimization/81443
2629 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
2630 from inner REGs to paradoxical SUBREGs.
2632 2018-02-16 Richard Biener <rguenther@suse.de>
2634 PR tree-optimization/84399
2635 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2636 For operands we can analyze at their definition make sure we can
2637 analyze them at each use as well.
2639 2018-02-16 Richard Biener <rguenther@suse.de>
2641 PR tree-optimization/84190
2642 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
2643 volatile accesses if the decl isn't volatile.
2645 2018-02-15 Jason Merrill <jason@redhat.com>
2647 PR c++/84314 - ICE with templates and fastcall attribute.
2648 * attribs.c (build_type_attribute_qual_variant): Don't clobber
2649 TYPE_CANONICAL on an existing type.
2651 2018-02-15 Jakub Jelinek <jakub@redhat.com>
2653 PR tree-optimization/84383
2654 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
2655 dstoff nor call operand_equal_p if dstbase is NULL.
2657 PR tree-optimization/84334
2658 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
2659 also a CONSTANT_CLASS_P, punt.
2661 2018-02-14 Jim Wilson <jimw@sifive.com>
2663 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
2664 first SMALL_OPERAND check. New local min_second_step. Move assert
2665 to where locals are set. Add TARGET_RVC support.
2666 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
2668 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
2670 * doc/invoke.texi: Correct -Wformat-overflow code sample.
2672 2018-02-14 Martin Sebor <msebor@redhat.com>
2674 PR tree-optimization/83698
2675 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
2676 arrays constrain the offset range to their bounds.
2677 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
2678 (builtin_access::overlap): Avoid setting the size of overlap if it's
2680 (maybe_diag_overlap): Also consider arrays when deciding what values
2681 of offsets to include in diagnostics.
2683 2018-02-14 Martin Sebor <msebor@redhat.com>
2686 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
2687 that correspond to the kind of a declaration.
2689 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
2692 * config/pa/pa.md: Load address of PIC label using the linkage table
2693 if the label is nonlocal.
2695 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
2697 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
2698 warning message if user requests -maltivec=be.
2699 * doc/invoke.texi: Document deprecation of -maltivec=be.
2701 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
2704 * config/rs6000/rs6000-c.c: Update definitions for
2705 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
2706 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
2708 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2711 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
2712 add _get_ssp intrinsics. Remove argument from
2713 __builtin_ia32_rdssp[d|q].
2714 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
2715 * config/i386/i386-builtin.def: Remove argument from
2716 __builtin_ia32_rdssp[d|q].
2717 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
2718 ix86_expand_special_args_builtin for _rdssp[d|q].
2719 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
2720 Clear register before usage.
2721 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
2722 Add documentation for new _get_ssp and _inc_ssp intrinsics.
2724 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
2726 PR tree-optimization/84357
2727 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
2728 operand 1 of an ARRAY_REF too.
2730 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
2733 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
2734 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2736 (set_of_reg): New struct.
2737 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
2738 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
2739 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
2741 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
2742 Split into bitclr, bitset, bitinvert patterns if appropriate.
2743 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
2744 use rx_fuse_in_memory_bitop.
2745 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
2746 to named insn, correct maximum insn length.
2748 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
2751 * machmode.def: Define a complex mode for PARTIAL_INT.
2752 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
2754 * doc/rtl.texi: Document CSPImode.
2755 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
2757 (msp430_hard_regno_nregs_with_padding): Likewise.
2759 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
2762 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
2764 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
2766 PR rtl-optimization/84169
2767 * combine.c (try_combine): New variable split_i2i3. Set it to true if
2768 we generated a parallel as new i3 and we split that to new i2 and i3
2769 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
2770 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
2771 those to i2, not i1. Partially rewrite this scan code.
2773 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2776 * stor-layout.c (place_field): For variable length fields, adjust
2777 offset_align afterwards not just based on the field's alignment,
2778 but also on the size.
2781 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
2782 of exps and logs in the use_exp2 case.
2784 2018-02-13 Jeff Law <law@redhat.com>
2786 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
2789 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
2792 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
2796 * common.opt (gas-loc-support, gas-locview-support): New.
2797 (ginline-points, ginternal-reset-location-views): New.
2798 * doc/invoke.texi: Document them. Use @itemx where intended.
2799 (gvariable-location-views): Adjust.
2800 * target.def (reset_location_view): New.
2801 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
2802 (TARGET_RESET_LOCATION_VIEW): New.
2803 * doc/tm.texi: Rebuilt.
2804 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
2805 (dwarf2out_default_as_locview_support): New.
2806 (output_asm_line_debug_info): Use option variables.
2807 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
2808 (output_loc_list): Likewise.
2809 (add_high_low_attributes): Check option variables.
2810 Don't output entry view attribute in strict mode.
2811 (gen_inlined_subroutine_die): Check option variables.
2812 (dwarf2out_inline_entry): Likewise.
2813 (init_sections_and_labels): Likewise.
2814 (dwarf2out_early_finish): Likewise.
2815 (maybe_reset_location_view): New, from...
2816 (dwarf2out_var_location): ... here. Call it.
2817 * debug.h (dwarf2out_default_as_loc_support): Declare.
2818 (dwarf2out_default_as_locview_support): Declare.
2819 * hooks.c (hook_int_rtx_insn_0): New.
2820 * hooks.h (hook_int_rtx_insn_0): Declare.
2821 * toplev.c (process_options): Take -gas-loc-support and
2822 -gas-locview-support from dwarf2out. Enable
2823 -gvariable-location-views by default only with locview
2824 assembler support. Enable -ginternal-reset-location-views by
2825 default only if the target defines the corresponding hook.
2826 Enable -ginline-points by default if location views are
2827 enabled; force it disabled if statement frontiers are
2829 * tree-inline.c (expand_call_inline): Check option variables.
2830 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2832 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
2834 PR tree-optimization/84321
2835 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
2836 handling. Also check whether the anti-range contains any values
2837 that satisfy the mask; switch to a VR_RANGE if not.
2839 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
2842 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
2844 2018-02-13 Martin Jambor <mjambor@suse.cz>
2847 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
2848 of call statements, also set location of a load to a temporary.
2850 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
2852 * config/rl78/rl78.c (add_vector_labels): New function.
2853 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
2854 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
2855 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
2856 which checks that no arguments are passed.
2857 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
2858 * doc/extend.texi: Documentation for the new attribute.
2860 2018-02-13 Andreas Schwab <schwab@suse.de>
2862 * config/riscv/linux.h (CPP_SPEC): Define.
2864 2018-02-13 Jakub Jelinek <jakub@redhat.com>
2867 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
2868 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
2869 OPTION_MASK_ISA_AES as first argument to def_builtin_const
2870 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
2871 instead of OPTION_MASK_ISA_PCLMUL as first argument to
2872 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
2873 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
2874 temporarily for AES and PCLMUL builtins.
2876 PR tree-optimization/84339
2877 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
2878 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
2882 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
2883 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
2884 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
2886 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
2888 * omp-simd-clone.h: New file.
2889 * omp-simd-clone.c: Include omp-simd-clone.h.
2890 (expand_simd_clones): No longer static.
2891 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
2892 cgraph.h and omp-simd-clone.h.
2893 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
2894 (vect_recog_widen_shift_pattern): Formatting fix.
2895 (vect_pattern_recog_1): Don't check optab for calls.
2898 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
2899 operands[2] into a REG before using gen_lowpart on it.
2901 2018-02-12 Jeff Law <law@redhat.com>
2904 * config/sh/sh.c (find_barrier): Consider a sibling call
2907 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
2908 successfully back substituting a reg.
2910 2018-02-12 Richard Biener <rguenther@suse.de>
2912 PR tree-optimization/84037
2913 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
2914 parameter, move visited init to caller.
2915 (vect_slp_analyze_operations): Separate cost from validity
2916 check, initialize visited once for all instances.
2917 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
2919 * tree-vect-stmts.c (vect_model_simple_cost): Make early
2921 (vect_model_promotion_demotion_cost): Likewise.
2922 (vectorizable_bswap): Guard cost modeling with !slp_node
2923 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
2925 (vectorizable_call): Likewise.
2926 (vectorizable_conversion): Likewise.
2927 (vectorizable_assignment): Likewise.
2928 (vectorizable_shift): Likewise.
2929 (vectorizable_operation): Likewise.
2930 (vectorizable_store): Likewise.
2931 (vectorizable_load): Likewise.
2932 (vectorizable_condition): Likewise.
2933 (vectorizable_comparison): Likewise.
2935 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
2938 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
2939 (ASAN_MARK): Fix fnspec to account for return value, change pointer
2940 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
2942 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2945 * params.def (inline-min-speedup): Increase from 8 to 15.
2946 (max-inline-insns-auto): Decrease from 40 to 30.
2947 * ipa-split.c (consider_split): Add some buffer for function to
2948 be considered inlining candidate.
2949 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
2952 2018-02-12 Richard Biener <rguenther@suse.de>
2954 PR tree-optimization/84037
2955 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
2956 matched stmts if we cannot swap the non-matched ones.
2958 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2960 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
2961 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
2962 _mm_maskz_scalef_round_ss): New intrinsics.
2963 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
2964 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
2965 __builtin_ia32_scalefss_round): Remove.
2966 (__builtin_ia32_scalefsd_mask_round,
2967 __builtin_ia32_scalefss_mask_round): New intrinsics.
2968 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
2969 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
2970 ((match_operand:VF_128 2 "<round_nimm_predicate>"
2971 "<round_constraint>")): Changed to ...
2972 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
2973 "<round_scalar_constraint>")): ... this.
2974 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
2975 %0, %1, %2<round_op3>}"): Changed to ...
2976 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
2977 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
2978 %2<round_scalar_mask_op3>}"): ... this.
2979 * config/i386/subst.md (round_scalar_nimm_predicate): New.
2981 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
2983 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
2984 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
2985 (_mm_maskz_sqrt_round_ss): New intrinsics.
2986 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
2987 (__builtin_ia32_sqrtsd_mask_round)
2988 (__builtin_ia32_sqrtss_mask_round): New builtins.
2989 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
2990 (__builtin_ia32_sqrtss_round): Remove.
2991 (__builtin_ia32_sqrtsd_mask_round)
2992 (__builtin_ia32_sqrtss_mask_round): New builtins.
2993 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
2994 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
2995 ((match_operand:VF_128 1 "vector_operand"
2996 "xBm,<round_constraint>")): Changed to ...
2997 ((match_operand:VF_128 1 "vector_operand"
2998 "xBm,<round_scalar_constraint>")): ... this.
2999 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
3000 %0, %2, %<iptr>1<round_op3>}): Changed to ...
3001 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
3002 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
3003 %<iptr>1<round_scalar_mask_op3>}): ... this.
3004 ((set_attr "prefix" "<round_prefix>")): Changed to ...
3005 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
3007 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
3010 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
3011 Cast vec_cmpeq result to correct type.
3012 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
3013 Cast vec_cmpgt result to correct type.
3015 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
3017 * final.c (final_scan_insn_1): Renamed from...
3018 (final_scan_insn): ... this. New wrapper, to recover
3019 seen from the outermost call in recursive ones.
3020 * config/sparc/sparc.c (output_return): Drop seen from call.
3021 (output_sibcall): Likewise.
3022 * config/visium/visium.c (output_branch): Likewise.
3024 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
3026 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
3029 2018-02-10 Alan Modra <amodra@gmail.com>
3032 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
3033 Specify LR as an input.
3035 2018-02-10 Jakub Jelinek <jakub@redhat.com>
3038 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
3039 remove_member_access_dummy_vars): New functions.
3040 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
3041 lower_omp_1, execute_lower_omp): Use them.
3043 PR rtl-optimization/84308
3044 * shrink-wrap.c (spread_components): Release todo vector.
3046 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
3048 PR rtl-optimization/57193
3049 * ira-color.c (struct allocno_color_data): Add member
3050 conflict_allocno_hard_prefs.
3051 (update_conflict_allocno_hard_prefs): New.
3052 (bucket_allocno_compare_func): Add a preference based on
3053 conflict_allocno_hard_prefs.
3054 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
3055 (color_allocnos): Remove a dead code. Initiate
3056 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
3058 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3061 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
3062 constraint from =wa to wa. Avoid a subreg on the output operand,
3063 instead use a pseudo and subreg it in a move.
3064 (p9_xxbrd_<mode>): Changed to ...
3065 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
3066 (p9_xxbrd_v2df): New expander.
3067 (p9_xxbrw_<mode>): Changed to ...
3068 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
3069 (p9_xxbrw_v4sf): New expander.
3071 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3073 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
3075 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
3078 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
3079 multiply in 32-bit mode.
3080 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
3081 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
3084 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3086 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
3087 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
3088 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
3089 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
3091 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
3094 * dwarf2out.c (is_trivial_indirect_ref): New function.
3095 (dwarf2out_late_global_decl): Do not generate a location
3096 attribute for variables that have a non-trivial DECL_VALUE_EXPR
3097 and that are not defined in the current unit.
3099 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3101 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
3102 instead of a libcall for UNORDERED.
3104 2018-02-09 Tamar Christina <tamar.christina@arm.com>
3107 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
3108 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
3110 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3113 * config/s390/s390.c (s390_set_current_function): Invoke
3114 s390_indirect_branch_settings also if fndecl didn't change.
3116 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3118 * config/rs6000/rs6000.md (blockage): Set length to zero.
3120 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3122 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
3124 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3127 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
3128 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
3132 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
3133 PARALLEL incoming that failed vt_get_decl_and_offset check.
3136 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
3137 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
3139 (get_variable_section): For decls in named .bss* sections pass true as
3140 second argument to bss_initializer_p.
3142 2018-02-09 Marek Polacek <polacek@redhat.com>
3143 Jakub Jelinek <jakub@redhat.com>
3146 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
3147 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
3148 Sync some changes from cxx_fold_indirect_ref.
3150 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3152 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
3154 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
3155 (BLOCK_INLINE_ENTRY_LABEL): New.
3156 (dwarf2out_var_location): Disregard inline entry markers.
3157 (inline_entry_data): New struct.
3158 (inline_entry_data_hasher): New hashtable type.
3159 (inline_entry_data_hasher::hash): New.
3160 (inline_entry_data_hasher::equal): New.
3161 (inline_entry_data_table): New variable.
3162 (add_high_low_attributes): Add DW_AT_entry_pc and
3163 DW_AT_GNU_entry_view attributes if a pending entry is found
3164 in inline_entry_data_table. Add old entry_pc attribute only
3165 if debug nonbinding markers are disabled.
3166 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
3167 markers are enabled.
3168 (block_within_block_p, dwarf2out_inline_entry): New.
3169 (dwarf2out_finish): Check that no entries remained in
3170 inline_entry_data_table.
3171 * final.c (reemit_insn_block_notes): Handle inline entry notes.
3172 (final_scan_insn, notice_source_line): Likewise.
3173 (rest_of_clean_state): Skip inline entry markers.
3174 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
3176 * gimple.c (gimple_build_debug_inline_entry): New.
3177 * gimple.h (enum gimple_debug_subcode): Add
3178 GIMPLE_DEBUG_INLINE_ENTRY.
3179 (gimple_build_debug_inline_entry): Declare.
3180 (gimple_debug_inline_entry_p): New.
3181 (gimple_debug_nonbind_marker_p): Adjust.
3182 * insn-notes.def (INLINE_ENTRY): New.
3183 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3184 inline entry marker notes.
3185 (print_insn): Likewise.
3186 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
3187 (INSN_DEBUG_MARKER_KIND): Likewise.
3188 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
3189 * tree-inline.c (expand_call_inline): Build and insert
3190 debug_inline_entry stmt.
3191 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
3192 inline entry blocks early, if nonbind markers are enabled.
3193 (dump_scope_block): Dump fragment info.
3194 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
3195 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
3196 (gimple_build_debug_inline_entry): New.
3197 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
3198 Enable/disable inline entry points too.
3199 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
3200 (DEBUG_INSN): Describe inline entry markers.
3202 * common.opt (gvariable-location-views): New.
3203 (gvariable-location-views=incompat5): New.
3204 * config.in: Rebuilt.
3205 * configure: Rebuilt.
3206 * configure.ac: Test assembler for view support.
3207 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
3208 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
3209 * dwarf2out.c (var_loc_view): New typedef.
3210 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
3211 (dwarf2out_locviews_in_attribute): New.
3212 (dwarf2out_locviews_in_loclist): New.
3213 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
3214 (enum dw_line_info_opcode): Add LI_adv_address.
3215 (struct dw_line_info_table): Add view.
3216 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
3217 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
3218 (zero_view_p): New variable.
3219 (ZERO_VIEW_P): New macro.
3220 (output_asm_line_debug_info): New.
3221 (struct var_loc_node): Add view.
3222 (add_AT_view_list, AT_loc_list): New.
3223 (add_var_loc_to_decl): Add view param. Test it against last.
3224 (new_loc_list): Add view params. Record them.
3225 (AT_loc_list_ptr): Handle loc and view lists.
3226 (view_list_to_loc_list_val_node): New.
3227 (print_dw_val): Handle dw_val_class_view_list.
3228 (size_of_die): Likewise.
3229 (value_format): Likewise.
3230 (loc_list_has_views): New.
3231 (gen_llsym): Set vl_symbol too.
3232 (maybe_gen_llsym, skip_loc_list_entry): New.
3233 (dwarf2out_maybe_output_loclist_view_pair): New.
3234 (output_loc_list): Output view list or entries too.
3235 (output_view_list_offset): New.
3236 (output_die): Handle dw_val_class_view_list.
3237 (output_dwarf_version): New.
3238 (output_compilation_unit_header): Use it.
3239 (output_skeleton_debug_sections): Likewise.
3240 (output_rnglists, output_line_info): Likewise.
3241 (output_pubnames, output_aranges): Update version comments.
3242 (output_one_line_info_table): Output view numbers in asm comments.
3243 (dw_loc_list): Determine current endview, pass it to new_loc_list.
3244 Call maybe_gen_llsym.
3245 (loc_list_from_tree_1): Adjust.
3246 (add_AT_location_description): Create view list attribute if
3247 needed, check it's absent otherwise.
3248 (convert_cfa_to_fb_loc_list): Adjust.
3249 (maybe_emit_file): Call output_asm_line_debug_info for test.
3250 (dwarf2out_var_location): Reset views as needed. Precompute
3251 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
3252 attribute. Set view.
3253 (new_line_info_table): Reset next view.
3254 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
3255 (dwarf2out_source_line): Likewise. Output view resets and labels to
3256 the assembler, or select appropriate line info opcodes.
3257 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
3258 (optimize_string_length): Catch it. Adjust.
3259 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
3260 dw_val_class_view_list, and remove it if no longer needed.
3261 (hash_loc_list): Hash view numbers.
3262 (loc_list_hasher::equal): Compare them.
3263 (optimize_location_lists): Check whether a view list symbol is
3264 needed, and whether the locview attribute is present, and
3265 whether they match. Remove the locview attribute if no longer
3267 (index_location_lists): Call skip_loc_list_entry for test.
3268 (dwarf2out_finish): Call output_asm_line_debug_info for test.
3269 Use output_dwarf_version.
3270 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
3271 (struct dw_val_node): Add val_view_list.
3272 * final.c (SEEN_NEXT_VIEW): New.
3273 (set_next_view_needed): New.
3274 (clear_next_view_needed): New.
3275 (maybe_output_next_view): New.
3276 (final_start_function): Rename to...
3277 (final_start_function_1): ... this. Take pointer to FIRST,
3278 add SEEN parameter. Emit param bindings in the initial view.
3279 (final_start_function): Reintroduce SEEN-less interface.
3280 (final): Rename to...
3281 (final_1): ... this. Take SEEN parameter. Output final pending
3282 next view at the end.
3283 (final): Reintroduce seen-less interface.
3284 (final_scan_insn): Output pending next view before switching
3285 sections or ending a block. Mark the next view as needed when
3286 outputting variable locations. Notify debug backend of section
3287 changes, and of location view changes.
3288 (rest_of_handle_final): Adjust.
3289 * toplev.c (process_options): Autodetect value for debug variable
3290 location views option. Warn on incompat5 without -gdwarf-5.
3291 * doc/invoke.texi (gvariable-location-views): New.
3292 (gvariable-location-views=incompat5): New.
3293 (gno-variable-location-views): New.
3295 2018-02-08 David Malcolm <dmalcolm@redhat.com>
3297 PR tree-optimization/84136
3298 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
3299 that the result of find_edge is non-NULL.
3301 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3304 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
3305 storing integer register in SImode. Fix cost of 256 and 512
3306 byte aligned SSE register store.
3308 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3310 * config/i386/i386.c (ix86_multiplication_cost): Fix
3311 multiplication cost for TARGET_AVX512DQ.
3313 2018-02-08 Marek Polacek <polacek@redhat.com>
3315 PR tree-optimization/84238
3316 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
3319 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3321 PR tree-optimization/84265
3322 * tree-vect-stmts.c (vectorizable_store): Don't treat
3323 VMAT_CONTIGUOUS accesses as grouped.
3324 (vectorizable_load): Likewise.
3326 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3328 PR tree-optimization/81635
3329 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
3330 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
3331 (test_round_for_mask): New functions.
3332 (wide_int_cc_tests): Call test_round_for_mask.
3333 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
3334 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
3335 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
3336 range returned by get_range_info.
3338 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3341 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
3342 * symtab.c: Include builtins.h
3343 (symtab_node::output_to_lto_symbol_table_p): Move here
3344 from lto-streamer-out.c:output_symbol_p.
3345 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
3346 (output_symbol_p): Move all logic to symtab.c
3347 (produce_symtab): Update.
3349 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3351 * config/s390/s390-opts.h (enum indirect_branch): Define.
3352 * config/s390/s390-protos.h (s390_return_addr_from_memory)
3353 (s390_indirect_branch_via_thunk)
3354 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
3355 (enum s390_indirect_branch_type): Define.
3356 * config/s390/s390.c (struct s390_frame_layout, struct
3357 machine_function): Remove.
3358 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
3359 (indirect_branch_table_label_no, indirect_branch_table_name):
3361 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
3362 (enum s390_indirect_branch_option): Define.
3363 (s390_return_addr_from_memory): New function.
3364 (s390_handle_string_attribute): New function.
3365 (s390_attribute_table): Add new attribute handler.
3366 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
3367 (s390_indirect_branch_via_thunk): New function.
3368 (s390_indirect_branch_via_inline_thunk): New function.
3369 (s390_function_ok_for_sibcall): When jumping via thunk disallow
3370 sibling call optimization for non z10 compiles.
3371 (s390_emit_call): Force indirect branch target to be a single
3372 register. Add r1 clobber for non-z10 compiles.
3373 (s390_emit_epilogue): Emit return jump via return_use expander.
3374 (s390_reorg): Handle JUMP_INSNs as execute targets.
3375 (s390_option_override_internal): Perform validity checks for the
3376 new command line options.
3377 (s390_indirect_branch_attrvalue): New function.
3378 (s390_indirect_branch_settings): New function.
3379 (s390_set_current_function): Invoke s390_indirect_branch_settings.
3380 (s390_output_indirect_thunk_function): New function.
3381 (s390_code_end): Implement target hook.
3382 (s390_case_values_threshold): Implement target hook.
3383 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
3385 * config/s390/s390.h (struct s390_frame_layout)
3386 (struct machine_function): Move here from s390.c.
3387 (TARGET_INDIRECT_BRANCH_NOBP_RET)
3388 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
3389 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
3390 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
3391 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
3392 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
3393 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
3394 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
3395 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
3396 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
3397 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
3398 (mnemonic attribute): Add values which aren't recognized
3400 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
3401 pattern for branch conversion. Fix mnemonic attribute.
3402 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
3403 indirect branch via thunk if requested.
3404 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
3405 ("*indirect_jump"): Disable for branch conversion using out of
3407 ("indirect_jump_via_thunk<mode>_z10")
3408 ("indirect_jump_via_thunk<mode>")
3409 ("indirect_jump_via_inlinethunk<mode>_z10")
3410 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
3411 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
3412 ("casesi_jump_via_inlinethunk<mode>_z10")
3413 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
3414 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
3415 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
3416 ("*indirect2_jump"): Disable for branch conversion.
3417 ("casesi_jump"): Turn into expander and expand patterns for branch
3419 ("return_use"): New expander.
3420 ("*return"): Emit return via thunk and rename it to ...
3421 ("*return<mode>"): ... this one.
3422 * config/s390/s390.opt: Add new options and and enum for the
3425 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3427 * lra-constraints.c (match_reload): Unconditionally use
3428 gen_lowpart_SUBREG, rather than selecting between that
3429 and equivalent gen_rtx_SUBREG code.
3431 2018-02-08 Richard Biener <rguenther@suse.de>
3433 PR tree-optimization/84233
3434 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
3435 changed flag instead of boguously re-using phi_inserted.
3437 2018-02-08 Martin Jambor <mjambor@suse.cz>
3439 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
3440 static local variables.
3442 2018-02-08 Richard Biener <rguenther@suse.de>
3444 PR tree-optimization/84278
3445 * tree-vect-stmts.c (vectorizable_store): When looking for
3446 smaller vector types to perform grouped strided loads/stores
3447 make sure the mode is supported by the target.
3448 (vectorizable_load): Likewise.
3450 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3452 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3453 Increase LDP/STP opportunities by adding adjacent callee-saves.
3455 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3457 PR rtl-optimization/84068
3458 PR rtl-optimization/83459
3459 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3461 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3463 PR tree-optimization/84224
3464 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3465 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3468 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3471 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3472 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3473 count, remove one USE.
3475 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3477 * doc/install.texi (Configuration): Document the
3478 --with-long-double-format={ibm,ieee} PowerPC configuration
3482 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3483 Convert from define_expand to be define_insn_and_split. Rework
3484 float/double/_Float128 conversions to QI/HI/SImode to work with
3485 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3486 conversions to QI/HImode types did a store and then a load to
3487 truncate the value. For conversions to VSX registers, don't split
3488 the insn, instead emit the code directly. Use the code iterator
3489 any_fix to combine signed and unsigned conversions.
3490 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3491 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3492 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3493 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3494 (fix_<mode>di2_hw): Likewise.
3495 (fixuns_<mode>di2_hw): Likewise.
3496 (fix_<mode>si2_hw): Likewise.
3497 (fixuns_<mode>si2_hw): Likewise.
3498 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3499 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3500 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3501 fix<uns>_trunc<SFDF:mode>si2_p8.
3502 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3504 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3505 (fix<uns>_<mode>_mem): Likewise.
3506 (fctiw<u>z_<mode>_mem): Likewise.
3507 (fix<uns>_<mode>_mem): Likewise.
3508 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3509 the register allocator from doing a direct move to the GPRs to do
3510 a store, and instead use the ISA 3.0 store byte/half-word from
3511 vector register instruction. For IEEE 128-bit floating point,
3512 also optimize stores of 32-bit ints.
3513 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
3515 2018-02-07 Alan Hayward <alan.hayward@arm.com>
3517 * genextract.c (push_pathstr_operand): New function to support
3519 (walk_rtx): Call push_pathstr_operand.
3520 (print_path): Support [a-zA-Z].
3522 2018-02-07 Richard Biener <rguenther@suse.de>
3524 PR tree-optimization/84037
3525 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
3526 (cse_and_gimplify_to_preheader): Declare.
3527 (vect_get_place_in_interleaving_chain): Likewise.
3528 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3530 (_loop_vec_info::~_loop_vec_info): Delete it.
3531 (cse_and_gimplify_to_preheader): New function.
3532 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
3533 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
3534 (vectorizable_load): Likewise. For grouped stores always base
3535 the IV on the first element.
3536 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
3537 condition before gimplifying.
3539 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3541 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
3542 *DIV_EXPR and *MOD_EXPR.
3544 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
3547 * config/i386/i386.c (ix86_option_override_internal): Mask out
3548 the CF_SET bit when checking -fcf-protection.
3550 2018-02-07 Tom de Vries <tom@codesourcery.com>
3553 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
3556 2018-02-07 Richard Biener <rguenther@suse.de>
3558 PR tree-optimization/84204
3559 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3562 PR tree-optimization/84205
3563 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
3564 special-case isl_ast_op_zdiv_r.
3566 PR tree-optimization/84223
3567 * graphite-scop-detection.c (gather_bbs::before_dom_children):
3568 Only add conditions from within the region.
3569 (gather_bbs::after_dom_children): Adjust.
3571 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
3574 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
3575 * config/avr/avr.md: Only post-reload split REG-REG moves if
3576 either register is GENERAL_REG_P.
3578 2018-02-07 Jakub Jelinek <jakub@redhat.com>
3580 PR tree-optimization/84235
3581 * tree-ssa-scopedtables.c
3582 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
3583 if the subtraction is performed in floating point type where NaNs are
3584 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
3585 build 1. Formatting fix.
3587 2018-02-06 Jakub Jelinek <jakub@redhat.com>
3590 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
3591 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
3592 and skip it regardless of bb boundaries. Use CALL_P macro,
3593 don't test INSN_P (insn) together with CALL_P or JUMP_P check
3594 unnecessarily, formatting fix.
3596 2018-02-06 Michael Collison <michael.collison@arm.com>
3598 * config/arm/thumb2.md:
3599 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
3600 (*thumb_mov_notscc): Ditto.
3602 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3605 * config/rs6000/rs6000.md (su code attribute): Use "u" for
3606 unsigned_fix, not "s".
3608 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3610 * configure.ac (gcc_fn_eh_frame_ro): New function.
3611 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
3612 correct .eh_frame permissions.
3613 * configure: Regenerate.
3615 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
3617 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
3620 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3622 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3623 Display warning message for -mno-speculate-indirect-jumps.
3625 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
3627 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
3629 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
3631 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
3633 PR tree-optimization/84225
3634 * tree-eh.c (find_trapping_overflow): Only call
3635 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
3637 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3640 * config/i386/i386.c: Reimplement the check of possible options
3641 -mibt/-mshstk conbination. Change error messages.
3642 * doc/invoke.texi: Fix a typo: remove extra '='.
3644 2018-02-06 Marek Polacek <polacek@redhat.com>
3646 PR tree-optimization/84228
3647 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
3649 2018-02-06 Tamar Christina <tamar.christina@arm.com>
3652 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
3653 emitted arch directives.
3654 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
3655 __ARM_FEATURE_COPROC before changing architectures.
3657 2018-02-06 Richard Biener <rguenther@suse.de>
3659 * config/i386/i386.c (print_reg): Fix typo.
3660 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
3662 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
3664 * configure: Regenerate.
3666 2018-02-05 Martin Sebor <msebor@redhat.com>
3668 PR tree-optimization/83369
3669 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
3672 2018-02-05 Martin Liska <mliska@suse.cz>
3674 * doc/invoke.texi: Cherry-pick upstream r323995.
3676 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
3678 * ira.c (ira_init_register_move_cost): Adjust comment.
3680 2018-02-05 Martin Liska <mliska@suse.cz>
3682 PR gcov-profile/84137
3683 * doc/gcov.texi: Fix typo in documentation.
3685 2018-02-05 Martin Liska <mliska@suse.cz>
3687 PR gcov-profile/83879
3688 * doc/gcov.texi: Document necessity of --dynamic-list-data when
3689 using dlopen functionality.
3691 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
3693 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
3694 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
3695 _mm_maskz_range_ss, _mm_mask_range_round_ss,
3696 _mm_maskz_range_round_ss): New intrinsics.
3697 (__builtin_ia32_rangesd128_round)
3698 (__builtin_ia32_rangess128_round): Remove.
3699 (__builtin_ia32_rangesd128_mask_round,
3700 __builtin_ia32_rangess128_mask_round): New builtins.
3701 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
3702 __builtin_ia32_rangess128_round): Remove.
3703 (__builtin_ia32_rangesd128_mask_round,
3704 __builtin_ia32_rangess128_mask_round): New builtins.
3705 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
3706 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
3707 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
3708 "<round_saeonly_constraint>")): Changed to ...
3709 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
3710 "<round_saeonly_scalar_constraint>")): ... this.
3711 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
3712 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
3713 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
3714 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
3715 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
3717 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
3719 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
3721 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
3722 Remove all values except native, 8540 and 8548.
3724 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
3726 * config/i386/i386.c (ix86_output_function_return): Pass
3727 INVALID_REGNUM, instead of -1, as invalid register number to
3728 indirect_thunk_name and output_indirect_thunk.
3730 2018-02-02 Julia Koval <julia.koval@intel.com>
3732 * config.gcc: Add -march=icelake.
3733 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
3734 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
3735 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
3736 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
3737 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
3738 (processor_target_table): Add icelake.
3739 (ix86_option_override_internal): Handle new PTAs.
3740 (get_builtin_code_for_version): Handle icelake.
3741 (M_INTEL_COREI7_ICELAKE): New.
3742 (fold_builtin_cpu): Handle icelake.
3743 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
3744 * doc/invoke.texi: Add -march=icelake.
3746 2018-02-02 Julia Koval <julia.koval@intel.com>
3748 * config/i386/i386.c (ix86_option_override_internal): Change flags type
3749 to wide_int_bitmask.
3750 * wide-int-bitmask.h: New.
3752 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3755 * config/i386/i386.md: Replace Pmode with word_mode in
3756 builtin_setjmp_setup and builtin_longjmp to support x32.
3758 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
3762 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
3764 (rs6000_supported_cpu_names): New static variable.
3765 (linux_cpu_translation_table): Likewise.
3766 (elf_platform) <cpu>: Define new static variable and use it.
3767 Translate kernel AT_PLATFORM name to canonical name if needed.
3768 Error if platform name is unknown.
3770 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
3773 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
3775 2018-02-01 Jeff Law <law@redhat.com>
3778 * config/i386/i386.c (release_scratch_register_on_entry): Add new
3779 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
3780 the scratch if RELEASE_VIA_POP is false.
3781 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
3782 If we have to save a temporary register, decrement SIZE appropriately.
3783 Pass new arguments to release_scratch_register_on_entry.
3784 (ix86_adjust_stack_and_probe): Likewise.
3785 (ix86_emit_probe_stack_range): Pass new arguments to
3786 release_scratch_register_on_entry.
3788 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
3790 PR rtl-optimization/84157
3791 * combine.c (change_zero_ext): Use REG_P predicate in
3792 front of HARD_REGISTER_P predicate.
3794 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
3796 * config/avr/avr.c (avr_option_override): Move disabling of
3797 -fdelete-null-pointer-checks to...
3798 * common/config/avr/avr-common.c (avr_option_optimization_table):
3801 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3803 PR tree-optimization/81635
3804 * tree-data-ref.c (split_constant_offset_1): For types that
3805 wrap on overflow, try to use range info to prove that wrapping
3808 2018-02-01 Renlin Li <renlin.li@arm.com>
3811 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
3813 (aarch64_register_move_cost): Likewise.
3814 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
3816 (REG_CLASS_NAMES): Likewise.
3817 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
3818 TAILCALL_ADDR_REGS. Remove IP registers.
3819 * config/aarch64/aarch64.md (Ucs): Update register constraint.
3821 2018-02-01 Richard Biener <rguenther@suse.de>
3823 * domwalk.h (dom_walker::dom_walker): Add additional constructor
3824 for specifying RPO order and allow NULL for that.
3825 * domwalk.c (dom_walker::dom_walker): Likewise.
3826 (dom_walker::walk): Handle NULL RPO order.
3827 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
3829 (rewrite_update_dom_walker): Likewise.
3830 (mark_def_dom_walker): Likewise.
3832 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3834 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
3835 (aarch64_maybe_expand_sve_subreg_move): Declare.
3836 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
3837 * config/aarch64/predicates.md (aarch64_any_register_operand): New
3839 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
3840 that are semantically a reverse operation.
3841 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
3842 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
3843 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
3845 (aarch64_can_change_mode_class): For big-endian, forbid changes
3846 between two SVE modes if they have different element sizes.
3848 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3850 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
3851 the TImode handling for big-endian targets.
3853 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3855 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
3856 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
3858 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
3859 Remove BSWAP handing for big-endian targets and use the form of
3860 LD1RQ appropariate for the mode.
3862 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3864 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
3865 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
3868 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
3871 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
3872 check for operands that need to go through aarch64_sve_reload_be.
3874 2018-02-01 Jakub Jelinek <jakub@redhat.com>
3876 PR tree-optimization/81661
3877 PR tree-optimization/84117
3878 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
3879 * tree-eh.c: Include gimplify.h.
3880 (find_trapping_overflow, replace_trapping_overflow,
3881 rewrite_to_non_trapping_overflow): New functions.
3882 * tree-vect-loop.c: Include tree-eh.h.
3883 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
3884 * tree-data-ref.c: Include tree-eh.h.
3885 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
3887 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
3889 PR rtl-optimization/84123
3890 * combine.c (change_zero_ext): Check if hard register satisfies
3891 can_change_dest_mode before calling gen_lowpart_SUBREG.
3893 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
3896 * ira.c (ira_init_register_move_cost): Remove assert.
3898 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3900 PR rtl-optimization/84071
3901 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
3902 * doc/tm.texi: Regenerate.
3904 2018-01-31 Richard Biener <rguenther@suse.de>
3906 PR tree-optimization/84132
3907 * tree-data-ref.c (analyze_miv_subscript): Properly
3908 check whether evolution_function_is_affine_multivariate_p
3909 before calling gcd_of_steps_may_divide_p.
3911 2018-01-31 Julia Koval <julia.koval@intel.com>
3914 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3915 * config/i386/i386.md (rdpid_rex64) New.
3916 (rdpid): Make 32bit only.
3918 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
3921 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
3922 an IDENTIFIER_NODE for FUNCTION_TYPE's.
3924 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3927 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
3929 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
3931 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
3933 PR rtl-optimization/84071
3934 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
3935 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
3937 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3939 * config/arc/arc.c (arc_handle_aux_attribute): New function.
3940 (arc_attribute_table): Add 'aux' attribute.
3941 (arc_in_small_data_p): Consider aux like variables.
3942 (arc_is_aux_reg_p): New function.
3943 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
3944 (arc_get_aux_arg): New function.
3945 (prepare_move_operands): Handle aux-register access.
3946 (arc_handle_aux_attribute): New function.
3947 * doc/extend.texi (ARC Variable attributes): Add subsection.
3949 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
3951 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
3952 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
3953 (arc_attribute_table): Add 'uncached' attribute.
3954 (arc_print_operand): Print '.di' flag for uncached memory
3956 (arc_in_small_data_p): Do not consider for small data the uncached
3958 (arc_is_uncached_mem_p): New function.
3959 * config/arc/predicates.md (compact_store_memory_operand): Check
3960 for uncached memory accesses.
3961 (nonvol_nonimm_operand): Likewise.
3962 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
3964 2018-01-31 Jakub Jelinek <jakub@redhat.com>
3967 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
3968 falign-loops=): Add Optimization flag.
3970 2018-01-30 Jeff Law <law@redhat.com>
3973 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
3974 INT_REGISTERS_SAVED. Check it prior to calling
3975 get_scratch_register_on_entry.
3976 (ix86_adjust_stack_and_probe): Similarly.
3977 (ix86_emit_probe_stack_range): Similarly.
3978 (ix86_expand_prologue): Corresponding changes.
3980 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3983 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
3984 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
3986 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
3989 * lra-constraints.c (curr_insn_transform): Process AND in the
3992 2018-01-30 Jakub Jelinek <jakub@redhat.com>
3994 PR rtl-optimization/83986
3995 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
3996 dependence against last_pending_memory_flush in addition to
3999 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
4001 PR tree-optimization/81611
4002 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
4005 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4008 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
4011 2018-01-30 Richard Biener <rguenther@suse.de>
4012 Jakub Jelinek <jakub@redhat.com>
4014 PR tree-optimization/84111
4015 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
4016 inner loops added during recursion, as they don't have up-to-date
4019 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4022 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
4023 (can_inline_edge_by_limits_p): ... here.
4024 (can_early_inline_edge_p, check_callers,
4025 update_caller_keys, update_callee_keys, recursive_inlining,
4026 add_new_edges_to_heap, speculation_useful_p,
4027 inline_small_functions,
4028 inline_small_functions, flatten_function,
4029 inline_to_all_callers_1): Update.
4031 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4033 * profile-count.c (profile_count::combine_with_ipa_count): Handle
4036 2018-01-30 Richard Biener <rguenther@suse.de>
4038 PR tree-optimization/83008
4039 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
4040 invariant and constant vector uses in stmts when they need
4043 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4046 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
4047 * configure: Regenerate.
4049 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4051 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
4053 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
4054 Use gen_rtx_REG rather than gen_lowpart.
4056 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4058 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
4059 rather than 0 when creating partial subregs.
4061 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4063 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
4066 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
4069 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
4070 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
4071 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
4072 flags. This restores the settings used before the 2017-07-24.
4073 Turning off pre increment/decrement/modify allows IVOPTS to
4074 optimize DF/SF loops where the index is an int.
4076 2018-01-29 Richard Biener <rguenther@suse.de>
4077 Kelvin Nilsen <kelvin@gcc.gnu.org>
4080 * tree-vect-stmts.c (vectorizable_call): Don't call
4081 targetm.vectorize_builtin_md_vectorized_function if callee is
4084 2018-01-22 Carl Love <cel@us.ibm.com>
4086 * doc/extend.tex: Fix typo in second arg in
4087 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
4089 2018-01-29 Richard Biener <rguenther@suse.de>
4091 PR tree-optimization/84086
4092 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
4093 (flush_ssaname_freelist): When SSA names were released reset
4094 the SCEV hash table.
4096 2018-01-29 Richard Biener <rguenther@suse.de>
4098 PR tree-optimization/84057
4099 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
4100 removed paths when removing edges.
4102 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
4104 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
4105 -mfunction-return=@var{choice}.
4107 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
4110 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
4112 (layout::print_source_line): Likewise.
4113 (test_get_line_width_without_trailing_whitespace): Add test cases.
4115 2018-01-27 Jakub Jelinek <jakub@redhat.com>
4118 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
4121 2018-01-26 Jim Wilson <jimw@sifive.com>
4123 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
4125 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
4128 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4130 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
4131 and CMP + SUB-immediate -> SUBS.
4133 2018-01-26 Martin Sebor <msebor@redhat.com>
4135 PR tree-optimization/83896
4136 * tree-ssa-strlen.c (get_string_len): Rename...
4137 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
4138 Avoid assuming length is constant.
4139 (handle_char_store): Use HOST_WIDE_INT for string length.
4141 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
4144 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
4145 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
4147 2018-01-26 Richard Biener <rguenther@suse.de>
4149 PR rtl-optimization/84003
4150 * dse.c (record_store): Only record redundant stores when
4151 the earlier store aliases at least all accesses the later one does.
4153 2018-01-26 Jakub Jelinek <jakub@redhat.com>
4155 PR rtl-optimization/83985
4156 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
4157 REG_CFA_RESTORE insns.
4158 (delete_unmarked_insns): Don't ignore separate shrink wrapping
4159 REG_CFA_RESTORE insns here.
4162 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
4163 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
4165 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4167 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
4168 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
4169 (arc_init): Likewise.
4170 (arc_override_options): Likewise.
4171 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
4173 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
4175 * config/arc/arc.h (TARGET_DBNZ): Define.
4176 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
4177 properly set the tune attribute.
4178 (dbnz): Use TARGET_DBNZ guard.
4179 * config/arc/arc.opt (mtune): Add core3 option.
4181 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4183 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
4184 recognize new pic like addresses.
4185 (arc_delegitimize_address): Clean up.
4187 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4189 * config/arc/arc-arches.def: Option mrf16 valid for all
4191 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
4192 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
4193 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
4194 * config/arc/arc-tables.opt: Regenerate.
4195 * config/arc/arc.c (arc_conditional_register_usage): Handle
4196 reduced register file case.
4197 (arc_file_start): Set must have build attributes.
4198 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
4200 * config/arc/arc.opt (mrf16): Add new option.
4201 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
4202 * config/arc/genmultilib.awk: Handle new mrf16 option.
4203 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
4204 * config/arc/t-multilib: Regenerate.
4205 * doc/invoke.texi (ARC Options): Document mrf16 option.
4207 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4209 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
4210 * config/arc/arc.c (arc_handle_secure_attribute): New function.
4211 (arc_attribute_table): Add 'secure_call' attribute.
4212 (arc_print_operand): Print secure call operand.
4213 (arc_function_ok_for_sibcall): Don't optimize tail calls when
4215 (arc_is_secure_call_p): New function. * config/arc/arc.md
4216 (call_i): Add support for sjli instruction.
4217 (call_value_i): Likewise.
4218 * config/arc/constraints.md (Csc): New constraint.
4220 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4221 John Eric Martin <John.Martin@emmicro-us.com>
4223 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
4224 * config/arc/arc.c (_arc_jli_section): New struct.
4225 (arc_jli_section): New type.
4226 (rc_jli_sections): New static variable.
4227 (arc_handle_jli_attribute): New function.
4228 (arc_attribute_table): Add jli_always and jli_fixed attribute.
4229 (arc_file_end): New function.
4230 (TARGET_ASM_FILE_END): Define.
4231 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
4232 (arc_add_jli_section): New function.
4233 (jli_call_scan): Likewise.
4234 (arc_reorg): Call jli_call_scan.
4235 (arc_output_addsi): Remove 'S' from printing asm operand.
4236 (arc_is_jli_call_p): New function.
4237 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
4239 (movhi_insn): Likewise.
4240 (movsi_insn): Likewise.
4241 (movsi_set_cc_insn): Likewise.
4242 (loadqi_update): Likewise.
4243 (load_zeroextendqisi_update): Likewise.
4244 (load_signextendqisi_update): Likewise.
4245 (loadhi_update): Likewise.
4246 (load_zeroextendhisi_update): Likewise.
4247 (load_signextendhisi_update): Likewise.
4248 (loadsi_update): Likewise.
4249 (loadsf_update): Likewise.
4250 (movsicc_insn): Likewise.
4251 (bset_insn): Likewise.
4252 (bxor_insn): Likewise.
4253 (bclr_insn): Likewise.
4254 (bmsk_insn): Likewise.
4255 (bicsi3_insn): Likewise.
4256 (cmpsi_cc_c_insn): Likewise.
4257 (movsi_ne): Likewise.
4258 (movsi_cond_exec): Likewise.
4259 (clrsbsi2): Likewise.
4270 (call_i): Remove 'S' asm letter, add jli instruction.
4271 (call_value_i): Likewise.
4272 * config/arc/arc.op (mjli-always): New option.
4273 * config/arc/constraints.md (Cji): New constraint.
4274 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
4276 (subsf3_fpx): Likewise.
4277 (mulsf3_fpx): Likewise.
4278 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
4280 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
4282 * doc/invoke.texi (ARC): Document mjli-always option.
4284 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
4286 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
4287 avoid addition with 0 and use incw and decw where possible.
4289 2018-01-26 Richard Biener <rguenther@suse.de>
4291 PR tree-optimization/81082
4292 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
4293 association if it requires casting to unsigned.
4294 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
4295 from fold_plusminus_mult_expr to catch important cases late when
4296 range info is available.
4298 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4300 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
4301 * configure.ac (hidden_linkonce): New test.
4302 * configure: Regenerate.
4303 * config.in: Regenerate.
4305 2018-01-26 Julia Koval <julia.koval@intel.com>
4307 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
4308 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
4309 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
4310 _mm_mask_bitshuffle_epi64_mask): Fix type.
4311 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
4312 USI_FTYPE_V4DI_V4DI_USI): Remove.
4313 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
4314 __builtin_ia32_vpshufbitqmb256_mask,
4315 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
4316 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
4317 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
4319 2018-01-26 Alan Modra <amodra@gmail.com>
4322 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
4323 UNSPEC_VBPERMQ. Sort other unspecs.
4325 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
4327 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
4329 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
4332 * predict.c (drop_profile): Do not push/pop cfun; update also
4334 (handle_missing_profiles): Fix logic looking for zero profiles.
4336 2018-01-25 Jakub Jelinek <jakub@redhat.com>
4339 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
4340 on functions with #pragma omp declare simd or functions with simd
4342 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
4343 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
4344 Remove trailing \n from warning_at calls.
4346 2018-01-25 Tom de Vries <tom@codesourcery.com>
4349 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4350 for neutered workers.
4352 2018-01-24 Joseph Myers <joseph@codesourcery.com>
4355 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
4356 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
4358 2018-01-24 Jeff Law <law@redhat.com>
4361 * i386.c (get_probe_interval): Move to earlier point.
4362 (ix86_compute_frame_layout): If -fstack-clash-protection and
4363 the frame is larger than the probe interval, then use pushes
4364 to save registers rather than reg->mem moves.
4365 (ix86_expand_prologue): Remove conditional for int_registers_saved
4368 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
4371 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
4372 min/max for never referenced object.
4374 2018-01-24 Jakub Jelinek <jakub@redhat.com>
4377 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
4379 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
4380 attributes from DECL_ATTRIBUTES (decl) without affecting
4381 DECL_ATTRIBUTES (current_function_decl).
4382 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
4383 functions with non-NULL DECL_ABSTRACT_ORIGIN.
4385 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
4387 PR tree-optimization/83979
4388 * fold-const.c (fold_comparison): Use constant_boolean_node
4389 instead of boolean_{true,false}_node.
4391 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
4393 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
4396 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4398 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4399 Simplify the clause that sets the length attribute.
4400 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4401 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
4402 clause that sets the length attribute.
4403 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4405 2018-01-24 Tom de Vries <tom@codesourcery.com>
4408 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
4409 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
4410 Add strict parameter.
4411 (prevent_branch_around_nothing): Insert dummy insn between branch to
4412 label and label with no ptx insn inbetween.
4413 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
4415 2018-01-24 Tom de Vries <tom@codesourcery.com>
4418 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4419 for neutered threads in warp.
4420 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
4422 2018-01-24 Richard Biener <rguenther@suse.de>
4424 PR tree-optimization/83176
4425 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
4428 2018-01-24 Richard Biener <rguenther@suse.de>
4430 PR tree-optimization/82819
4431 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
4432 code generating pluses that are no-ops in the target precision.
4434 2018-01-24 Richard Biener <rguenther@suse.de>
4437 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
4439 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4441 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
4442 to merge probabilities.
4443 * predict.c (probably_never_executed): Also mark as cold functions
4444 with global 0 profile and guessed local profile.
4445 * profile-count.c (profile_probability::combine_with_count): New
4447 * profile-count.h (profile_probability::operator*,
4448 profile_probability::operator*=, profile_probability::operator/,
4449 profile_probability::operator/=): Reduce precision to adjusted
4450 and set value to guessed on contradictory divisions.
4451 (profile_probability::combine_with_freq): Remove.
4452 (profile_probability::combine_wiht_count): Declare.
4453 (profile_count::force_nonzero):: Set to adjusted.
4454 (profile_count::probability_in):: Set quality to adjusted.
4455 * tree-ssa-tail-merge.c (replace_block_by): Use
4458 2018-01-23 Andrew Waterman <andrew@sifive.com>
4459 Jim Wilson <jimw@sifive.com>
4461 * config/riscv/riscv.c (riscv_stack_boundary): New.
4462 (riscv_option_override): Set riscv_stack_boundary. Handle
4463 riscv_preferred_stack_boundary_arg.
4464 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4465 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4466 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4467 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4468 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4469 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4471 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4474 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4475 of struct ix86_frame.
4476 (ix86_expand_epilogue): Likewise. Add a local variable for
4477 the reg_save_offset field in struct ix86_frame.
4479 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4481 PR tree-optimization/82604
4482 * tree-loop-distribution.c (enum partition_kind): New enum item
4483 PKIND_PARTIAL_MEMSET.
4484 (partition_builtin_p): Support above new enum item.
4485 (generate_code_for_partition): Ditto.
4486 (compute_access_range): Differentiate cases that equality can be
4487 proven at all loops, the innermost loops or no loops.
4488 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4489 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4490 (finalize_partitions, distribute_loop): Don't fuse partition of
4491 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4492 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4495 2018-01-23 Martin Liska <mliska@suse.cz>
4497 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4498 order to ignore the predictor.
4499 (PRED_POLYMORPHIC_CALL): Likewise.
4500 (PRED_RECURSIVE_CALL): Likewise.
4502 2018-01-23 Martin Liska <mliska@suse.cz>
4504 * tree-profile.c (tree_profiling): Print function header to
4505 aware reader which function we are working on.
4506 * value-prof.c (gimple_find_values_to_profile): Do not print
4507 not interesting value histograms.
4509 2018-01-23 Martin Liska <mliska@suse.cz>
4511 * profile-count.h (enum profile_quality): Add
4512 profile_uninitialized as the first value. Do not number values
4513 as they are zero based.
4514 (profile_count::verify): Update sanity check.
4515 (profile_probability::verify): Likewise.
4517 2018-01-23 Nathan Sidwell <nathan@acm.org>
4519 * doc/invoke.texi (ffor-scope): Deprecate.
4521 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4523 PR tree-optimization/83510
4524 * domwalk.c (set_all_edges_as_executable): New function.
4525 (dom_walker::dom_walker): Convert bool param
4526 "skip_unreachable_blocks" to enum reachability. Move setup of
4527 edge flags to set_all_edges_as_executable and only do it when
4528 reachability is REACHABLE_BLOCKS.
4529 * domwalk.h (enum dom_walker::reachability): New enum.
4530 (dom_walker::dom_walker): Convert bool param
4531 "skip_unreachable_blocks" to enum reachability.
4532 (set_all_edges_as_executable): New decl.
4533 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
4534 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
4536 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
4537 but converting true to REACHABLE_BLOCKS.
4538 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
4540 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
4541 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
4542 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
4544 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
4545 if check_all_array_refs will be called.
4547 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4549 * tree.c (selftest::test_location_wrappers): Add more test
4552 2018-01-23 David Malcolm <dmalcolm@redhat.com>
4554 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
4555 (selftest::test_bit_in_range): Likewise.
4557 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4560 * doc/sourcebuild.texi (vect_float): Say that the selector
4561 only describes the situation when -funsafe-math-optimizations is on.
4562 (vect_float_strict): Document.
4564 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
4566 PR tree-optimization/83965
4567 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
4568 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
4569 instead of checking only for a reduction.
4570 (vect_recog_widen_sum_pattern): Likewise.
4572 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4574 * predict.c (probably_never_executed): Only use precise profile info.
4575 (compute_function_frequency): Skip after inlining hack since we now
4576 have quality checking.
4578 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4580 * profile-count.h (profile_probability::very_unlikely,
4581 profile_probability::unlikely, profile_probability::even): Set
4582 precision to guessed.
4584 2018-01-23 Richard Biener <rguenther@suse.de>
4586 PR tree-optimization/83963
4587 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
4588 Properly terminate dominator walk when crossing the exit edge not
4589 when visiting its source block.
4591 2018-01-23 Jakub Jelinek <jakub@redhat.com>
4594 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
4595 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
4597 2018-01-22 Jakub Jelinek <jakub@redhat.com>
4599 PR tree-optimization/83957
4600 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
4601 semicolon after for body surrounded by braces.
4603 PR tree-optimization/83081
4604 * profile-count.h (profile_probability::split): New method.
4605 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
4606 Use profile_probability::split.
4607 (do_compare_rtx_and_jump): Fix adjustment of probabilities
4608 when splitting a single conditional jump into 2.
4610 2018-01-22 David Malcolm <dmalcolm@redhat.com>
4612 PR tree-optimization/69452
4613 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
4616 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4618 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
4619 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
4620 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
4622 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4624 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
4625 * config/rl78/rl78.md (movdi): New define_expand.
4626 * config/rl78/rl78.c (rl78_split_movdi): New function.
4628 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
4631 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
4633 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
4634 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
4635 128-bit to produce an UNSPEC move to get the double word with the
4636 signbit and then a shift directly to do signbit.
4637 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
4638 implementation with a new version that just does either a direct
4639 move or a regular move. Move memory interface to separate insns.
4640 Move insns so they are next to the expander.
4641 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
4642 with signbit move. Split big and little endian case.
4643 (signbit<mode>2_dm_mem_le): Likewise.
4644 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
4645 (signbit<mode>2_dm2): Likewise.
4647 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4649 * config/rl78/rl78.md (anddi3): New define_expand.
4651 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4653 * config/rl78/rl78.md (umindi3): New define_expand.
4655 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4657 * config/rl78/rl78.md (smindi3): New define_expand.
4659 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4661 * config/rl78/rl78.md (smaxdi3): New define_expand.
4663 2018-01-22 Carl Love <cel@us.ibm.com>
4665 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
4666 LVX_V1TI): Add macro expansion.
4667 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
4668 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
4669 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
4670 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
4671 Change check to determine if the instruction is a byte reversing
4672 entry. Fix typo in comment.
4673 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
4674 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
4675 Add def_builtin calls for new builtins.
4676 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
4677 Add define_insn expansion.
4679 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4681 * config/rl78/rl78.md (umaxdi3): New define_expand.
4683 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
4685 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
4686 for non-QImode registers.
4688 2018-01-22 Richard Biener <rguenther@suse.de>
4690 PR tree-optimization/83963
4691 * graphite-scop-detection.c (scop_detection::get_sese): Delay
4692 including the loop exit block.
4693 (scop_detection::merge_sese): Likewise.
4694 (scop_detection::add_scop): Do it here instead.
4696 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4698 * doc/sourcebuild.texi (arm_softfloat): Document.
4700 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
4703 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
4704 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
4705 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
4707 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4708 David Edelsohn <dje.gcc@gmail.com>
4711 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4712 Change "crset eq" to "crset 2".
4713 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4714 (*call_indirect_aix<mode>_nospec): Likewise.
4715 (*call_value_indirect_aix<mode>_nospec): Likewise.
4716 (*call_indirect_elfv2<mode>_nospec): Likewise.
4717 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
4718 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
4719 change assembly output from . to $.
4720 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4721 (indirect_jump<mode>_nospec): Change assembly output from . to $.
4722 (*tablejump<mode>_internal1_nospec): Likewise.
4724 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
4727 * config/sh/sh_optimize_sett_clrt.cc:
4728 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
4730 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
4732 PR tree-optimization/83940
4733 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
4734 offset_dt to vect_constant_def rather than vect_unknown_def_type.
4735 (vect_check_load_store_mask): Add a mask_dt_out parameter and
4736 use it to pass back the definition type.
4737 (vect_check_store_rhs): Likewise rhs_dt_out.
4738 (vect_build_gather_load_calls): Add a mask_dt argument and use
4739 it instead of a call to vect_is_simple_use.
4740 (vectorizable_store): Update calls to vect_check_load_store_mask
4741 and vect_check_store_rhs. Use the dt returned by the latter instead
4742 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
4743 instead of calls to vect_is_simple_use. Pass the scalar rather
4744 than the vector operand to vect_is_simple_use when handling
4745 second and subsequent copies of an rhs value.
4746 (vectorizable_load): Update calls to vect_check_load_store_mask
4747 and vect_build_gather_load_calls. Use the cached mask_dt and
4748 gs_info.offset_dt instead of calls to vect_is_simple_use.
4750 2018-01-20 Jakub Jelinek <jakub@redhat.com>
4753 * tree-emutls.c: Include gimplify.h.
4754 (lower_emutls_2): New function.
4755 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
4756 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
4757 it before further processing.
4760 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
4761 UINTVAL (trueop1) instead of INTVAL (op1).
4763 2018-01-19 Jakub Jelinek <jakub@redhat.com>
4767 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
4768 INCOMING_FRAME_SP_OFFSET if not defined.
4769 (scan_trace): Add ENTRY argument. If true and
4770 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
4771 emit a note to adjust the CFA offset.
4772 (create_cfi_notes): Adjust scan_trace callers.
4773 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
4774 INCOMING_FRAME_SP_OFFSET in the CIE.
4775 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
4776 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
4778 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
4779 * doc/tm.texi: Regenerated.
4781 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4783 PR rtl-optimization/83147
4784 * lra-constraints.c (remove_inheritance_pseudos): Use
4785 lra_substitute_pseudo_within_insn.
4787 2018-01-19 Tom de Vries <tom@codesourcery.com>
4788 Cesar Philippidis <cesar@codesourcery.com>
4791 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
4793 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
4796 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
4797 spaces for function labels.
4799 2018-01-19 Martin Liska <mliska@suse.cz>
4801 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
4802 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
4803 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
4804 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
4805 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
4806 (PRED_CONST_RETURN): Change from 69 to 65.
4807 (PRED_NULL_RETURN): Change from 91 to 71.
4808 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
4809 (PRED_LOOP_GUARD): Change from 66 to 73.
4811 2018-01-19 Martin Liska <mliska@suse.cz>
4813 * predict.c (predict_insn_def): Add new assert.
4814 (struct branch_predictor): Change type to signed integer.
4815 (test_prediction_value_range): Amend test to cover
4817 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
4818 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
4819 (PRED_LOOP_ITERATIONS_MAX): Likewise.
4820 (PRED_LOOP_IV_COMPARE): Likewise.
4821 * predict.h (PROB_UNINITIALIZED): Define new constant.
4823 2018-01-19 Martin Liska <mliska@suse.cz>
4825 * predict.c (dump_prediction): Add new format for
4826 analyze_brprob.py script which is enabled with -details
4828 * profile-count.h (precise_p): New function.
4830 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4832 PR tree-optimization/83922
4833 * tree-vect-loop.c (vect_verify_full_masking): Return false if
4834 there are no statements that need masking.
4835 (vect_active_double_reduction_p): New function.
4836 (vect_analyze_loop_operations): Use it when handling phis that
4837 are not in the loop header.
4839 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
4841 PR tree-optimization/83914
4842 * tree-vect-loop.c (vectorizable_induction): Don't convert
4843 init_expr or apply the peeling adjustment for inductions
4844 that are nested within the vectorized loop.
4846 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4848 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
4851 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4855 * function.h (gimplify_parameters): Add gimple_seq * argument.
4856 * function.c: Include gimple.h and options.h.
4857 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
4858 for the added local temporaries if needed.
4859 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
4860 if there are any parameter cleanups, wrap whole body into a
4861 try/finally with the cleanups.
4863 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
4866 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
4867 Use GET_MODE_CLASS for scalar floating point.
4869 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4873 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4874 Fix call of call_cgraph_insertion_hooks.
4876 2018-01-18 Martin Sebor <msebor@redhat.com>
4878 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
4880 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
4883 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
4886 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
4889 * common.opt: (-ffile-prefix-map): New option.
4890 * opts.c (common_handle_option): Defer it.
4891 * opts-global.c (handle_common_deferred_options): Handle it.
4892 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
4893 * file-prefix-map.h: New file.
4894 (remap_debug_filename, add_debug_prefix_map): ...here.
4895 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
4896 * final.c (debug_prefix_map, add_debug_prefix_map
4897 remap_debug_filename): Move to...
4898 * file-prefix-map.c: New file.
4899 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
4900 generalize, get rid of alloca(), use strrchr() instead of strchr().
4901 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
4902 Implement in terms of add_prefix_map().
4903 (remap_macro_filename, remap_debug_filename): Implement in term of
4905 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
4906 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
4907 * dbxout.c: Include file-prefix-map.h.
4908 * varasm.c: Likewise.
4909 * vmsdbgout.c: Likewise.
4910 * xcoffout.c: Likewise.
4911 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
4912 * doc/cppopts.texi (-fmacro-prefix-map): Document.
4913 * doc/invoke.texi (-ffile-prefix-map): Document.
4914 (-fdebug-prefix-map): Update description.
4916 2018-01-18 Martin Liska <mliska@suse.cz>
4918 * config/i386/i386.c (indirect_thunk_name): Document that also
4920 (output_indirect_thunk): Document why both instructions
4921 (pause and lfence) are generated.
4923 2018-01-18 Richard Biener <rguenther@suse.de>
4925 PR tree-optimization/83887
4926 * graphite-scop-detection.c
4927 (scop_detection::get_nearest_dom_with_single_entry): Remove.
4928 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
4929 (scop_detection::merge_sese): Re-implement with a flood-fill
4930 algorithm that properly finds a SESE region if it exists.
4932 2018-01-18 Jakub Jelinek <jakub@redhat.com>
4935 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
4936 pointer_diff optimizations use view_convert instead of convert.
4938 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4940 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4941 Generate different code for -mno-speculate-indirect-jumps.
4942 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4943 (*call_indirect_aix<mode>): Disable for
4944 -mno-speculate-indirect-jumps.
4945 (*call_indirect_aix<mode>_nospec): New define_insn.
4946 (*call_value_indirect_aix<mode>): Disable for
4947 -mno-speculate-indirect-jumps.
4948 (*call_value_indirect_aix<mode>_nospec): New define_insn.
4949 (*sibcall_nonlocal_sysv<mode>): Generate different code for
4950 -mno-speculate-indirect-jumps.
4951 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4953 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4955 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
4956 long double type, set the flags for noting the default long double
4957 type, even if we don't pass or return a long double type.
4959 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
4962 * ipa-inline.c (flatten_function): Do not overwrite final inlining
4965 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
4967 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
4968 support for merge[hl].
4969 (fold_mergehl_helper): New helper function.
4970 (tree-vector-builder.h): New #include for tree_vector_builder usage.
4971 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
4972 (altivec_vmrglw_direct): Add xxmrglw insn.
4974 2018-01-17 Andrew Waterman <andrew@sifive.com>
4976 * config/riscv/riscv.c (riscv_conditional_register_usage): If
4977 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
4979 2018-01-17 David Malcolm <dmalcolm@redhat.com>
4982 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
4983 call the lto_location_cache before reading the
4984 DECL_SOURCE_LOCATION of the types.
4986 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4987 Richard Sandiford <richard.sandiford@linaro.org>
4989 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
4990 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
4991 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
4992 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
4993 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
4995 * config/aarch64/constraints.md (aarch64_movti_operand):
4997 * config/aarch64/predicates.md (Uti): Add new constraint.
4999 2018-01-17 Carl Love <cel@us.ibm.com>
5001 * config/rs6000/vsx.md (define_expand xl_len_r,
5002 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
5003 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
5005 (define_expand, define_insn): Move the shift left from the
5006 define_insn to the define_expand for lxvl and stxvl instructions.
5007 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
5008 and XL_LEN_R definitions to PURE.
5010 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
5012 * config/i386/i386.c (indirect_thunk_name): Declare regno
5013 as unsigned int. Compare regno with INVALID_REGNUM.
5014 (output_indirect_thunk): Ditto.
5015 (output_indirect_thunk_function): Ditto.
5016 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
5017 in the call to output_indirect_thunk_function.
5019 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
5022 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
5023 rather than the size of inner_type to determine the stack slot size
5024 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
5026 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
5029 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
5032 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
5034 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
5035 endian Linux systems to optionally enable multilibs for selecting
5036 the long double type if the user configured an explicit type.
5037 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
5038 have no long double multilibs if not defined.
5039 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
5040 warn if the user used -mabi={ieee,ibm}longdouble and we built
5041 multilibs for long double.
5042 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
5043 appropriate multilib option.
5044 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
5046 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
5047 for building long double multilibs.
5048 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
5050 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
5052 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
5055 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
5057 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
5060 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
5063 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
5066 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5068 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
5069 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
5071 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5073 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
5074 different rtl trees depending on TARGET_64BIT.
5075 (rs6000_gen_lvx): Likewise.
5077 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5079 * config/visium/visium.md (nop): Tweak comment.
5080 (hazard_nop): Likewise.
5082 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5084 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
5085 -mspeculate-indirect-jumps.
5086 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
5087 for -mno-speculate-indirect-jumps.
5088 (*call_indirect_elfv2<mode>_nospec): New define_insn.
5089 (*call_value_indirect_elfv2<mode>): Disable for
5090 -mno-speculate-indirect-jumps.
5091 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
5092 (indirect_jump): Emit different RTL for
5093 -mno-speculate-indirect-jumps.
5094 (*indirect_jump<mode>): Disable for
5095 -mno-speculate-indirect-jumps.
5096 (*indirect_jump<mode>_nospec): New define_insn.
5097 (tablejump): Emit different RTL for
5098 -mno-speculate-indirect-jumps.
5099 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
5100 (tablejumpsi_nospec): New define_expand.
5101 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
5102 (tablejumpdi_nospec): New define_expand.
5103 (*tablejump<mode>_internal1): Disable for
5104 -mno-speculate-indirect-jumps.
5105 (*tablejump<mode>_internal1_nospec): New define_insn.
5106 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
5109 2018-01-16 Artyom Skrobov tyomitch@gmail.com
5111 * caller-save.c (insert_save): Drop unnecessary parameter. All
5114 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5115 Richard Biener <rguenth@suse.de>
5118 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
5119 return early, inline manually is_gimple_sizepos. Make sure if we
5120 call gimplify_expr we don't end up with a gimple constant.
5121 * tree.c (variably_modified_type_p): Don't return true for
5122 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
5123 * gimplify.h (is_gimple_sizepos): Remove.
5125 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5127 PR tree-optimization/83857
5128 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
5129 vectorizable_live_operation for pure SLP statements.
5130 (vectorizable_live_operation): Handle PHIs.
5132 2018-01-16 Richard Biener <rguenther@suse.de>
5134 PR tree-optimization/83867
5135 * tree-vect-stmts.c (vect_transform_stmt): Precompute
5136 nested_in_vect_loop_p since the scalar stmt may get invalidated.
5138 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5141 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
5142 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
5143 If off is not INTEGER_CST, issue a may not be aligned warning
5144 rather than isn't aligned. Use isn%'t rather than isn't.
5145 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
5147 <case MULT_EXPR>: Improve the case when bottom and one of the
5148 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
5149 operand, in that case check if the other operand is multiple of
5150 bottom divided by the INTEGER_CST operand.
5152 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5155 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
5156 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
5157 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
5158 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
5159 * config/pa/pa.c (pa_function_arg_advance): Likewise.
5160 (pa_function_arg, pa_arg_partial_bytes): Likewise.
5161 (pa_function_arg_size): New function.
5163 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5165 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
5166 in a separate statement.
5168 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5170 PR tree-optimization/83847
5171 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
5172 group gathers and scatters.
5174 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5176 PR rtl-optimization/86620
5177 * params.def (max-sched-ready-insns): Bump minimum value to 1.
5179 PR rtl-optimization/83213
5180 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
5181 to last if both are JUMP_INSNs.
5183 PR tree-optimization/83843
5184 * gimple-ssa-store-merging.c
5185 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
5186 store_immediate_info for bswap/nop orig_stores.
5188 2018-01-15 Andrew Waterman <andrew@sifive.com>
5190 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
5192 <UDIV>: Increase cost if !TARGET_DIV.
5194 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
5196 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
5197 (define_attr "cr_logical_3op"): New.
5198 (cceq_ior_compare): Adjust.
5199 (cceq_ior_compare_complement): Adjust.
5200 (*cceq_rev_compare): Adjust.
5201 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
5202 (is_cracked_insn): Adjust.
5203 (insn_must_be_first_in_group): Adjust.
5204 * config/rs6000/40x.md: Adjust.
5205 * config/rs6000/440.md: Adjust.
5206 * config/rs6000/476.md: Adjust.
5207 * config/rs6000/601.md: Adjust.
5208 * config/rs6000/603.md: Adjust.
5209 * config/rs6000/6xx.md: Adjust.
5210 * config/rs6000/7450.md: Adjust.
5211 * config/rs6000/7xx.md: Adjust.
5212 * config/rs6000/8540.md: Adjust.
5213 * config/rs6000/cell.md: Adjust.
5214 * config/rs6000/e300c2c3.md: Adjust.
5215 * config/rs6000/e500mc.md: Adjust.
5216 * config/rs6000/e500mc64.md: Adjust.
5217 * config/rs6000/e5500.md: Adjust.
5218 * config/rs6000/e6500.md: Adjust.
5219 * config/rs6000/mpc.md: Adjust.
5220 * config/rs6000/power4.md: Adjust.
5221 * config/rs6000/power5.md: Adjust.
5222 * config/rs6000/power6.md: Adjust.
5223 * config/rs6000/power7.md: Adjust.
5224 * config/rs6000/power8.md: Adjust.
5225 * config/rs6000/power9.md: Adjust.
5226 * config/rs6000/rs64.md: Adjust.
5227 * config/rs6000/titan.md: Adjust.
5229 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5231 * config/i386/predicates.md (indirect_branch_operand): Rewrite
5232 ix86_indirect_branch_register logic.
5234 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5236 * config/i386/constraints.md (Bs): Update
5237 ix86_indirect_branch_register check. Don't check
5238 ix86_indirect_branch_register with GOT_memory_operand.
5240 * config/i386/predicates.md (GOT_memory_operand): Don't check
5241 ix86_indirect_branch_register here.
5242 (GOT32_symbol_operand): Likewise.
5244 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5246 * config/i386/predicates.md (constant_call_address_operand):
5247 Rewrite ix86_indirect_branch_register logic.
5248 (sibcall_insn_operand): Likewise.
5250 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5252 * config/i386/constraints.md (Bs): Replace
5253 ix86_indirect_branch_thunk_register with
5254 ix86_indirect_branch_register.
5256 * config/i386/i386.md (indirect_jump): Likewise.
5257 (tablejump): Likewise.
5258 (*sibcall_memory): Likewise.
5259 (*sibcall_value_memory): Likewise.
5260 Peepholes of indirect call and jump via memory: Likewise.
5261 * config/i386/i386.opt: Likewise.
5262 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5263 (GOT_memory_operand): Likewise.
5264 (call_insn_operand): Likewise.
5265 (sibcall_insn_operand): Likewise.
5266 (GOT32_symbol_operand): Likewise.
5268 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5271 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
5272 type rather than type addr's type points to.
5273 (expand_omp_atomic_mutex): Likewise.
5274 (expand_omp_atomic): Likewise.
5276 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5279 * config/i386/i386.c (output_indirect_thunk_function): Use
5280 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
5281 for __x86_return_thunk.
5283 2018-01-15 Richard Biener <rguenther@suse.de>
5286 * expmed.c (extract_bit_field_1): Fix typo.
5288 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5291 * config/arm/iterators.md (VF): New mode iterator.
5292 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
5293 Remove integer-related logic from pattern.
5294 (neon_vabd<mode>_3): Likewise.
5296 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5299 * common.opt (fstrict-overflow): No longer an alias.
5300 (fwrapv-pointer): New option.
5301 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
5302 also for pointer types based on flag_wrapv_pointer.
5303 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
5304 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
5305 opts->x_flag_wrapv got set.
5306 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
5307 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
5308 POINTER_TYPE_OVERFLOW_UNDEFINED.
5309 * match.pd: Likewise in address comparison pattern.
5310 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
5312 2018-01-15 Richard Biener <rguenther@suse.de>
5315 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
5316 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
5317 Reset type names to their identifier if their TYPE_DECL doesn't
5318 have linkage (and thus is used for ODR and devirt).
5319 (save_debug_info_for_decl): Remove.
5320 (save_debug_info_for_type): Likewise.
5321 (add_tree_to_fld_list): Adjust.
5322 * tree-pretty-print.c (dump_generic_node): Make dumping of
5323 type names more robust.
5325 2018-01-15 Richard Biener <rguenther@suse.de>
5327 * BASE-VER: Bump to 8.0.1.
5329 2018-01-14 Martin Sebor <msebor@redhat.com>
5332 * builtins.c (check_access): Avoid warning when the no-warning bit
5335 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
5337 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
5338 * ira-color (allocno_hard_regs_compare): Likewise.
5340 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
5343 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
5344 Use .pushsection/.popsection.
5346 2018-01-14 Martin Sebor <msebor@redhat.com>
5349 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
5351 2018-01-14 Jakub Jelinek <jakub@redhat.com>
5353 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
5354 entry from extra_headers.
5355 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
5356 extra_headers, make the list bitwise identical to the i?86-*-* one.
5358 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5360 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
5361 -mcmodel=large with -mindirect-branch=thunk,
5362 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
5363 -mfunction-return=thunk-extern.
5364 * doc/invoke.texi: Document -mcmodel=large is incompatible with
5365 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
5366 -mfunction-return=thunk and -mfunction-return=thunk-extern.
5368 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5370 * config/i386/i386.c (print_reg): Print the name of the full
5371 integer register without '%'.
5372 (ix86_print_operand): Handle 'V'.
5373 * doc/extend.texi: Document 'V' modifier.
5375 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5377 * config/i386/constraints.md (Bs): Disallow memory operand for
5378 -mindirect-branch-register.
5380 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5381 (GOT_memory_operand): Likewise.
5382 (call_insn_operand): Likewise.
5383 (sibcall_insn_operand): Likewise.
5384 (GOT32_symbol_operand): Likewise.
5385 * config/i386/i386.md (indirect_jump): Call convert_memory_address
5386 for -mindirect-branch-register.
5387 (tablejump): Likewise.
5388 (*sibcall_memory): Likewise.
5389 (*sibcall_value_memory): Likewise.
5390 Disallow peepholes of indirect call and jump via memory for
5391 -mindirect-branch-register.
5392 (*call_pop): Replace m with Bw.
5393 (*call_value_pop): Likewise.
5394 (*sibcall_pop_memory): Replace m with Bs.
5395 * config/i386/i386.opt (mindirect-branch-register): New option.
5396 * doc/invoke.texi: Document -mindirect-branch-register option.
5398 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5400 * config/i386/i386-protos.h (ix86_output_function_return): New.
5401 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
5402 set function_return_type.
5403 (indirect_thunk_name): Add ret_p to indicate thunk for function
5405 (output_indirect_thunk_function): Pass false to
5406 indirect_thunk_name.
5407 (ix86_output_indirect_branch_via_reg): Likewise.
5408 (ix86_output_indirect_branch_via_push): Likewise.
5409 (output_indirect_thunk_function): Create alias for function
5410 return thunk if regno < 0.
5411 (ix86_output_function_return): New function.
5412 (ix86_handle_fndecl_attribute): Handle function_return.
5413 (ix86_attribute_table): Add function_return.
5414 * config/i386/i386.h (machine_function): Add
5415 function_return_type.
5416 * config/i386/i386.md (simple_return_internal): Use
5417 ix86_output_function_return.
5418 (simple_return_internal_long): Likewise.
5419 * config/i386/i386.opt (mfunction-return=): New option.
5420 (indirect_branch): Mention -mfunction-return=.
5421 * doc/extend.texi: Document function_return function attribute.
5422 * doc/invoke.texi: Document -mfunction-return= option.
5424 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5426 * config/i386/i386-opts.h (indirect_branch): New.
5427 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
5428 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
5429 with local indirect jump when converting indirect call and jump.
5430 (ix86_set_indirect_branch_type): New.
5431 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
5432 (indirectlabelno): New.
5433 (indirect_thunk_needed): Likewise.
5434 (indirect_thunk_bnd_needed): Likewise.
5435 (indirect_thunks_used): Likewise.
5436 (indirect_thunks_bnd_used): Likewise.
5437 (INDIRECT_LABEL): Likewise.
5438 (indirect_thunk_name): Likewise.
5439 (output_indirect_thunk): Likewise.
5440 (output_indirect_thunk_function): Likewise.
5441 (ix86_output_indirect_branch_via_reg): Likewise.
5442 (ix86_output_indirect_branch_via_push): Likewise.
5443 (ix86_output_indirect_branch): Likewise.
5444 (ix86_output_indirect_jmp): Likewise.
5445 (ix86_code_end): Call output_indirect_thunk_function if needed.
5446 (ix86_output_call_insn): Call ix86_output_indirect_branch if
5448 (ix86_handle_fndecl_attribute): Handle indirect_branch.
5449 (ix86_attribute_table): Add indirect_branch.
5450 * config/i386/i386.h (machine_function): Add indirect_branch_type
5451 and has_local_indirect_jump.
5452 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5454 (tablejump): Likewise.
5455 (*indirect_jump): Use ix86_output_indirect_jmp.
5456 (*tablejump_1): Likewise.
5457 (simple_return_indirect_internal): Likewise.
5458 * config/i386/i386.opt (mindirect-branch=): New option.
5459 (indirect_branch): New.
5462 (thunk-inline): Likewise.
5463 (thunk-extern): Likewise.
5464 * doc/extend.texi: Document indirect_branch function attribute.
5465 * doc/invoke.texi: Document -mindirect-branch= option.
5467 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5470 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5472 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5474 * ipa-inline.c (want_inline_small_function_p): Return false if
5475 inlining has already failed with CIF_FINAL_ERROR.
5476 (update_caller_keys): Call want_inline_small_function_p before
5478 (update_callee_keys): Likewise.
5480 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5482 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5484 (rs6000_quadword_masked_address_p): Likewise.
5485 (quad_aligned_load_p): Likewise.
5486 (quad_aligned_store_p): Likewise.
5487 (const_load_sequence_p): Add comment to describe the outer-most loop.
5488 (mimic_memory_attributes_and_flags): New function.
5489 (rs6000_gen_stvx): Likewise.
5490 (replace_swapped_aligned_store): Likewise.
5491 (rs6000_gen_lvx): Likewise.
5492 (replace_swapped_aligned_load): Likewise.
5493 (replace_swapped_load_constant): Capitalize argument name in
5494 comment describing this function.
5495 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5496 and stores that access quad-word aligned addresses and replace
5497 with stvx or lvx instructions when appropriate.
5498 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5499 New function prototype.
5500 (rs6000_quadword_masked_address_p): Likewise.
5501 (rs6000_gen_lvx): Likewise.
5502 (rs6000_gen_stvx): Likewise.
5503 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5504 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5505 when memory address is aligned.
5506 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5507 this split to select lvx instruction when memory address is aligned.
5508 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5509 instruction when memory address is aligned.
5510 (*vsx_le_perm_load_v16qi): Likewise.
5511 (four unnamed splitters): Modify to select the stvx instruction
5512 when memory is aligned.
5514 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
5516 * predict.c (determine_unlikely_bbs): Handle correctly BBs
5517 which appears in the queue multiple times.
5519 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5520 Alan Hayward <alan.hayward@arm.com>
5521 David Sherwood <david.sherwood@arm.com>
5523 * tree-vectorizer.h (vec_lower_bound): New structure.
5524 (_loop_vec_info): Add check_nonzero and lower_bounds.
5525 (LOOP_VINFO_CHECK_NONZERO): New macro.
5526 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
5527 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
5528 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
5529 fields. Make seg_len the distance travelled, not including the
5531 (dr_direction_indicator): Declare.
5532 (dr_zero_step_indicator): Likewise.
5533 (dr_known_forward_stride_p): Likewise.
5534 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
5536 (runtime_alias_check_p): Allow runtime alias checks with
5538 (operator ==): Compare access_size and align.
5539 (prune_runtime_alias_test_list): Rework for new distinction between
5540 the access_size and seg_len.
5541 (create_intersect_range_checks_index): Likewise. Cope with polynomial
5543 (get_segment_min_max): New function.
5544 (create_intersect_range_checks): Use it.
5545 (dr_step_indicator): New function.
5546 (dr_direction_indicator): Likewise.
5547 (dr_zero_step_indicator): Likewise.
5548 (dr_known_forward_stride_p): Likewise.
5549 * tree-loop-distribution.c (data_ref_segment_size): Return
5550 DR_STEP * (niters - 1).
5551 (compute_alias_check_pairs): Update call to the dr_with_seg_len
5553 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
5554 (vect_preserves_scalar_order_p): New function, split out from...
5555 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
5556 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
5557 (vect_vfa_access_size): New function.
5558 (vect_vfa_align): Likewise.
5559 (vect_compile_time_alias): Take access_size_a and access_b arguments.
5560 (dump_lower_bound): New function.
5561 (vect_check_lower_bound): Likewise.
5562 (vect_small_gap_p): Likewise.
5563 (vectorizable_with_step_bound_p): Likewise.
5564 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
5565 depencies if the vectorization factor is 1. Convert the checks
5566 for nonzero steps into checks on the bounds of DR_STEP. Try using
5567 a bunds check for variable steps if the minimum required step is
5568 relatively small. Update calls to the dr_with_seg_len
5569 constructor and to vect_compile_time_alias.
5570 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
5572 (vect_loop_versioning): Call it.
5573 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
5575 (vect_estimate_min_profitable_iters): Account for any bounds checks.
5577 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5578 Alan Hayward <alan.hayward@arm.com>
5579 David Sherwood <david.sherwood@arm.com>
5581 * doc/sourcebuild.texi (vect_scatter_store): Document.
5582 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
5584 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
5586 * genopinit.c (main): Add supports_vec_scatter_store and
5587 supports_vec_scatter_store_cached to target_optabs.
5588 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
5589 IFN_MASK_SCATTER_STORE.
5590 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
5592 * internal-fn.h (internal_store_fn_p): Declare.
5593 (internal_fn_stored_value_index): Likewise.
5594 * internal-fn.c (scatter_store_direct): New macro.
5595 (expand_scatter_store_optab_fn): New function.
5596 (direct_scatter_store_optab_supported_p): New macro.
5597 (internal_store_fn_p): New function.
5598 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
5599 IFN_MASK_SCATTER_STORE.
5600 (internal_fn_mask_index): Likewise.
5601 (internal_fn_stored_value_index): New function.
5602 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
5604 * optabs-query.h (supports_vec_scatter_store_p): Declare.
5605 * optabs-query.c (supports_vec_scatter_store_p): New function.
5606 * tree-vectorizer.h (vect_get_store_rhs): Declare.
5607 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
5608 true for scatter stores.
5609 (vect_gather_scatter_fn_p): Handle scatter stores too.
5610 (vect_check_gather_scatter): Consider using scatter stores if
5611 supports_vec_scatter_store_p.
5612 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
5614 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5615 internal_fn_stored_value_index.
5616 (check_load_store_masking): Handle scatter stores too.
5617 (vect_get_store_rhs): Make public.
5618 (vectorizable_call): Use internal_store_fn_p.
5619 (vectorizable_store): Handle scatter store internal functions.
5620 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
5621 when deciding whether the end of the group has been reached.
5622 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
5623 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
5624 (mask_scatter_store<mode>): New insns.
5626 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5627 Alan Hayward <alan.hayward@arm.com>
5628 David Sherwood <david.sherwood@arm.com>
5630 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
5631 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
5632 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
5634 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
5635 Use vect_truncate_gather_scatter_offset if we can't treat the
5636 operation as a normal gather load or scatter store.
5637 (get_group_load_store_type): Take the gather_scatter_info
5638 as argument. Try using a gather load or scatter store for
5639 single-element groups.
5640 (get_load_store_type): Update calls to get_group_load_store_type
5641 and vect_use_strided_gather_scatters_p.
5643 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5644 Alan Hayward <alan.hayward@arm.com>
5645 David Sherwood <david.sherwood@arm.com>
5647 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
5648 optional tree argument.
5649 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
5651 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
5652 but continue to use the current value as a fallback.
5653 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
5654 to compare the updates.
5655 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
5656 (get_load_store_type): Use it when handling a strided access.
5657 (vect_get_strided_load_store_ops): New function.
5658 (vect_get_data_ptr_increment): Likewise.
5659 (vectorizable_load): Handle strided gather loads. Always pass
5660 a step to vect_create_data_ref_ptr and bump_vector_ptr.
5662 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5663 Alan Hayward <alan.hayward@arm.com>
5664 David Sherwood <david.sherwood@arm.com>
5666 * doc/md.texi (gather_load@var{m}): Document.
5667 (mask_gather_load@var{m}): Likewise.
5668 * genopinit.c (main): Add supports_vec_gather_load and
5669 supports_vec_gather_load_cached to target_optabs.
5670 * optabs-tree.c (init_tree_optimization_optabs): Use
5671 ggc_cleared_alloc to allocate target_optabs.
5672 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
5673 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
5675 * internal-fn.h (internal_load_fn_p): Declare.
5676 (internal_gather_scatter_fn_p): Likewise.
5677 (internal_fn_mask_index): Likewise.
5678 (internal_gather_scatter_fn_supported_p): Likewise.
5679 * internal-fn.c (gather_load_direct): New macro.
5680 (expand_gather_load_optab_fn): New function.
5681 (direct_gather_load_optab_supported_p): New macro.
5682 (direct_internal_fn_optab): New function.
5683 (internal_load_fn_p): Likewise.
5684 (internal_gather_scatter_fn_p): Likewise.
5685 (internal_fn_mask_index): Likewise.
5686 (internal_gather_scatter_fn_supported_p): Likewise.
5687 * optabs-query.c (supports_at_least_one_mode_p): New function.
5688 (supports_vec_gather_load_p): Likewise.
5689 * optabs-query.h (supports_vec_gather_load_p): Declare.
5690 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
5691 and memory_type field.
5692 (NUM_PATTERNS): Bump to 15.
5693 * tree-vect-data-refs.c: Include internal-fn.h.
5694 (vect_gather_scatter_fn_p): New function.
5695 (vect_describe_gather_scatter_call): Likewise.
5696 (vect_check_gather_scatter): Try using internal functions for
5697 gather loads. Recognize existing calls to a gather load function.
5698 (vect_analyze_data_refs): Consider using gather loads if
5699 supports_vec_gather_load_p.
5700 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
5701 (vect_get_gather_scatter_offset_type): Likewise.
5702 (vect_convert_mask_for_vectype): Likewise.
5703 (vect_add_conversion_to_patterm): Likewise.
5704 (vect_try_gather_scatter_pattern): Likewise.
5705 (vect_recog_gather_scatter_pattern): New pattern recognizer.
5706 (vect_vect_recog_func_ptrs): Add it.
5707 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
5708 internal_fn_mask_index and internal_gather_scatter_fn_p.
5709 (check_load_store_masking): Take the gather_scatter_info as an
5710 argument and handle gather loads.
5711 (vect_get_gather_scatter_ops): New function.
5712 (vectorizable_call): Check internal_load_fn_p.
5713 (vectorizable_load): Likewise. Handle gather load internal
5715 (vectorizable_store): Update call to check_load_store_masking.
5716 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
5717 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
5718 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
5719 (aarch64_gather_scale_operand_d): New predicates.
5720 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
5721 (mask_gather_load<mode>): New insns.
5723 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5724 Alan Hayward <alan.hayward@arm.com>
5725 David Sherwood <david.sherwood@arm.com>
5727 * optabs.def (fold_left_plus_optab): New optab.
5728 * doc/md.texi (fold_left_plus_@var{m}): Document.
5729 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
5730 * internal-fn.c (fold_left_direct): Define.
5731 (expand_fold_left_optab_fn): Likewise.
5732 (direct_fold_left_optab_supported_p): Likewise.
5733 * fold-const-call.c (fold_const_fold_left): New function.
5734 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
5735 * tree-parloops.c (valid_reduction_p): New function.
5736 (gather_scalar_reductions): Use it.
5737 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
5738 (vect_finish_replace_stmt): Declare.
5739 * tree-vect-loop.c (fold_left_reduction_fn): New function.
5740 (needs_fold_left_reduction_p): New function, split out from...
5741 (vect_is_simple_reduction): ...here. Accept reductions that
5742 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
5743 (vect_force_simple_reduction): Also store the reduction type in
5744 the assignment's STMT_VINFO_REDUC_TYPE.
5745 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
5746 (merge_with_identity): New function.
5747 (vect_expand_fold_left): Likewise.
5748 (vectorize_fold_left_reduction): Likewise.
5749 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
5750 scalar phi in place for it. Check for target support and reject
5751 cases that would reassociate the operation. Defer the transform
5752 phase to vectorize_fold_left_reduction.
5753 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
5754 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
5755 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
5757 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5759 * tree-if-conv.c (predicate_mem_writes): Remove redundant
5760 call to ifc_temp_var.
5762 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5763 Alan Hayward <alan.hayward@arm.com>
5764 David Sherwood <david.sherwood@arm.com>
5766 * target.def (legitimize_address_displacement): Take the original
5767 offset as a poly_int.
5768 * targhooks.h (default_legitimize_address_displacement): Update
5770 * targhooks.c (default_legitimize_address_displacement): Likewise.
5771 * doc/tm.texi: Regenerate.
5772 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
5773 as an argument, moving assert of ad->disp == ad->disp_term to...
5774 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
5775 Try calling targetm.legitimize_address_displacement before expanding
5776 the address rather than afterwards, and adjust for the new interface.
5777 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
5778 Match the new hook interface. Handle SVE addresses.
5779 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
5782 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5784 * Makefile.in (OBJS): Add early-remat.o.
5785 * target.def (select_early_remat_modes): New hook.
5786 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
5787 * doc/tm.texi: Regenerate.
5788 * targhooks.h (default_select_early_remat_modes): Declare.
5789 * targhooks.c (default_select_early_remat_modes): New function.
5790 * timevar.def (TV_EARLY_REMAT): New timevar.
5791 * passes.def (pass_early_remat): New pass.
5792 * tree-pass.h (make_pass_early_remat): Declare.
5793 * early-remat.c: New file.
5794 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
5796 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
5798 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5799 Alan Hayward <alan.hayward@arm.com>
5800 David Sherwood <david.sherwood@arm.com>
5802 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
5803 vfm1 with a bound_epilog parameter.
5804 (vect_do_peeling): Update calls accordingly, and move the prologue
5805 call earlier in the function. Treat the base bound_epilog as 0 for
5806 fully-masked loops and retain vf - 1 for other loops. Add 1 to
5807 this base when peeling for gaps.
5808 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
5809 with fully-masked loops.
5810 (vect_estimate_min_profitable_iters): Handle the single peeled
5811 iteration in that case.
5813 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5814 Alan Hayward <alan.hayward@arm.com>
5815 David Sherwood <david.sherwood@arm.com>
5817 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
5818 single-element interleaving even if the size is not a power of 2.
5819 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
5820 accesses for single-element interleaving if the group size is
5823 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5824 Alan Hayward <alan.hayward@arm.com>
5825 David Sherwood <david.sherwood@arm.com>
5827 * doc/md.texi (fold_extract_last_@var{m}): Document.
5828 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
5829 * optabs.def (fold_extract_last_optab): New optab.
5830 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
5831 * internal-fn.c (fold_extract_direct): New macro.
5832 (expand_fold_extract_optab_fn): Likewise.
5833 (direct_fold_extract_optab_supported_p): Likewise.
5834 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
5835 * tree-vect-loop.c (vect_model_reduction_cost): Handle
5836 EXTRACT_LAST_REDUCTION.
5837 (get_initial_def_for_reduction): Do not create an initial vector
5838 for EXTRACT_LAST_REDUCTION reductions.
5839 (vectorizable_reduction): Leave the scalar phi in place for
5840 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
5841 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
5842 epilogue code for EXTRACT_LAST_REDUCTION and defer the
5843 transform phase to vectorizable_condition.
5844 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
5846 (vect_finish_stmt_generation): ...here.
5847 (vect_finish_replace_stmt): New function.
5848 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
5849 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
5851 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
5853 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5854 Alan Hayward <alan.hayward@arm.com>
5855 David Sherwood <david.sherwood@arm.com>
5857 * doc/md.texi (extract_last_@var{m}): Document.
5858 * optabs.def (extract_last_optab): New optab.
5859 * internal-fn.def (EXTRACT_LAST): New internal function.
5860 * internal-fn.c (cond_unary_direct): New macro.
5861 (expand_cond_unary_optab_fn): Likewise.
5862 (direct_cond_unary_optab_supported_p): Likewise.
5863 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
5864 loops using EXTRACT_LAST.
5865 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
5866 (extract_last_<mode>): ...this optab.
5867 (vec_extract<mode><Vel>): Update accordingly.
5869 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5870 Alan Hayward <alan.hayward@arm.com>
5871 David Sherwood <david.sherwood@arm.com>
5873 * target.def (empty_mask_is_expensive): New hook.
5874 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
5875 * doc/tm.texi: Regenerate.
5876 * targhooks.h (default_empty_mask_is_expensive): Declare.
5877 * targhooks.c (default_empty_mask_is_expensive): New function.
5878 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
5879 if the target says that empty masks are expensive.
5880 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
5882 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
5884 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5885 Alan Hayward <alan.hayward@arm.com>
5886 David Sherwood <david.sherwood@arm.com>
5888 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
5889 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
5890 (vect_use_loop_mask_for_alignment_p): New function.
5891 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
5892 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
5893 niters_skip argument. Make sure that the first niters_skip elements
5894 of the first iteration are inactive.
5895 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
5896 Update call to vect_set_loop_masks_directly.
5897 (get_misalign_in_elems): New function, split out from...
5898 (vect_gen_prolog_loop_niters): ...here.
5899 (vect_update_init_of_dr): Take a code argument that specifies whether
5900 the adjustment should be added or subtracted.
5901 (vect_update_init_of_drs): Likewise.
5902 (vect_prepare_for_masked_peels): New function.
5903 (vect_do_peeling): Skip prologue peeling if we're using a mask
5904 instead. Update call to vect_update_inits_of_drs.
5905 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5907 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
5908 alignment. Do not include the number of peeled iterations in
5909 the minimum threshold in that case.
5910 (vectorizable_induction): Adjust the start value down by
5911 LOOP_VINFO_MASK_SKIP_NITERS iterations.
5912 (vect_transform_loop): Call vect_prepare_for_masked_peels.
5913 Take the number of skipped iterations into account when calculating
5915 * tree-vect-stmts.c (vect_gen_while_not): New function.
5917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5918 Alan Hayward <alan.hayward@arm.com>
5919 David Sherwood <david.sherwood@arm.com>
5921 * doc/sourcebuild.texi (vect_fully_masked): Document.
5922 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
5924 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
5926 (vect_analyze_loop_2): ...here. Don't check the vectorization
5927 factor against the number of loop iterations if the loop is
5930 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5931 Alan Hayward <alan.hayward@arm.com>
5932 David Sherwood <david.sherwood@arm.com>
5934 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
5935 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
5936 (dump_groups): Update accordingly.
5937 (iv_use::mem_type): New member variable.
5938 (address_p): New function.
5939 (record_use): Add a mem_type argument and initialize the new
5941 (record_group_use): Add a mem_type argument. Use address_p.
5942 Remove obsolete null checks of base_object. Update call to record_use.
5943 (find_interesting_uses_op): Update call to record_group_use.
5944 (find_interesting_uses_cond): Likewise.
5945 (find_interesting_uses_address): Likewise.
5946 (get_mem_type_for_internal_fn): New function.
5947 (find_address_like_use): Likewise.
5948 (find_interesting_uses_stmt): Try find_address_like_use before
5949 calling find_interesting_uses_op.
5950 (addr_offset_valid_p): Use the iv mem_type field as the type
5951 of the addressed memory.
5952 (add_autoinc_candidates): Likewise.
5953 (get_address_cost): Likewise.
5954 (split_small_address_groups_p): Use address_p.
5955 (split_address_groups): Likewise.
5956 (add_iv_candidate_for_use): Likewise.
5957 (autoinc_possible_for_pair): Likewise.
5958 (rewrite_groups): Likewise.
5959 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
5960 (determine_group_iv_cost): Update after split of USE_ADDRESS.
5961 (get_alias_ptr_type_for_ptr_address): New function.
5962 (rewrite_use_address): Rewrite address uses in calls that were
5963 identified by find_address_like_use.
5965 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5966 Alan Hayward <alan.hayward@arm.com>
5967 David Sherwood <david.sherwood@arm.com>
5969 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
5971 * gimple-expr.h (is_gimple_addressable: Likewise.
5972 * gimple-expr.c (is_gimple_address): Likewise.
5973 * internal-fn.c (expand_call_mem_ref): New function.
5974 (expand_mask_load_optab_fn): Use it.
5975 (expand_mask_store_optab_fn): Likewise.
5977 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5978 Alan Hayward <alan.hayward@arm.com>
5979 David Sherwood <david.sherwood@arm.com>
5981 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
5982 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
5983 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
5984 (cond_umax@var{mode}): Document.
5985 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
5986 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
5987 (cond_umin_optab, cond_umax_optab): New optabs.
5988 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
5989 (COND_IOR, COND_XOR): New internal functions.
5990 * internal-fn.h (get_conditional_internal_fn): Declare.
5991 * internal-fn.c (cond_binary_direct): New macro.
5992 (expand_cond_binary_optab_fn): Likewise.
5993 (direct_cond_binary_optab_supported_p): Likewise.
5994 (get_conditional_internal_fn): New function.
5995 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
5996 Cope with reduction statements that are vectorized as calls rather
5998 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
5999 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
6000 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
6001 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6002 (UNSPEC_COND_EOR): New unspecs.
6003 (optab): Add mappings for them.
6004 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
6005 (sve_int_op, sve_fp_op): New int attributes.
6007 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6008 Alan Hayward <alan.hayward@arm.com>
6009 David Sherwood <david.sherwood@arm.com>
6011 * optabs.def (while_ult_optab): New optab.
6012 * doc/md.texi (while_ult@var{m}@var{n}): Document.
6013 * internal-fn.def (WHILE_ULT): New internal function.
6014 * internal-fn.h (direct_internal_fn_supported_p): New override
6015 that takes two types as argument.
6016 * internal-fn.c (while_direct): New macro.
6017 (expand_while_optab_fn): New function.
6018 (convert_optab_supported_p): Likewise.
6019 (direct_while_optab_supported_p): New macro.
6020 * wide-int.h (wi::udiv_ceil): New function.
6021 * tree-vectorizer.h (rgroup_masks): New structure.
6022 (vec_loop_masks): New typedef.
6023 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
6025 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
6026 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
6027 (vect_max_vf): New function.
6028 (slpeel_make_loop_iterate_ntimes): Delete.
6029 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
6030 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
6031 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
6032 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
6033 internal-fn.h, stor-layout.h and optabs-query.h.
6034 (vect_set_loop_mask): New function.
6035 (add_preheader_seq): Likewise.
6036 (add_header_seq): Likewise.
6037 (interleave_supported_p): Likewise.
6038 (vect_maybe_permute_loop_masks): Likewise.
6039 (vect_set_loop_masks_directly): Likewise.
6040 (vect_set_loop_condition_masked): Likewise.
6041 (vect_set_loop_condition_unmasked): New function, split out from
6042 slpeel_make_loop_iterate_ntimes.
6043 (slpeel_make_loop_iterate_ntimes): Rename to..
6044 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
6045 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
6046 (vect_do_peeling): Update call accordingly.
6047 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
6049 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6050 mask_compare_type, can_fully_mask_p and fully_masked_p.
6051 (release_vec_loop_masks): New function.
6052 (_loop_vec_info): Use it to free the loop masks.
6053 (can_produce_all_loop_masks_p): New function.
6054 (vect_get_max_nscalars_per_iter): Likewise.
6055 (vect_verify_full_masking): Likewise.
6056 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
6057 retries, and free the mask rgroups before retrying. Check loop-wide
6058 reasons for disallowing fully-masked loops. Make the final decision
6059 about whether use a fully-masked loop or not.
6060 (vect_estimate_min_profitable_iters): Do not assume that peeling
6061 for the number of iterations will be needed for fully-masked loops.
6062 (vectorizable_reduction): Disable fully-masked loops.
6063 (vectorizable_live_operation): Likewise.
6064 (vect_halve_mask_nunits): New function.
6065 (vect_double_mask_nunits): Likewise.
6066 (vect_record_loop_mask): Likewise.
6067 (vect_get_loop_mask): Likewise.
6068 (vect_transform_loop): Handle the case in which the final loop
6069 iteration might handle a partial vector. Call vect_set_loop_condition
6070 instead of slpeel_make_loop_iterate_ntimes.
6071 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
6072 (check_load_store_masking): New function.
6073 (prepare_load_store_mask): Likewise.
6074 (vectorizable_store): Handle fully-masked loops.
6075 (vectorizable_load): Likewise.
6076 (supportable_widening_operation): Use vect_halve_mask_nunits for
6078 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
6079 (vect_gen_while): New function.
6080 * config/aarch64/aarch64.md (umax<mode>3): New expander.
6081 (aarch64_uqdec<mode>): New insn.
6083 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6084 Alan Hayward <alan.hayward@arm.com>
6085 David Sherwood <david.sherwood@arm.com>
6087 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
6088 (reduc_xor_scal_optab): New optabs.
6089 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
6090 (reduc_xor_scal_@var{m}): Document.
6091 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
6092 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
6094 * fold-const-call.c (fold_const_call): Handle them.
6095 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
6096 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
6097 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
6098 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
6099 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
6100 (UNSPEC_XORV): New unspecs.
6101 (optab): Add entries for them.
6102 (BITWISEV): New int iterator.
6103 (bit_reduc_op): New int attributes.
6105 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6106 Alan Hayward <alan.hayward@arm.com>
6107 David Sherwood <david.sherwood@arm.com>
6109 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
6110 * internal-fn.def (VEC_SHL_INSERT): New internal function.
6111 * optabs.def (vec_shl_insert_optab): New optab.
6112 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
6113 (duplicate_and_interleave): Likewise.
6114 * tree-vect-loop.c: Include internal-fn.h.
6115 (neutral_op_for_slp_reduction): New function, split out from
6116 get_initial_defs_for_reduction.
6117 (get_initial_def_for_reduction): Handle option 2 for variable-length
6118 vectors by loading the neutral value into a vector and then shifting
6119 the initial value into element 0.
6120 (get_initial_defs_for_reduction): Replace the code argument with
6121 the neutral value calculated by neutral_op_for_slp_reduction.
6122 Use gimple_build_vector for constant-length vectors.
6123 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
6124 but the first group_size elements have a neutral value.
6125 Use duplicate_and_interleave otherwise.
6126 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
6127 Update call to get_initial_defs_for_reduction. Handle SLP
6128 reductions for variable-length vectors by creating one vector
6129 result for each scalar result, with the elements associated
6130 with other scalar results stubbed out with the neutral value.
6131 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
6132 Require IFN_VEC_SHL_INSERT for double reductions on
6133 variable-length vectors, or SLP reductions that have
6134 a neutral value. Require can_duplicate_and_interleave_p
6135 support for variable-length unchained SLP reductions if there
6136 is no neutral value, such as for MIN/MAX reductions. Also require
6137 the number of vector elements to be a multiple of the number of
6138 SLP statements when doing variable-length unchained SLP reductions.
6139 Update call to vect_create_epilog_for_reduction.
6140 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
6141 and remove initial values.
6142 (duplicate_and_interleave): Make public.
6143 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
6144 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
6146 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6147 Alan Hayward <alan.hayward@arm.com>
6148 David Sherwood <david.sherwood@arm.com>
6150 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
6151 (can_duplicate_and_interleave_p): New function.
6152 (vect_get_and_check_slp_defs): Take the vector of statements
6153 rather than just the current one. Remove excess parentheses.
6154 Restriction rejectinon of vect_constant_def and vect_external_def
6155 for variable-length vectors to boolean types, or types for which
6156 can_duplicate_and_interleave_p is false.
6157 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
6158 (duplicate_and_interleave): New function.
6159 (vect_get_constant_vectors): Use gimple_build_vector for
6160 constant-length vectors and suitable variable-length constant
6161 vectors. Use duplicate_and_interleave for other variable-length
6162 vectors. Don't defer the update when inserting new statements.
6164 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6165 Alan Hayward <alan.hayward@arm.com>
6166 David Sherwood <david.sherwood@arm.com>
6168 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
6169 min_profitable_iters doesn't go negative.
6171 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6172 Alan Hayward <alan.hayward@arm.com>
6173 David Sherwood <david.sherwood@arm.com>
6175 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
6176 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
6177 * optabs.def (vec_mask_load_lanes_optab): New optab.
6178 (vec_mask_store_lanes_optab): Likewise.
6179 * internal-fn.def (MASK_LOAD_LANES): New internal function.
6180 (MASK_STORE_LANES): Likewise.
6181 * internal-fn.c (mask_load_lanes_direct): New macro.
6182 (mask_store_lanes_direct): Likewise.
6183 (expand_mask_load_optab_fn): Handle masked operations.
6184 (expand_mask_load_lanes_optab_fn): New macro.
6185 (expand_mask_store_optab_fn): Handle masked operations.
6186 (expand_mask_store_lanes_optab_fn): New macro.
6187 (direct_mask_load_lanes_optab_supported_p): Likewise.
6188 (direct_mask_store_lanes_optab_supported_p): Likewise.
6189 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
6191 (vect_load_lanes_supported): Likewise.
6192 * tree-vect-data-refs.c (strip_conversion): New function.
6193 (can_group_stmts_p): Likewise.
6194 (vect_analyze_data_ref_accesses): Use it instead of checking
6195 for a pair of assignments.
6196 (vect_store_lanes_supported): Take a masked_p parameter.
6197 (vect_load_lanes_supported): Likewise.
6198 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
6199 vect_store_lanes_supported and vect_load_lanes_supported.
6200 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6201 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
6202 parameter. Don't allow gaps for masked accesses.
6203 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
6204 and vect_load_lanes_supported.
6205 (get_load_store_type): Take a masked_p parameter and update
6206 call to get_group_load_store_type.
6207 (vectorizable_store): Update call to get_load_store_type.
6208 Handle IFN_MASK_STORE_LANES.
6209 (vectorizable_load): Update call to get_load_store_type.
6210 Handle IFN_MASK_LOAD_LANES.
6212 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6213 Alan Hayward <alan.hayward@arm.com>
6214 David Sherwood <david.sherwood@arm.com>
6216 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
6218 * config/aarch64/aarch64-protos.h
6219 (aarch64_sve_struct_memory_operand_p): Declare.
6220 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
6221 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
6222 (VPRED, vpred): Handle SVE structure modes.
6223 * config/aarch64/constraints.md (Utx): New constraint.
6224 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
6225 (aarch64_sve_struct_nonimmediate_operand): New predicates.
6226 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
6227 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
6228 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
6229 structure modes. Split into pieces after RA.
6230 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
6231 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
6233 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
6234 SVE structure modes.
6235 (aarch64_classify_address): Likewise.
6236 (sizetochar): Move earlier in file.
6237 (aarch64_print_operand): Handle SVE register lists.
6238 (aarch64_array_mode): New function.
6239 (aarch64_sve_struct_memory_operand_p): Likewise.
6240 (TARGET_ARRAY_MODE): Redefine.
6242 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6243 Alan Hayward <alan.hayward@arm.com>
6244 David Sherwood <david.sherwood@arm.com>
6246 * target.def (array_mode): New target hook.
6247 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
6248 * doc/tm.texi: Regenerate.
6249 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
6250 * hooks.c (hook_optmode_mode_uhwi_none): New function.
6251 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
6253 * stor-layout.c (mode_for_array): Likewise. Support polynomial
6256 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6257 Alan Hayward <alan.hayward@arm.com>
6258 David Sherwood <david.sherwood@arm.com>
6260 * fold-const.c (fold_binary_loc): Check the argument types
6261 rather than the result type when testing for a vector operation.
6263 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6265 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
6266 * doc/tm.texi: Regenerate.
6268 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6269 Alan Hayward <alan.hayward@arm.com>
6270 David Sherwood <david.sherwood@arm.com>
6272 * doc/invoke.texi (-msve-vector-bits=): Document new option.
6273 (sve): Document new AArch64 extension.
6274 * doc/md.texi (w): Extend the description of the AArch64
6275 constraint to include SVE vectors.
6276 (Upl, Upa): Document new AArch64 predicate constraints.
6277 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
6279 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
6280 (msve-vector-bits=): New option.
6281 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
6282 SVE when these are disabled.
6283 (sve): New extension.
6284 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
6285 modes. Adjust their number of units based on aarch64_sve_vg.
6286 (MAX_BITSIZE_MODE_ANY_MODE): Define.
6287 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
6288 aarch64_addr_query_type.
6289 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
6290 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
6291 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
6292 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
6293 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
6294 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
6295 (aarch64_simd_imm_zero_p): Delete.
6296 (aarch64_check_zero_based_sve_index_immediate): Declare.
6297 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6298 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6299 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6300 (aarch64_sve_float_mul_immediate_p): Likewise.
6301 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6303 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
6304 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
6305 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
6306 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
6307 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
6308 (aarch64_regmode_natural_size): Likewise.
6309 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
6310 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
6312 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
6313 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
6314 for VG and the SVE predicate registers.
6315 (V_ALIASES): Add a "z"-prefixed alias.
6316 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
6317 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
6318 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
6319 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
6320 (REG_CLASS_NAMES): Add entries for them.
6321 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
6322 and the predicate registers.
6323 (aarch64_sve_vg): Declare.
6324 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
6325 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
6326 (REGMODE_NATURAL_SIZE): Define.
6327 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
6329 * config/aarch64/aarch64.c: Include cfgrtl.h.
6330 (simd_immediate_info): Add a constructor for series vectors,
6331 and an associated step field.
6332 (aarch64_sve_vg): New variable.
6333 (aarch64_dbx_register_number): Handle VG and the predicate registers.
6334 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
6335 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
6336 (VEC_ANY_DATA, VEC_STRUCT): New constants.
6337 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
6338 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
6339 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
6340 (aarch64_get_mask_mode): New functions.
6341 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
6342 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
6343 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
6344 predicate modes and predicate registers. Explicitly restrict
6345 GPRs to modes of 16 bytes or smaller. Only allow FP registers
6346 to store a vector mode if it is recognized by
6347 aarch64_classify_vector_mode.
6348 (aarch64_regmode_natural_size): New function.
6349 (aarch64_hard_regno_caller_save_mode): Return the original mode
6351 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
6352 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
6353 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
6354 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
6356 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
6357 does not overlap dest if the function is frame-related. Handle
6359 (aarch64_split_add_offset): New function.
6360 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
6361 them aarch64_add_offset.
6362 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
6363 and update call to aarch64_sub_sp.
6364 (aarch64_add_cfa_expression): New function.
6365 (aarch64_expand_prologue): Pass extra temporary registers to the
6366 functions above. Handle the case in which we need to emit new
6367 DW_CFA_expressions for registers that were originally saved
6368 relative to the stack pointer, but now have to be expressed
6369 relative to the frame pointer.
6370 (aarch64_output_mi_thunk): Pass extra temporary registers to the
6372 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
6373 IP0 and IP1 values for SVE frames.
6374 (aarch64_expand_vec_series): New function.
6375 (aarch64_expand_sve_widened_duplicate): Likewise.
6376 (aarch64_expand_sve_const_vector): Likewise.
6377 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
6378 Handle SVE constants. Use emit_move_insn to move a force_const_mem
6379 into the register, rather than emitting a SET directly.
6380 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
6381 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
6382 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
6383 (offset_9bit_signed_scaled_p): New functions.
6384 (aarch64_replicate_bitmask_imm): New function.
6385 (aarch64_bitmask_imm): Use it.
6386 (aarch64_cannot_force_const_mem): Reject expressions involving
6387 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
6388 (aarch64_classify_index): Handle SVE indices, by requiring
6389 a plain register index with a scale that matches the element size.
6390 (aarch64_classify_address): Handle SVE addresses. Assert that
6391 the mode of the address is VOIDmode or an integer mode.
6392 Update call to aarch64_classify_symbol.
6393 (aarch64_classify_symbolic_expression): Update call to
6394 aarch64_classify_symbol.
6395 (aarch64_const_vec_all_in_range_p): New function.
6396 (aarch64_print_vector_float_operand): Likewise.
6397 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
6398 "vN" for FP registers with SVE modes. Handle (const ...) vectors
6399 and the FP immediates 1.0 and 0.5.
6400 (aarch64_print_address_internal): Handle SVE addresses.
6401 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
6402 (aarch64_regno_regclass): Handle predicate registers.
6403 (aarch64_secondary_reload): Handle big-endian reloads of SVE
6405 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
6406 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
6407 (aarch64_convert_sve_vector_bits): New function.
6408 (aarch64_override_options): Use it to handle -msve-vector-bits=.
6409 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6411 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
6412 Handle SVE vector and predicate modes. Accept VL-based constants
6413 that need only one temporary register, and VL offsets that require
6414 no temporary registers.
6415 (aarch64_conditional_register_usage): Mark the predicate registers
6416 as fixed if SVE isn't available.
6417 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
6418 Return true for SVE vector and predicate modes.
6419 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
6420 rather than an unsigned int. Handle SVE modes.
6421 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6423 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6425 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6426 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6427 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6428 (aarch64_sve_float_mul_immediate_p): New functions.
6429 (aarch64_sve_valid_immediate): New function.
6430 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
6431 Explicitly reject structure modes. Check for INDEX constants.
6432 Handle PTRUE and PFALSE constants.
6433 (aarch64_check_zero_based_sve_index_immediate): New function.
6434 (aarch64_simd_imm_zero_p): Delete.
6435 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
6436 vector modes. Accept constants in the range of CNT[BHWD].
6437 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
6438 ask for an Advanced SIMD mode.
6439 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
6440 (aarch64_simd_vector_alignment): Handle SVE predicates.
6441 (aarch64_vectorize_preferred_vector_alignment): New function.
6442 (aarch64_simd_vector_alignment_reachable): Use it instead of
6444 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
6445 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
6447 (MAX_VECT_LEN): Delete.
6448 (expand_vec_perm_d): Add a vec_flags field.
6449 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
6450 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
6451 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6453 (aarch64_evpc_rev): Rename to...
6454 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6455 (aarch64_evpc_rev_global): New function.
6456 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6457 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6459 (aarch64_evpc_sve_tbl): New function.
6460 (aarch64_expand_vec_perm_const_1): Update after rename of
6461 aarch64_evpc_rev. Handle SVE permutes too, trying
6462 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6463 than aarch64_evpc_tbl.
6464 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6465 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6466 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6467 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6468 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6469 (aarch64_expand_sve_vcond): New functions.
6470 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6471 of aarch64_vector_mode_p.
6472 (aarch64_dwarf_poly_indeterminate_value): New function.
6473 (aarch64_compute_pressure_classes): Likewise.
6474 (aarch64_can_change_mode_class): Likewise.
6475 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6476 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6477 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6478 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6479 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6480 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6481 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6482 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6484 (Dn, Dl, Dr): Accept const as well as const_vector.
6485 (Dz): Likewise. Compare against CONST0_RTX.
6486 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6487 of "vector" where appropriate.
6488 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6489 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6490 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6491 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6492 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6493 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6494 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6495 (v_int_equiv): Extend to SVE modes.
6496 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6498 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6499 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6500 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6501 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6502 (SVE_COND_FP_CMP): New int iterators.
6503 (perm_hilo): Handle the new unpack unspecs.
6504 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6506 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6507 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6508 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6509 (aarch64_equality_operator, aarch64_constant_vector_operand)
6510 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
6511 (aarch64_sve_nonimmediate_operand): Likewise.
6512 (aarch64_sve_general_operand): Likewise.
6513 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
6514 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
6515 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
6516 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
6517 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
6518 (aarch64_sve_float_arith_immediate): Likewise.
6519 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
6520 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
6521 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
6522 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
6523 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
6524 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
6525 (aarch64_sve_float_arith_operand): Likewise.
6526 (aarch64_sve_float_arith_with_sub_operand): Likewise.
6527 (aarch64_sve_float_mul_operand): Likewise.
6528 (aarch64_sve_vec_perm_operand): Likewise.
6529 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
6530 (aarch64_mov_operand): Accept const_poly_int and const_vector.
6531 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
6532 as well as const_vector.
6533 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
6534 in file. Use CONST0_RTX and CONSTM1_RTX.
6535 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
6536 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
6537 Use aarch64_simd_imm_zero.
6538 * config/aarch64/aarch64-sve.md: New file.
6539 * config/aarch64/aarch64.md: Include it.
6540 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
6541 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
6542 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
6543 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
6544 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
6545 (sve): New attribute.
6546 (enabled): Disable instructions with the sve attribute unless
6548 (movqi, movhi): Pass CONST_POLY_INT operaneds through
6549 aarch64_expand_mov_immediate.
6550 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
6551 CNT[BHSD] immediates.
6552 (movti): Split CONST_POLY_INT moves into two halves.
6553 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
6554 Split additions that need a temporary here if the destination
6555 is the stack pointer.
6556 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
6557 (*add<mode>3_poly_1): New instruction.
6558 (set_clobber_cc): New expander.
6560 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6562 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
6563 parameter and use it instead of GET_MODE_SIZE (innermode). Use
6564 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
6565 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
6566 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
6567 Change innermode from fixed_mode_size to machine_mode.
6568 (simplify_subreg): Update call accordingly. Handle a constant-sized
6569 subreg of a variable-length CONST_VECTOR.
6571 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6572 Alan Hayward <alan.hayward@arm.com>
6573 David Sherwood <david.sherwood@arm.com>
6575 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
6576 (add_offset_to_base): New function, split out from...
6577 (create_mem_ref): ...here. When handling a scale other than 1,
6578 check first whether the address is valid without the offset.
6579 Add it into the base if so, leaving the index and scale as-is.
6581 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6584 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
6585 fold_for_warn before checking if arg2 is INTEGER_CST.
6587 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
6589 * config/rs6000/predicates.md (load_multiple_operation): Delete.
6590 (store_multiple_operation): Delete.
6591 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
6592 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
6593 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
6594 guarded by TARGET_STRING.
6595 (rs6000_output_load_multiple): Delete.
6596 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
6597 OPTION_MASK_STRING / TARGET_STRING handling.
6598 (print_operand) <'N', 'O'>: Add comment that these are unused now.
6599 (const rs6000_opt_masks) <"string">: Change mask to 0.
6600 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
6601 (MASK_STRING): Delete.
6602 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
6604 (load_multiple): Delete.
6611 (store_multiple): Delete.
6618 (movmemsi_8reg): Delete.
6619 (corresponding unnamed define_insn): Delete.
6620 (movmemsi_6reg): Delete.
6621 (corresponding unnamed define_insn): Delete.
6622 (movmemsi_4reg): Delete.
6623 (corresponding unnamed define_insn): Delete.
6624 (movmemsi_2reg): Delete.
6625 (corresponding unnamed define_insn): Delete.
6626 (movmemsi_1reg): Delete.
6627 (corresponding unnamed define_insn): Delete.
6628 * config/rs6000/rs6000.opt (mno-string): New.
6629 (mstring): Replace by deprecation warning stub.
6630 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
6632 2018-01-12 Jakub Jelinek <jakub@redhat.com>
6634 * regrename.c (regrename_do_replace): If replacing the same
6635 reg multiple times, try to reuse last created gen_raw_REG.
6638 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
6639 main to workaround a bug in GDB.
6641 2018-01-12 Tom de Vries <tom@codesourcery.com>
6644 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
6646 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
6648 PR rtl-optimization/80481
6649 * ira-color.c (get_cap_member): New function.
6650 (allocnos_conflict_by_live_ranges_p): Use it.
6651 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
6652 (setup_slot_coalesced_allocno_live_ranges): Ditto.
6654 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
6657 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
6658 (*saddl_se_1): Ditto.
6660 (*ssubl_se_1): Ditto.
6662 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6664 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
6665 rather than wi::to_widest for DR_INITs.
6666 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
6667 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
6668 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
6670 (vect_analyze_group_access_1): Note that here.
6672 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6674 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
6675 polynomial type sizes.
6677 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
6679 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
6680 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
6681 (gimple_add_tmp_var): Likewise.
6683 2018-01-12 Martin Liska <mliska@suse.cz>
6685 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
6686 (gimple_alloc_sizes): Likewise.
6687 (dump_gimple_statistics): Use PRIu64 in printf format.
6688 * gimple.h: Change uint64_t to int.
6690 2018-01-12 Martin Liska <mliska@suse.cz>
6692 * tree-core.h: Use uint64_t instead of int.
6693 * tree.c (tree_node_counts): Likewise.
6694 (tree_node_sizes): Likewise.
6695 (dump_tree_statistics): Use PRIu64 in printf format.
6697 2018-01-12 Martin Liska <mliska@suse.cz>
6699 * Makefile.in: As qsort_chk is implemented in vec.c, add
6700 vec.o to linkage of gencfn-macros.
6701 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
6702 passing the info to record_node_allocation_statistics.
6703 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
6705 * ggc-common.c (struct ggc_usage): Add operator== and use
6706 it in operator< and compare function.
6707 * mem-stats.h (struct mem_usage): Likewise.
6708 * vec.c (struct vec_usage): Remove operator< and compare
6709 function. Can be simply inherited.
6711 2018-01-12 Martin Jambor <mjambor@suse.cz>
6714 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
6715 * tree-ssa-math-opts.c: Include domwalk.h.
6716 (convert_mult_to_fma_1): New function.
6717 (fma_transformation_info): New type.
6718 (fma_deferring_state): Likewise.
6719 (cancel_fma_deferring): New function.
6720 (result_of_phi): Likewise.
6721 (last_fma_candidate_feeds_initial_phi): Likewise.
6722 (convert_mult_to_fma): Added deferring logic, split actual
6723 transformation to convert_mult_to_fma_1.
6724 (math_opts_dom_walker): New type.
6725 (math_opts_dom_walker::after_dom_children): New method, body moved
6726 here from pass_optimize_widening_mul::execute, added deferring logic
6728 (pass_optimize_widening_mul::execute): Moved most of code to
6729 math_opts_dom_walker::after_dom_children.
6730 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
6731 * config/i386/i386.c (ix86_option_override_internal): Added
6732 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
6734 2018-01-12 Richard Biener <rguenther@suse.de>
6737 * dwarf2out.c (gen_variable_die): Do not reset old_die for
6738 inline instance vars.
6740 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
6743 * config/rx/rx.c (rx_is_restricted_memory_address):
6746 2018-01-12 Richard Biener <rguenther@suse.de>
6748 PR tree-optimization/80846
6749 * target.def (split_reduction): New target hook.
6750 * targhooks.c (default_split_reduction): New function.
6751 * targhooks.h (default_split_reduction): Declare.
6752 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
6753 target requests first reduce vectors by combining low and high
6755 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
6756 (get_vectype_for_scalar_type_and_size): Export.
6757 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
6758 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
6759 * doc/tm.texi: Regenerate.
6760 * config/i386/i386.c (ix86_split_reduction): Implement
6761 TARGET_VECTORIZE_SPLIT_REDUCTION.
6763 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6766 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
6767 in PIC mode except for TARGET_VXWORKS_RTP.
6768 * config/sparc/sparc.c: Include cfgrtl.h.
6769 (TARGET_INIT_PIC_REG): Define.
6770 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
6771 (sparc_pic_register_p): New predicate.
6772 (sparc_legitimate_address_p): Use it.
6773 (sparc_legitimize_pic_address): Likewise.
6774 (sparc_delegitimize_address): Likewise.
6775 (sparc_mode_dependent_address_p): Likewise.
6776 (gen_load_pcrel_sym): Remove 4th parameter.
6777 (load_got_register): Adjust call to above. Remove obsolete stuff.
6778 (sparc_expand_prologue): Do not call load_got_register here.
6779 (sparc_flat_expand_prologue): Likewise.
6780 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
6781 (sparc_use_pseudo_pic_reg): New function.
6782 (sparc_init_pic_reg): Likewise.
6783 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6784 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
6786 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
6788 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
6789 Add item for branch_cost.
6791 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6793 PR rtl-optimization/83565
6794 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
6795 not extend the result to a larger mode for rotate operations.
6796 (num_sign_bit_copies1): Likewise.
6798 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6801 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
6803 Use values-Xc.o for -pedantic.
6804 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
6806 2018-01-12 Martin Liska <mliska@suse.cz>
6809 * ipa-devirt.c (final_warning_record::grow_type_warnings):
6811 (possible_polymorphic_call_targets): Use it.
6812 (ipa_devirt): Likewise.
6814 2018-01-12 Martin Liska <mliska@suse.cz>
6816 * profile-count.h (enum profile_quality): Use 0 as invalid
6817 enum value of profile_quality.
6819 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
6821 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
6822 -mext-string options.
6824 2018-01-12 Richard Biener <rguenther@suse.de>
6826 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
6827 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
6828 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
6830 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
6832 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6834 * configure.ac (--with-long-double-format): Add support for the
6835 configuration option to change the default long double format on
6837 * config.gcc (powerpc*-linux*-*): Likewise.
6838 * configure: Regenerate.
6839 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
6840 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
6841 used without modification.
6843 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6845 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
6846 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
6847 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
6848 MISC_BUILTIN_SPEC_BARRIER.
6849 (rs6000_init_builtins): Likewise.
6850 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
6852 (speculation_barrier): New define_insn.
6853 * doc/extend.texi: Document __builtin_speculation_barrier.
6855 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6858 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
6859 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
6860 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
6862 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
6863 integral modes instead of "ss" and "sd".
6864 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
6865 vectors with 32-bit and 64-bit elements.
6866 (vecdupssescalarmodesuffix): New mode attribute.
6867 (vec_dup<mode>): Use it.
6869 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
6872 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
6873 frame if argument is passed on stack.
6875 2018-01-11 Jakub Jelinek <jakub@redhat.com>
6878 * ree.c (combine_reaching_defs): Optimize also
6879 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
6880 reg2=any_extend(exp); reg1=reg2;, formatting fix.
6882 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6885 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
6887 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
6890 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
6891 after they are computed.
6893 2018-01-11 Bin Cheng <bin.cheng@arm.com>
6895 PR tree-optimization/83695
6896 * gimple-loop-linterchange.cc
6897 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
6898 reset cached scev information after interchange.
6899 (pass_linterchange::execute): Remove call to scev_reset_htab.
6901 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6903 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
6904 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
6905 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
6906 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
6907 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
6908 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
6909 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
6910 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
6911 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
6912 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
6913 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
6914 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
6915 (V_lane_reg): Likewise.
6916 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
6918 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
6919 (vfmal_lane_low<mode>_intrinsic,
6920 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
6921 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
6922 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
6923 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
6924 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
6925 vfmsl_lane_high<mode>_intrinsic): New define_insns.
6927 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6929 * config/arm/arm-cpus.in (fp16fml): New feature.
6930 (ALL_SIMD): Add fp16fml.
6931 (armv8.2-a): Add fp16fml as an option.
6932 (armv8.3-a): Likewise.
6933 (armv8.4-a): Add fp16fml as part of fp16.
6934 * config/arm/arm.h (TARGET_FP16FML): Define.
6935 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
6937 * config/arm/arm-modes.def (V2HF): Define.
6938 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
6939 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
6940 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
6941 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
6942 vfmsl_low, vfmsl_high): New set of builtins.
6943 * config/arm/iterators.md (PLUSMINUS): New code iterator.
6944 (vfml_op): New code attribute.
6945 (VFMLHALVES): New int iterator.
6946 (VFML, VFMLSEL): New mode attributes.
6947 (V_reg): Define mapping for V2HF.
6948 (V_hi, V_lo): New mode attributes.
6949 (VF_constraint): Likewise.
6950 (vfml_half, vfml_half_selector): New int attributes.
6951 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
6953 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
6954 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
6956 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
6957 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
6958 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
6959 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
6961 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
6962 Document new effective target and option set.
6964 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6966 * config/arm/arm-cpus.in (armv8_4): New feature.
6967 (ARMv8_4a): New fgroup.
6968 (armv8.4-a): New arch.
6969 * config/arm/arm-tables.opt: Regenerate.
6970 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
6971 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
6972 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
6973 Add matching rules for -march=armv8.4-a and extensions.
6974 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
6976 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
6979 * config/rx/rx.md (BW): New mode attribute.
6980 (sync_lock_test_and_setsi): Add mode suffix to insn output.
6982 2018-01-11 Richard Biener <rguenther@suse.de>
6984 PR tree-optimization/83435
6985 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
6986 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
6987 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
6989 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6990 Alan Hayward <alan.hayward@arm.com>
6991 David Sherwood <david.sherwood@arm.com>
6993 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
6995 (aarch64_classify_address): Initialize it. Track polynomial offsets.
6996 (aarch64_print_address_internal): Use it to check for a zero offset.
6998 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
6999 Alan Hayward <alan.hayward@arm.com>
7000 David Sherwood <david.sherwood@arm.com>
7002 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
7003 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
7004 Return a poly_int64 rather than a HOST_WIDE_INT.
7005 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
7006 rather than a HOST_WIDE_INT.
7007 * config/aarch64/aarch64.h (aarch64_frame): Protect with
7008 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
7009 hard_fp_offset, frame_size, initial_adjust, callee_offset and
7010 final_offset from HOST_WIDE_INT to poly_int64.
7011 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
7012 to_constant when getting the number of units in an Advanced SIMD
7014 (aarch64_builtin_vectorized_function): Check for a constant number
7016 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
7018 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
7019 attribute instead of GET_MODE_NUNITS.
7020 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
7021 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
7022 GET_MODE_SIZE for fixed-size registers.
7023 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
7024 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
7025 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
7026 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
7027 (aarch64_print_operand, aarch64_print_address_internal)
7028 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
7029 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
7030 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
7031 Handle polynomial GET_MODE_SIZE.
7032 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
7033 wider than SImode without modification.
7034 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
7035 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
7036 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
7037 passing and returning SVE modes.
7038 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
7039 rather than GEN_INT.
7040 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
7041 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
7042 (aarch64_allocate_and_probe_stack_space): Likewise.
7043 (aarch64_layout_frame): Cope with polynomial offsets.
7044 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
7045 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
7047 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
7048 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
7049 poly_int64 rather than a HOST_WIDE_INT.
7050 (aarch64_get_separate_components, aarch64_process_components)
7051 (aarch64_expand_prologue, aarch64_expand_epilogue)
7052 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
7053 (aarch64_anchor_offset): New function, split out from...
7054 (aarch64_legitimize_address): ...here.
7055 (aarch64_builtin_vectorization_cost): Handle polynomial
7056 TYPE_VECTOR_SUBPARTS.
7057 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
7059 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
7060 number of elements from the PARALLEL rather than the mode.
7061 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
7062 rather than GET_MODE_BITSIZE.
7063 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
7064 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
7065 (aarch64_expand_vec_perm_const_1): Handle polynomial
7066 d->perm.length () and d->perm elements.
7067 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
7068 Apply to_constant to d->perm elements.
7069 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
7070 polynomial CONST_VECTOR_NUNITS.
7071 (aarch64_move_pointer): Take amount as a poly_int64 rather
7073 (aarch64_progress_pointer): Avoid temporary variable.
7074 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
7075 the mode attribute instead of GET_MODE.
7077 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7078 Alan Hayward <alan.hayward@arm.com>
7079 David Sherwood <david.sherwood@arm.com>
7081 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
7082 x exists before using it.
7083 (aarch64_add_constant_internal): Rename to...
7084 (aarch64_add_offset_1): ...this. Replace regnum with separate
7085 src and dest rtxes. Handle the case in which they're different,
7086 including when the offset is zero. Replace scratchreg with an rtx.
7087 Use 2 additions if there is no spare register into which we can
7088 move a 16-bit constant.
7089 (aarch64_add_constant): Delete.
7090 (aarch64_add_offset): Replace reg with separate src and dest
7091 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
7092 Use aarch64_add_offset_1.
7093 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
7094 an rtx rather than an int. Take the delta as a poly_int64
7095 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
7096 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
7097 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
7098 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
7099 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
7101 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
7102 aarch64_add_constant.
7104 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7106 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
7107 Use scalar_float_mode.
7109 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7111 * config/aarch64/aarch64-simd.md
7112 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
7113 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
7114 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
7115 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
7116 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
7117 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
7118 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
7119 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
7120 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
7121 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
7123 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7126 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
7127 targ_options->x_arm_arch_string is non NULL.
7129 2018-01-11 Tamar Christina <tamar.christina@arm.com>
7131 * config/aarch64/aarch64.h
7132 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
7134 2018-01-11 Sudakshina Das <sudi.das@arm.com>
7137 * expmed.c (emit_store_flag_force): Swap if const op0
7138 and change VOIDmode to mode of op0.
7140 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7142 PR rtl-optimization/83761
7143 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
7144 than bytes to mode_for_size.
7146 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7149 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
7150 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
7153 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7156 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
7157 when in layout mode.
7158 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
7159 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
7162 2018-01-10 Michael Collison <michael.collison@arm.com>
7164 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
7165 * config/aarch64/aarch64-option-extension.def: Add
7166 AARCH64_OPT_EXTENSION of 'fp16fml'.
7167 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7168 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
7169 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
7170 * config/aarch64/constraints.md (Ui7): New constraint.
7171 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
7172 (VFMLA_SEL_W): Ditto.
7175 (VFMLA16_LOW): New int iterator.
7176 (VFMLA16_HIGH): Ditto.
7177 (UNSPEC_FMLAL): New unspec.
7178 (UNSPEC_FMLSL): Ditto.
7179 (UNSPEC_FMLAL2): Ditto.
7180 (UNSPEC_FMLSL2): Ditto.
7181 (f16mac): New code attribute.
7182 * config/aarch64/aarch64-simd-builtins.def
7183 (aarch64_fmlal_lowv2sf): Ditto.
7184 (aarch64_fmlsl_lowv2sf): Ditto.
7185 (aarch64_fmlalq_lowv4sf): Ditto.
7186 (aarch64_fmlslq_lowv4sf): Ditto.
7187 (aarch64_fmlal_highv2sf): Ditto.
7188 (aarch64_fmlsl_highv2sf): Ditto.
7189 (aarch64_fmlalq_highv4sf): Ditto.
7190 (aarch64_fmlslq_highv4sf): Ditto.
7191 (aarch64_fmlal_lane_lowv2sf): Ditto.
7192 (aarch64_fmlsl_lane_lowv2sf): Ditto.
7193 (aarch64_fmlal_laneq_lowv2sf): Ditto.
7194 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
7195 (aarch64_fmlalq_lane_lowv4sf): Ditto.
7196 (aarch64_fmlsl_lane_lowv4sf): Ditto.
7197 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
7198 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
7199 (aarch64_fmlal_lane_highv2sf): Ditto.
7200 (aarch64_fmlsl_lane_highv2sf): Ditto.
7201 (aarch64_fmlal_laneq_highv2sf): Ditto.
7202 (aarch64_fmlsl_laneq_highv2sf): Ditto.
7203 (aarch64_fmlalq_lane_highv4sf): Ditto.
7204 (aarch64_fmlsl_lane_highv4sf): Ditto.
7205 (aarch64_fmlalq_laneq_highv4sf): Ditto.
7206 (aarch64_fmlsl_laneq_highv4sf): Ditto.
7207 * config/aarch64/aarch64-simd.md:
7208 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
7209 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7210 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
7211 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7212 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
7213 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
7214 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
7215 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
7216 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
7217 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
7218 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
7219 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
7220 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
7221 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
7222 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
7223 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
7224 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
7225 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
7226 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
7227 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
7228 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
7229 (vfmlsl_low_u32): Ditto.
7230 (vfmlalq_low_u32): Ditto.
7231 (vfmlslq_low_u32): Ditto.
7232 (vfmlal_high_u32): Ditto.
7233 (vfmlsl_high_u32): Ditto.
7234 (vfmlalq_high_u32): Ditto.
7235 (vfmlslq_high_u32): Ditto.
7236 (vfmlal_lane_low_u32): Ditto.
7237 (vfmlsl_lane_low_u32): Ditto.
7238 (vfmlal_laneq_low_u32): Ditto.
7239 (vfmlsl_laneq_low_u32): Ditto.
7240 (vfmlalq_lane_low_u32): Ditto.
7241 (vfmlslq_lane_low_u32): Ditto.
7242 (vfmlalq_laneq_low_u32): Ditto.
7243 (vfmlslq_laneq_low_u32): Ditto.
7244 (vfmlal_lane_high_u32): Ditto.
7245 (vfmlsl_lane_high_u32): Ditto.
7246 (vfmlal_laneq_high_u32): Ditto.
7247 (vfmlsl_laneq_high_u32): Ditto.
7248 (vfmlalq_lane_high_u32): Ditto.
7249 (vfmlslq_lane_high_u32): Ditto.
7250 (vfmlalq_laneq_high_u32): Ditto.
7251 (vfmlslq_laneq_high_u32): Ditto.
7252 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
7253 (AARCH64_FL_FOR_ARCH8_4): New.
7254 (AARCH64_ISA_F16FML): New ISA flag.
7255 (TARGET_F16FML): New feature flag for fp16fml.
7256 (doc/invoke.texi): Document new fp16fml option.
7258 2018-01-10 Michael Collison <michael.collison@arm.com>
7260 * config/aarch64/aarch64-builtins.c:
7261 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
7262 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7263 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
7264 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
7265 (AARCH64_ISA_SHA3): New ISA flag.
7266 (TARGET_SHA3): New feature flag for sha3.
7267 * config/aarch64/iterators.md (sha512_op): New int attribute.
7268 (CRYPTO_SHA512): New int iterator.
7269 (UNSPEC_SHA512H): New unspec.
7270 (UNSPEC_SHA512H2): Ditto.
7271 (UNSPEC_SHA512SU0): Ditto.
7272 (UNSPEC_SHA512SU1): Ditto.
7273 * config/aarch64/aarch64-simd-builtins.def
7274 (aarch64_crypto_sha512hqv2di): New builtin.
7275 (aarch64_crypto_sha512h2qv2di): Ditto.
7276 (aarch64_crypto_sha512su0qv2di): Ditto.
7277 (aarch64_crypto_sha512su1qv2di): Ditto.
7278 (aarch64_eor3qv8hi): Ditto.
7279 (aarch64_rax1qv2di): Ditto.
7280 (aarch64_xarqv2di): Ditto.
7281 (aarch64_bcaxqv8hi): Ditto.
7282 * config/aarch64/aarch64-simd.md:
7283 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
7284 (aarch64_crypto_sha512su0qv2di): Ditto.
7285 (aarch64_crypto_sha512su1qv2di): Ditto.
7286 (aarch64_eor3qv8hi): Ditto.
7287 (aarch64_rax1qv2di): Ditto.
7288 (aarch64_xarqv2di): Ditto.
7289 (aarch64_bcaxqv8hi): Ditto.
7290 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
7291 (vsha512h2q_u64): Ditto.
7292 (vsha512su0q_u64): Ditto.
7293 (vsha512su1q_u64): Ditto.
7294 (veor3q_u16): Ditto.
7295 (vrax1q_u64): Ditto.
7297 (vbcaxq_u16): Ditto.
7298 * config/arm/types.md (crypto_sha512): New type attribute.
7299 (crypto_sha3): Ditto.
7300 (doc/invoke.texi): Document new sha3 option.
7302 2018-01-10 Michael Collison <michael.collison@arm.com>
7304 * config/aarch64/aarch64-builtins.c:
7305 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
7306 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7307 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
7308 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
7309 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
7310 (AARCH64_ISA_SM4): New ISA flag.
7311 (TARGET_SM4): New feature flag for sm4.
7312 * config/aarch64/aarch64-simd-builtins.def
7313 (aarch64_sm3ss1qv4si): Ditto.
7314 (aarch64_sm3tt1aq4si): Ditto.
7315 (aarch64_sm3tt1bq4si): Ditto.
7316 (aarch64_sm3tt2aq4si): Ditto.
7317 (aarch64_sm3tt2bq4si): Ditto.
7318 (aarch64_sm3partw1qv4si): Ditto.
7319 (aarch64_sm3partw2qv4si): Ditto.
7320 (aarch64_sm4eqv4si): Ditto.
7321 (aarch64_sm4ekeyqv4si): Ditto.
7322 * config/aarch64/aarch64-simd.md:
7323 (aarch64_sm3ss1qv4si): Ditto.
7324 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
7325 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
7326 (aarch64_sm4eqv4si): Ditto.
7327 (aarch64_sm4ekeyqv4si): Ditto.
7328 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
7329 (sm3part_op): Ditto.
7330 (CRYPTO_SM3TT): Ditto.
7331 (CRYPTO_SM3PART): Ditto.
7332 (UNSPEC_SM3SS1): New unspec.
7333 (UNSPEC_SM3TT1A): Ditto.
7334 (UNSPEC_SM3TT1B): Ditto.
7335 (UNSPEC_SM3TT2A): Ditto.
7336 (UNSPEC_SM3TT2B): Ditto.
7337 (UNSPEC_SM3PARTW1): Ditto.
7338 (UNSPEC_SM3PARTW2): Ditto.
7339 (UNSPEC_SM4E): Ditto.
7340 (UNSPEC_SM4EKEY): Ditto.
7341 * config/aarch64/constraints.md (Ui2): New constraint.
7342 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
7343 * config/arm/types.md (crypto_sm3): New type attribute.
7344 (crypto_sm4): Ditto.
7345 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
7346 (vsm3tt1aq_u32): Ditto.
7347 (vsm3tt1bq_u32): Ditto.
7348 (vsm3tt2aq_u32): Ditto.
7349 (vsm3tt2bq_u32): Ditto.
7350 (vsm3partw1q_u32): Ditto.
7351 (vsm3partw2q_u32): Ditto.
7352 (vsm4eq_u32): Ditto.
7353 (vsm4ekeyq_u32): Ditto.
7354 (doc/invoke.texi): Document new sm4 option.
7356 2018-01-10 Michael Collison <michael.collison@arm.com>
7358 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
7359 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
7360 (AARCH64_FL_FOR_ARCH8_4): New.
7361 (AARCH64_FL_V8_4): New flag.
7362 (doc/invoke.texi): Document new armv8.4-a option.
7364 2018-01-10 Michael Collison <michael.collison@arm.com>
7366 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7367 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
7368 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
7369 * config/aarch64/aarch64-option-extension.def: Add
7370 AARCH64_OPT_EXTENSION of 'sha2'.
7371 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
7372 (crypto): Disable sha2 and aes if crypto disabled.
7373 (crypto): Enable aes and sha2 if enabled.
7374 (simd): Disable sha2 and aes if simd disabled.
7375 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
7377 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
7378 (TARGET_SHA2): New feature flag for sha2.
7379 (TARGET_AES): New feature flag for aes.
7380 * config/aarch64/aarch64-simd.md:
7381 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
7382 conditional on TARGET_AES.
7383 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
7384 (aarch64_crypto_sha1hsi): Make pattern conditional
7386 (aarch64_crypto_sha1hv4si): Ditto.
7387 (aarch64_be_crypto_sha1hv4si): Ditto.
7388 (aarch64_crypto_sha1su1v4si): Ditto.
7389 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
7390 (aarch64_crypto_sha1su0v4si): Ditto.
7391 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
7392 (aarch64_crypto_sha256su0v4si): Ditto.
7393 (aarch64_crypto_sha256su1v4si): Ditto.
7394 (doc/invoke.texi): Document new aes and sha2 options.
7396 2018-01-10 Martin Sebor <msebor@redhat.com>
7398 PR tree-optimization/83781
7399 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
7402 2018-01-11 Martin Sebor <msebor@gmail.com>
7403 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7405 PR tree-optimization/83501
7406 PR tree-optimization/81703
7408 * tree-ssa-strlen.c (get_string_cst): Rename...
7409 (get_string_len): ...to this. Handle global constants.
7410 (handle_char_store): Adjust.
7412 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
7413 Jim Wilson <jimw@sifive.com>
7415 * config/riscv/riscv-protos.h (riscv_output_return): New.
7416 * config/riscv/riscv.c (struct machine_function): New naked_p field.
7417 (riscv_attribute_table, riscv_output_return),
7418 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
7419 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
7420 (riscv_compute_frame_info): Only compute frame->mask if not a naked
7422 (riscv_expand_prologue): Add early return for naked function.
7423 (riscv_expand_epilogue): Likewise.
7424 (riscv_function_ok_for_sibcall): Return false for naked function.
7425 (riscv_set_current_function): New.
7426 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7427 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
7428 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
7429 * doc/extend.texi (RISC-V Function Attributes): New.
7431 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7433 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
7434 check for 128-bit long double before checking TCmode.
7435 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
7436 128-bit long doubles before checking TFmode or TCmode.
7437 (FLOAT128_IBM_P): Likewise.
7439 2018-01-10 Martin Sebor <msebor@redhat.com>
7441 PR tree-optimization/83671
7442 * builtins.c (c_strlen): Unconditionally return zero for the empty
7444 Use -Warray-bounds for warnings.
7445 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
7446 for non-constant array indices with COMPONENT_REF, arrays of
7447 arrays, and pointers to arrays.
7448 (gimple_fold_builtin_strlen): Determine and set length range for
7449 non-constant character arrays.
7451 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7454 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7457 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7459 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7461 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7464 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7465 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7466 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7467 indexed_or_indirect_operand predicate.
7468 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7469 (*vsx_le_perm_load_v8hi): Likewise.
7470 (*vsx_le_perm_load_v16qi): Likewise.
7471 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7472 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7473 (*vsx_le_perm_store_v8hi): Likewise.
7474 (*vsx_le_perm_store_v16qi): Likewise.
7475 (eight unnamed splitters): Likewise.
7477 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7479 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7480 * config/rs6000/emmintrin.h: Likewise.
7481 * config/rs6000/mmintrin.h: Likewise.
7482 * config/rs6000/xmmintrin.h: Likewise.
7484 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7487 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7489 * tree.c (tree_nop_conversion): Return true for location wrapper
7491 (maybe_wrap_with_location): New function.
7492 (selftest::check_strip_nops): New function.
7493 (selftest::test_location_wrappers): New function.
7494 (selftest::tree_c_tests): Call it.
7495 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7496 (maybe_wrap_with_location): New decl.
7497 (EXPR_LOCATION_WRAPPER_P): New macro.
7498 (location_wrapper_p): New inline function.
7499 (tree_strip_any_location_wrapper): New inline function.
7501 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7504 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7505 stack_realign_offset for the largest alignment of stack slot
7507 (ix86_find_max_used_stack_alignment): New function.
7508 (ix86_finalize_stack_frame_flags): Use it. Set
7509 max_used_stack_alignment if we don't realign stack.
7510 * config/i386/i386.h (machine_function): Add
7511 max_used_stack_alignment.
7513 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
7515 * config/arm/arm.opt (-mbranch-cost): New option.
7516 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
7519 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
7522 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
7523 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
7525 2018-01-10 Richard Biener <rguenther@suse.de>
7528 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
7529 early out so it also covers the case where we have a non-NULL
7532 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7534 PR tree-optimization/83753
7535 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
7536 for non-strided grouped accesses if the number of elements is 1.
7538 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7541 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
7542 * i386.h (TARGET_USE_GATHER): Define.
7543 * x86-tune.def (X86_TUNE_USE_GATHER): New.
7545 2018-01-10 Martin Liska <mliska@suse.cz>
7548 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
7549 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
7551 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
7552 CLEANUP_NO_PARTITIONING is not set.
7554 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
7556 * doc/rtl.texi: Remove documentation of (const ...) wrappers
7557 for vectors, as a partial revert of r254296.
7558 * rtl.h (const_vec_p): Delete.
7559 (const_vec_duplicate_p): Don't test for vector CONSTs.
7560 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
7561 * expmed.c (make_tree): Likewise.
7564 * common.md (E, F): Use CONSTANT_P instead of checking for
7566 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
7567 checking for CONST_VECTOR.
7569 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7572 * predict.c (force_edge_cold): Handle in more sane way edges
7575 2018-01-09 Carl Love <cel@us.ibm.com>
7577 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
7579 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
7580 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
7581 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
7582 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
7583 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
7584 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
7585 * config/rs6000/rs6000-protos.h: Add extern defition for
7586 rs6000_generate_float2_double_code.
7587 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
7589 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
7590 (float2_v2df): Add define_expand.
7592 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
7595 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
7596 op_mode in the force_to_mode call.
7598 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7600 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
7601 instead of checking each element individually.
7602 (aarch64_evpc_uzp): Likewise.
7603 (aarch64_evpc_zip): Likewise.
7604 (aarch64_evpc_ext): Likewise.
7605 (aarch64_evpc_rev): Likewise.
7606 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
7607 instead of checking each element individually. Return true without
7609 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
7610 whether all selected elements come from the same input, instead of
7611 checking each element individually. Remove calls to gen_rtx_REG,
7612 start_sequence and end_sequence and instead assert that no rtl is
7615 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7617 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
7618 order of HIGH and CONST checks.
7620 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
7622 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
7623 if the destination isn't an SSA_NAME.
7625 2018-01-09 Richard Biener <rguenther@suse.de>
7627 PR tree-optimization/83668
7628 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
7630 (canonicalize_loop_form): ... here, renamed from ...
7631 (canonicalize_loop_closed_ssa_form): ... this and amended to
7632 swap successor edges for loop exit blocks to make us use
7633 the RPO order we need for initial schedule generation.
7635 2018-01-09 Joseph Myers <joseph@codesourcery.com>
7637 PR tree-optimization/64811
7638 * match.pd: When optimizing comparisons with Inf, avoid
7639 introducing or losing exceptions from comparisons with NaN.
7641 2018-01-09 Martin Liska <mliska@suse.cz>
7644 * asan.c (shadow_mem_size): Add gcc_assert.
7646 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
7648 Don't save registers in main().
7651 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
7652 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
7653 * config/avr/avr.c (avr_set_current_function): Don't error if
7654 naked, OS_task or OS_main are specified at the same time.
7655 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
7657 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
7659 * common/config/avr/avr-common.c (avr_option_optimization_table):
7660 Switch on -mmain-is-OS_task for optimizing compilations.
7662 2018-01-09 Richard Biener <rguenther@suse.de>
7664 PR tree-optimization/83572
7665 * graphite.c: Include cfganal.h.
7666 (graphite_transform_loops): Connect infinite loops to exit
7667 and remove fake edges at the end.
7669 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7671 * ipa-inline.c (edge_badness): Revert accidental checkin.
7673 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
7676 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
7677 symbols; not inline clones.
7679 2018-01-09 Jakub Jelinek <jakub@redhat.com>
7682 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
7683 hard registers. Formatting fixes.
7685 PR preprocessor/83722
7686 * gcc.c (try_generate_repro): Pass
7687 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
7688 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
7691 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
7692 Kito Cheng <kito.cheng@gmail.com>
7694 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
7695 (riscv_leaf_function_p): Delete.
7696 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
7698 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7700 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
7702 (do_ifelse): New function.
7703 (do_isel): New function.
7704 (do_sub3): New function.
7705 (do_add3): New function.
7706 (do_load_mask_compare): New function.
7707 (do_overlap_load_compare): New function.
7708 (expand_compare_loop): New function.
7709 (expand_block_compare): Call expand_compare_loop() when appropriate.
7710 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
7712 (-mblock-compare-inline-loop-limit): New option.
7714 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7717 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
7718 Reverse order of second and third operands in first alternative.
7719 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
7720 of first and second elements in UNSPEC_VPERMR vector.
7721 (altivec_expand_vec_perm_le): Likewise.
7723 2018-01-08 Jeff Law <law@redhat.com>
7725 PR rtl-optimizatin/81308
7726 * tree-switch-conversion.c (cfg_altered): New file scoped static.
7727 (process_switch): If group_case_labels makes a change, then set
7729 (pass_convert_switch::execute): If a switch is converted, then
7730 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
7732 PR rtl-optimization/81308
7733 * recog.c (split_all_insns): Conditionally cleanup the CFG after
7736 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
7738 PR target/83663 - Revert r255946
7739 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
7740 generation for cases where splatting a value is not useful.
7741 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
7742 across a vec_duplicate and a paradoxical subreg forming a vector
7743 mode to a vec_concat.
7745 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7747 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
7748 -march=armv8.3-a variants.
7749 * config/arm/t-multilib: Likewise.
7750 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
7752 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7754 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
7756 (cceq_ior_compare_complement): Give it a name so I can use it, and
7757 change boolean_or_operator predicate to boolean_operator so it can
7758 be used to generate a crand.
7759 (eqne): New code iterator.
7760 (bd/bd_neg): New code_attrs.
7761 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
7762 a single define_insn.
7763 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
7764 decrement (bdnzt/bdnzf/bdzt/bdzf).
7765 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
7766 with the new names of the branch decrement patterns, and added the
7767 names of the branch decrement conditional patterns.
7769 2018-01-08 Richard Biener <rguenther@suse.de>
7771 PR tree-optimization/83563
7772 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
7775 2018-01-08 Richard Biener <rguenther@suse.de>
7778 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
7780 2018-01-08 Richard Biener <rguenther@suse.de>
7782 PR tree-optimization/83685
7783 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
7784 references to abnormals.
7786 2018-01-08 Richard Biener <rguenther@suse.de>
7789 * dwarf2out.c (output_indirect_strings): Handle empty
7790 skeleton_debug_str_hash.
7791 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
7793 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7795 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
7796 (emit_store_direct): Likewise.
7797 (arc_trampoline_adjust_address): Likewise.
7798 (arc_asm_trampoline_template): New function.
7799 (arc_initialize_trampoline): Use asm_trampoline_template.
7800 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
7801 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
7802 * config/arc/arc.md (flush_icache): Delete pattern.
7804 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
7806 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
7807 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
7810 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7813 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
7814 by not USED_FOR_TARGET.
7815 (make_pass_resolve_sw_modes): Likewise.
7817 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
7819 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
7822 2018-01-08 Richard Biener <rguenther@suse.de>
7825 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
7827 2018-01-08 Richard Biener <rguenther@suse.de>
7830 * match.pd ((t * 2) / 2) -> t): Add missing :c.
7832 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
7835 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
7836 basic blocks with a small number of successors.
7837 (convert_control_dep_chain_into_preds): Improve handling of
7839 (dump_predicates): Split apart into...
7840 (dump_pred_chain): ...here...
7841 (dump_pred_info): ...and here.
7842 (can_one_predicate_be_invalidated_p): Add debugging printfs.
7843 (can_chain_union_be_invalidated_p): Improve check for invalidation
7845 (uninit_uses_cannot_happen): Avoid unnecessary if
7846 convert_control_dep_chain_into_preds yielded nothing.
7848 2018-01-06 Martin Sebor <msebor@redhat.com>
7850 PR tree-optimization/83640
7851 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
7852 subtracting negative offset from size.
7853 (builtin_access::overlap): Adjust offset bounds of the access to fall
7854 within the size of the object if possible.
7856 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
7858 PR rtl-optimization/83699
7859 * expmed.c (extract_bit_field_1): Restrict the vector usage of
7860 extract_bit_field_as_subreg to cases in which the extracted
7861 value is also a vector.
7863 * lra-constraints.c (process_alt_operands): Test for the equivalence
7864 substitutions when detecting a possible reload cycle.
7866 2018-01-06 Jakub Jelinek <jakub@redhat.com>
7869 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
7870 by default if flag_selective_schedling{,2}. Formatting fixes.
7872 PR rtl-optimization/83682
7873 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
7874 if it has non-VECTOR_MODE element mode.
7875 (vec_duplicate_p): Likewise.
7878 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
7879 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
7881 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7884 * config/i386/i386-builtin.def
7885 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
7886 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
7887 Require also OPTION_MASK_ISA_AVX512F in addition to
7888 OPTION_MASK_ISA_GFNI.
7889 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
7890 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
7891 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
7892 to OPTION_MASK_ISA_GFNI.
7893 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
7894 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
7895 OPTION_MASK_ISA_AVX512BW.
7896 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
7897 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
7898 addition to OPTION_MASK_ISA_GFNI.
7899 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
7900 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
7901 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
7902 to OPTION_MASK_ISA_GFNI.
7903 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
7904 a requirement for all ISAs rather than any of them with a few
7906 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
7908 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
7909 bitmasks to be enabled with 3 exceptions, instead of requiring any
7910 enabled ISA with lots of exceptions.
7911 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
7912 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
7913 Change avx512bw in isa attribute to avx512f.
7914 * config/i386/sgxintrin.h: Add license boilerplate.
7915 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
7916 to __AVX512F__ and __AVX512VL to __AVX512VL__.
7917 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
7918 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
7920 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
7921 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
7922 temporarily sse2 rather than sse if not enabled already.
7925 * config/i386/sse.md (VI248_VLBW): Rename to ...
7926 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
7927 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
7928 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
7929 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
7930 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
7931 mode iterator instead of VI248_VLBW.
7933 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
7935 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
7936 (record_modified): Skip clobbers; add debug output.
7937 (param_change_prob): Use sreal frequencies.
7939 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7941 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
7942 punt for user-aligned variables.
7944 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
7946 * tree-chrec.c (chrec_contains_symbols): Return true for
7949 2018-01-05 Sudakshina Das <sudi.das@arm.com>
7952 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
7953 of (x|y) == x for BICS pattern.
7955 2018-01-05 Jakub Jelinek <jakub@redhat.com>
7957 PR tree-optimization/83605
7958 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
7959 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
7962 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
7964 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
7965 * config/epiphany/rtems.h: New file.
7967 2018-01-04 Jakub Jelinek <jakub@redhat.com>
7968 Uros Bizjak <ubizjak@gmail.com>
7971 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
7972 QIreg_operand instead of register_operand predicate.
7973 * config/i386/i386.c (ix86_rop_should_change_byte_p,
7974 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
7975 comments instead of -fmitigate[-_]rop.
7977 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7980 * cgraphunit.c (symbol_table::compile): Switch to text_section
7981 before calling assembly_start debug hook.
7982 * run-rtl-passes.c (run_rtl_passes): Likewise.
7985 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7987 * tree-vrp.c (extract_range_from_binary_expr_1): Check
7988 range_int_cst_p rather than !symbolic_range_p before calling
7989 extract_range_from_multiplicative_op_1.
7991 2018-01-04 Jeff Law <law@redhat.com>
7993 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
7994 redundant test in assertion.
7996 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
7998 * doc/rtl.texi: Document machine_mode wrapper classes.
8000 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8002 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
8005 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8007 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
8008 the VEC_PERM_EXPR fold to fail.
8010 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8013 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
8014 to switched_sections.
8016 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8019 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
8022 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
8025 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
8026 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
8028 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8031 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
8032 is BLKmode and bitpos not zero or mode change is needed.
8034 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8037 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
8040 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
8043 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
8044 instead of MULT rtx. Update all corresponding splitters.
8046 (*ssub<modesuffix>): Ditto.
8048 (*cmp_sadd_di): Update split patterns.
8049 (*cmp_sadd_si): Ditto.
8050 (*cmp_sadd_sidi): Ditto.
8051 (*cmp_ssub_di): Ditto.
8052 (*cmp_ssub_si): Ditto.
8053 (*cmp_ssub_sidi): Ditto.
8054 * config/alpha/predicates.md (const23_operand): New predicate.
8055 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
8056 Look for ASHIFT, not MULT inner operand.
8057 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
8059 2018-01-04 Martin Liska <mliska@suse.cz>
8061 PR gcov-profile/83669
8062 * gcov.c (output_intermediate_file): Add version to intermediate
8064 * doc/gcov.texi: Document new field 'version' in intermediate
8065 file format. Fix location of '-k' option of gcov command.
8067 2018-01-04 Martin Liska <mliska@suse.cz>
8070 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
8072 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8074 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
8076 2018-01-03 Martin Sebor <msebor@redhat.com>
8078 PR tree-optimization/83655
8079 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
8080 checking calls with invalid arguments.
8082 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8084 * tree-vect-stmts.c (vect_get_store_rhs): New function.
8085 (vectorizable_mask_load_store): Delete.
8086 (vectorizable_call): Return false for masked loads and stores.
8087 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
8088 instead of gimple_assign_rhs1.
8089 (vectorizable_load): Handle IFN_MASK_LOAD.
8090 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
8092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8094 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
8096 (vectorizable_mask_load_store): ...here.
8097 (vectorizable_load): ...and here.
8099 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8101 * tree-vect-stmts.c (vect_build_all_ones_mask)
8102 (vect_build_zero_merge_argument): New functions, split out from...
8103 (vectorizable_load): ...here.
8105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8107 * tree-vect-stmts.c (vect_check_store_rhs): New function,
8109 (vectorizable_mask_load_store): ...here.
8110 (vectorizable_store): ...and here.
8112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8114 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
8116 (vectorizable_mask_load_store): ...here.
8118 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8120 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
8121 (vect_model_store_cost): Take a vec_load_store_type instead of a
8123 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
8124 (vect_model_store_cost): Take a vec_load_store_type instead of a
8126 (vectorizable_mask_load_store): Update accordingly.
8127 (vectorizable_store): Likewise.
8128 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
8130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8132 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
8133 IFN_MASK_LOAD calls here rather than...
8134 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
8136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8137 Alan Hayward <alan.hayward@arm.com>
8138 David Sherwood <david.sherwood@arm.com>
8140 * expmed.c (extract_bit_field_1): For vector extracts,
8141 fall back to extract_bit_field_as_subreg if vec_extract
8144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8145 Alan Hayward <alan.hayward@arm.com>
8146 David Sherwood <david.sherwood@arm.com>
8148 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
8149 they are variable or constant sized.
8150 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
8151 slots for constant-sized data.
8153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8154 Alan Hayward <alan.hayward@arm.com>
8155 David Sherwood <david.sherwood@arm.com>
8157 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
8158 handling COND_EXPRs with boolean comparisons, try to find a better
8159 basis for the mask type than the boolean itself.
8161 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8163 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
8164 is calculated and how it can be overridden.
8165 * genmodes.c (max_bitsize_mode_any_mode): New variable.
8166 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
8168 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
8171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8172 Alan Hayward <alan.hayward@arm.com>
8173 David Sherwood <david.sherwood@arm.com>
8175 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
8176 Remove the mode argument.
8177 (aarch64_simd_valid_immediate): Remove the mode and inverse
8179 * config/aarch64/iterators.md (bitsize): New iterator.
8180 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
8181 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
8182 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
8183 aarch64_simd_valid_immediate.
8184 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
8185 (aarch64_reg_or_bic_imm): Likewise.
8186 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
8187 with an insn_type enum and msl with a modifier_type enum.
8188 Replace element_width with a scalar_mode. Change the shift
8189 to unsigned int. Add constructors for scalar_float_mode and
8190 scalar_int_mode elements.
8191 (aarch64_vect_float_const_representable_p): Delete.
8192 (aarch64_can_const_movi_rtx_p)
8193 (aarch64_simd_scalar_immediate_valid_for_move)
8194 (aarch64_simd_make_constant): Update call to
8195 aarch64_simd_valid_immediate.
8196 (aarch64_advsimd_valid_immediate_hs): New function.
8197 (aarch64_advsimd_valid_immediate): Likewise.
8198 (aarch64_simd_valid_immediate): Remove mode and inverse
8199 arguments. Rewrite to use the above. Use const_vec_duplicate_p
8200 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
8201 and aarch64_float_const_representable_p on the result.
8202 (aarch64_output_simd_mov_immediate): Remove mode argument.
8203 Update call to aarch64_simd_valid_immediate and use of
8204 simd_immediate_info.
8205 (aarch64_output_scalar_simd_mov_immediate): Update call
8208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8209 Alan Hayward <alan.hayward@arm.com>
8210 David Sherwood <david.sherwood@arm.com>
8212 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
8213 (mode_nunits): Likewise CONST_MODE_NUNITS.
8214 * machmode.def (ADJUST_NUNITS): Document.
8215 * genmodes.c (mode_data::need_nunits_adj): New field.
8216 (blank_mode): Update accordingly.
8217 (adj_nunits): New variable.
8218 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
8220 (emit_mode_size_inline): Set need_bytesize_adj for all modes
8221 listed in adj_nunits.
8222 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
8223 listed in adj_nunits. Don't emit case statements for such modes.
8224 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
8225 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
8226 nothing if adj_nunits is nonnull.
8227 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
8228 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
8229 (emit_mode_fbit): Update use of print_maybe_const_decl.
8230 (emit_move_size): Likewise. Treat the array as non-const
8232 (emit_mode_adjustments): Handle adj_nunits.
8234 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8236 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
8237 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
8238 (VECTOR_MODES): Use it.
8239 (make_vector_modes): Take the prefix as an argument.
8241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8242 Alan Hayward <alan.hayward@arm.com>
8243 David Sherwood <david.sherwood@arm.com>
8245 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
8246 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
8247 for MODE_VECTOR_BOOL.
8248 * machmode.def (VECTOR_BOOL_MODE): Document.
8249 * genmodes.c (VECTOR_BOOL_MODE): New macro.
8250 (make_vector_bool_mode): New function.
8251 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
8253 * lto-streamer-in.c (lto_input_mode_table): Likewise.
8254 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
8256 * stor-layout.c (int_mode_for_mode): Likewise.
8257 * tree.c (build_vector_type_for_mode): Likewise.
8258 * varasm.c (output_constant_pool_2): Likewise.
8259 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
8260 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
8261 for MODE_VECTOR_BOOL.
8262 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
8263 of mode class checks.
8264 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
8265 instead of a list of mode class checks.
8266 (expand_vector_scalar_condition): Likewise.
8267 (type_for_widest_vector_mode): Handle BImode as an inner mode.
8269 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8270 Alan Hayward <alan.hayward@arm.com>
8271 David Sherwood <david.sherwood@arm.com>
8273 * machmode.h (mode_size): Change from unsigned short to
8275 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
8276 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8277 or if measurement_type is not polynomial.
8278 (fixed_size_mode::includes_p): Check for constant-sized modes.
8279 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
8280 return a poly_uint16 rather than an unsigned short.
8281 (emit_mode_size): Change the type of mode_size from unsigned short
8282 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
8283 (emit_mode_adjustments): Cope with polynomial vector sizes.
8284 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8286 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8288 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
8289 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
8290 * caller-save.c (setup_save_areas): Likewise.
8291 (replace_reg_with_saved_mem): Likewise.
8292 * calls.c (emit_library_call_value_1): Likewise.
8293 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
8294 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
8295 (gen_lowpart_for_combine): Likewise.
8296 * convert.c (convert_to_integer_1): Likewise.
8297 * cse.c (equiv_constant, cse_insn): Likewise.
8298 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
8299 (cselib_subst_to_values): Likewise.
8300 * dce.c (word_dce_process_block): Likewise.
8301 * df-problems.c (df_word_lr_mark_ref): Likewise.
8302 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
8303 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
8304 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
8305 (rtl_for_decl_location): Likewise.
8306 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
8307 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
8308 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
8309 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
8310 (expand_expr_real_1): Likewise.
8311 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
8312 (pad_below): Likewise.
8313 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8314 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
8315 * ira.c (get_subreg_tracking_sizes): Likewise.
8316 * ira-build.c (ira_create_allocno_objects): Likewise.
8317 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
8318 (ira_sort_regnos_for_alter_reg): Likewise.
8319 * ira-costs.c (record_operand_costs): Likewise.
8320 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
8321 (resolve_simple_move): Likewise.
8322 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
8323 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
8324 (lra_constraints): Likewise.
8325 (CONST_POOL_OK_P): Reject variable-sized modes.
8326 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
8327 (add_pseudo_to_slot, lra_spill): Likewise.
8328 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8329 * optabs-query.c (get_best_extraction_insn): Likewise.
8330 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8331 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
8332 (expand_mult_highpart, valid_multiword_target_p): Likewise.
8333 * recog.c (offsettable_address_addr_space_p): Likewise.
8334 * regcprop.c (maybe_mode_change): Likewise.
8335 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
8336 * regrename.c (build_def_use): Likewise.
8337 * regstat.c (dump_reg_info): Likewise.
8338 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
8339 (find_reloads, find_reloads_subreg_address): Likewise.
8340 * reload1.c (eliminate_regs_1): Likewise.
8341 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
8342 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
8343 (simplify_binary_operation_1, simplify_subreg): Likewise.
8344 * targhooks.c (default_function_arg_padding): Likewise.
8345 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
8346 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
8347 (verify_gimple_assign_ternary): Likewise.
8348 * tree-inline.c (estimate_move_cost): Likewise.
8349 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8350 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
8351 (get_address_cost_ainc): Likewise.
8352 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
8353 (vect_supportable_dr_alignment): Likewise.
8354 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8355 (vectorizable_reduction): Likewise.
8356 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
8357 (vectorizable_operation, vectorizable_load): Likewise.
8358 * tree.c (build_same_sized_truth_vector_type): Likewise.
8359 * valtrack.c (cleanup_auto_inc_dec): Likewise.
8360 * var-tracking.c (emit_note_insn_var_location): Likewise.
8361 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
8362 (ADDR_VEC_ALIGN): Likewise.
8364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8365 Alan Hayward <alan.hayward@arm.com>
8366 David Sherwood <david.sherwood@arm.com>
8368 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
8370 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8371 or if measurement_type is polynomial.
8372 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
8373 * combine.c (make_extraction): Likewise.
8374 * dse.c (find_shift_sequence): Likewise.
8375 * dwarf2out.c (mem_loc_descriptor): Likewise.
8376 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
8377 (extract_bit_field, extract_low_bits): Likewise.
8378 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
8379 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
8380 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
8381 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
8382 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8383 * reload.c (find_reloads): Likewise.
8384 * reload1.c (alter_reg): Likewise.
8385 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
8386 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
8387 * tree-if-conv.c (predicate_mem_writes): Likewise.
8388 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
8389 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
8390 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
8391 * valtrack.c (dead_debug_insert_temp): Likewise.
8392 * varasm.c (mergeable_constant_section): Likewise.
8393 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
8395 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8396 Alan Hayward <alan.hayward@arm.com>
8397 David Sherwood <david.sherwood@arm.com>
8399 * expr.c (expand_assignment): Cope with polynomial mode sizes
8400 when assigning to a CONCAT.
8402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8403 Alan Hayward <alan.hayward@arm.com>
8404 David Sherwood <david.sherwood@arm.com>
8406 * machmode.h (mode_precision): Change from unsigned short to
8408 (mode_to_precision): Return a poly_uint16 rather than an unsigned
8410 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
8411 or if measurement_type is not polynomial.
8412 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
8413 in which the mode is already known to be a scalar_int_mode.
8414 * genmodes.c (emit_mode_precision): Change the type of mode_precision
8415 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
8417 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8418 for GET_MODE_PRECISION.
8419 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8420 for GET_MODE_PRECISION.
8421 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8423 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8424 (expand_field_assignment, make_extraction): Likewise.
8425 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8426 (get_last_value): Likewise.
8427 * convert.c (convert_to_integer_1): Likewise.
8428 * cse.c (cse_insn): Likewise.
8429 * expr.c (expand_expr_real_1): Likewise.
8430 * lra-constraints.c (simplify_operand_subreg): Likewise.
8431 * optabs-query.c (can_atomic_load_p): Likewise.
8432 * optabs.c (expand_atomic_load): Likewise.
8433 (expand_atomic_store): Likewise.
8434 * ree.c (combine_reaching_defs): Likewise.
8435 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
8436 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
8437 * tree.h (type_has_mode_precision_p): Likewise.
8438 * ubsan.c (instrument_si_overflow): Likewise.
8440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8441 Alan Hayward <alan.hayward@arm.com>
8442 David Sherwood <david.sherwood@arm.com>
8444 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
8445 polynomial numbers of units.
8446 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
8447 (valid_vector_subparts_p): New function.
8448 (build_vector_type): Remove temporary shim and take the number
8449 of units as a poly_uint64 rather than an int.
8450 (build_opaque_vector_type): Take the number of units as a
8451 poly_uint64 rather than an int.
8452 * tree.c (build_vector_from_ctor): Handle polynomial
8453 TYPE_VECTOR_SUBPARTS.
8454 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8455 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8456 (build_vector_from_val): If the number of units is variable,
8457 use build_vec_duplicate_cst for constant operands and
8458 VEC_DUPLICATE_EXPR otherwise.
8459 (make_vector_type): Remove temporary is_constant ().
8460 (build_vector_type, build_opaque_vector_type): Take the number of
8461 units as a poly_uint64 rather than an int.
8462 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8464 * cfgexpand.c (expand_debug_expr): Likewise.
8465 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8466 (store_constructor, expand_expr_real_1): Likewise.
8467 (const_scalar_mask_from_tree): Likewise.
8468 * fold-const-call.c (fold_const_reduction): Likewise.
8469 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8470 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8471 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8472 (fold_relational_const): Likewise.
8473 (native_interpret_vector): Likewise. Change the size from an
8474 int to an unsigned int.
8475 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8476 TYPE_VECTOR_SUBPARTS.
8477 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8478 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8479 duplicating a non-constant operand into a variable-length vector.
8480 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8481 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8482 * ipa-icf.c (sem_variable::equals): Likewise.
8483 * match.pd: Likewise.
8484 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8485 * print-tree.c (print_node): Likewise.
8486 * stor-layout.c (layout_type): Likewise.
8487 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8488 * tree-cfg.c (verify_gimple_comparison): Likewise.
8489 (verify_gimple_assign_binary): Likewise.
8490 (verify_gimple_assign_ternary): Likewise.
8491 (verify_gimple_assign_single): Likewise.
8492 * tree-pretty-print.c (dump_generic_node): Likewise.
8493 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8494 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8495 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8496 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8497 (vect_shift_permute_load_chain): Likewise.
8498 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8499 (expand_vector_condition, optimize_vector_constructor): Likewise.
8500 (lower_vec_perm, get_compute_type): Likewise.
8501 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8502 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8503 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8504 (vect_recog_mask_conversion_pattern): Likewise.
8505 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8506 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8507 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8508 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8509 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
8510 (vectorizable_shift, vectorizable_operation, vectorizable_store)
8511 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
8512 (supportable_widening_operation): Likewise.
8513 (supportable_narrowing_operation): Likewise.
8514 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
8516 * varasm.c (output_constant): Likewise.
8518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8519 Alan Hayward <alan.hayward@arm.com>
8520 David Sherwood <david.sherwood@arm.com>
8522 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
8523 so that both the length == 3 and length != 3 cases set up their
8524 own permute vectors. Add comments explaining why we know the
8525 number of elements is constant.
8526 (vect_permute_load_chain): Likewise.
8528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8529 Alan Hayward <alan.hayward@arm.com>
8530 David Sherwood <david.sherwood@arm.com>
8532 * machmode.h (mode_nunits): Change from unsigned char to
8534 (ONLY_FIXED_SIZE_MODES): New macro.
8535 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
8536 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
8537 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
8539 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
8540 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
8541 or if measurement_type is not polynomial.
8542 * genmodes.c (ZERO_COEFFS): New macro.
8543 (emit_mode_nunits_inline): Make mode_nunits_inline return a
8545 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
8546 Use ZERO_COEFFS when emitting initializers.
8547 * data-streamer.h (bp_pack_poly_value): New function.
8548 (bp_unpack_poly_value): Likewise.
8549 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8550 for GET_MODE_NUNITS.
8551 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8552 for GET_MODE_NUNITS.
8553 * tree.c (make_vector_type): Remove temporary shim and make
8554 the real function take the number of units as a poly_uint64
8556 (build_vector_type_for_mode): Handle polynomial nunits.
8557 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
8558 * emit-rtl.c (const_vec_series_p_1): Likewise.
8559 (gen_rtx_CONST_VECTOR): Likewise.
8560 * fold-const.c (test_vec_duplicate_folding): Likewise.
8561 * genrecog.c (validate_pattern): Likewise.
8562 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
8563 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8564 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
8565 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
8566 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
8567 * rtlanal.c (subreg_get_info): Likewise.
8568 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8569 (vect_grouped_load_supported): Likewise.
8570 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
8571 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
8572 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8573 (simplify_const_unary_operation, simplify_binary_operation_1)
8574 (simplify_const_binary_operation, simplify_ternary_operation)
8575 (test_vector_ops_duplicate, test_vector_ops): Likewise.
8576 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
8577 instead of CONST_VECTOR_NUNITS.
8578 * varasm.c (output_constant_pool_2): Likewise.
8579 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
8580 explicit-encoded elements in the XVEC for variable-length vectors.
8582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8584 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
8586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8587 Alan Hayward <alan.hayward@arm.com>
8588 David Sherwood <david.sherwood@arm.com>
8590 * coretypes.h (fixed_size_mode): Declare.
8591 (fixed_size_mode_pod): New typedef.
8592 * builtins.h (target_builtins::x_apply_args_mode)
8593 (target_builtins::x_apply_result_mode): Change type to
8594 fixed_size_mode_pod.
8595 * builtins.c (apply_args_size, apply_result_size, result_vector)
8596 (expand_builtin_apply_args_1, expand_builtin_apply)
8597 (expand_builtin_return): Update accordingly.
8599 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8601 * cse.c (hash_rtx_cb): Hash only the encoded elements.
8602 * cselib.c (cselib_hash_rtx): Likewise.
8603 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
8604 CONST_VECTOR encoding.
8606 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8607 Jeff Law <law@redhat.com>
8610 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
8611 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
8612 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
8613 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
8616 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
8617 explicitly probe *sp in a noreturn function if there were any callee
8618 register saves or frame pointer is needed.
8620 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8623 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
8624 BLKmode for ternary, binary or unary expressions.
8627 * var-tracking.c (delete_vta_debug_insn): New inline function.
8628 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
8629 insns from get_insns () to NULL instead of each bb separately.
8630 Use delete_vta_debug_insn. No longer static.
8631 (vt_debug_insns_local, variable_tracking_main_1): Adjust
8632 delete_vta_debug_insns callers.
8633 * rtl.h (delete_vta_debug_insns): Declare.
8634 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
8635 instead of variable_tracking_main.
8637 2018-01-03 Martin Sebor <msebor@redhat.com>
8639 PR tree-optimization/83603
8640 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
8641 arguments past the endof the argument list in functions declared
8642 without a prototype.
8643 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8644 Avoid checking when arguments are null.
8646 2018-01-03 Martin Sebor <msebor@redhat.com>
8649 * doc/extend.texi (attribute const): Fix a typo.
8650 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
8651 issuing -Wsuggest-attribute for void functions.
8653 2018-01-03 Martin Sebor <msebor@redhat.com>
8655 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
8656 offset_int::from instead of wide_int::to_shwi.
8657 (maybe_diag_overlap): Remove assertion.
8658 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
8659 * gimple-ssa-sprintf.c (format_directive): Same.
8660 (parse_directive): Same.
8661 (sprintf_dom_walker::compute_format_length): Same.
8662 (try_substitute_return_value): Same.
8664 2018-01-03 Jeff Law <law@redhat.com>
8667 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
8668 non-constant residual for zero at runtime and avoid probing in
8669 that case. Reorganize code for trailing problem to mirror handling
8672 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8674 PR tree-optimization/83501
8675 * tree-ssa-strlen.c (get_string_cst): New.
8676 (handle_char_store): Call get_string_cst.
8678 2018-01-03 Martin Liska <mliska@suse.cz>
8680 PR tree-optimization/83593
8681 * tree-ssa-strlen.c: Include tree-cfg.h.
8682 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
8683 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
8684 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
8686 (strlen_dom_walker::before_dom_children): Call
8687 gimple_purge_dead_eh_edges. Dump tranformation with details
8689 (strlen_dom_walker::before_dom_children): Update call by adding
8690 new argument cleanup_eh.
8691 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
8693 2018-01-03 Martin Liska <mliska@suse.cz>
8696 * cif-code.def (VARIADIC_THUNK): New enum value.
8697 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
8700 2018-01-03 Jan Beulich <jbeulich@suse.com>
8702 * sse.md (mov<mode>_internal): Tighten condition for when to use
8703 vmovdqu<ssescalarsize> for TI and OI modes.
8705 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8707 Update copyright years.
8709 2018-01-03 Martin Liska <mliska@suse.cz>
8712 * ipa-visibility.c (function_and_variable_visibility): Skip
8713 functions with noipa attribure.
8715 2018-01-03 Jakub Jelinek <jakub@redhat.com>
8717 * gcc.c (process_command): Update copyright notice dates.
8718 * gcov-dump.c (print_version): Ditto.
8719 * gcov.c (print_version): Ditto.
8720 * gcov-tool.c (print_version): Ditto.
8721 * gengtype.c (create_file): Ditto.
8722 * doc/cpp.texi: Bump @copying's copyright year.
8723 * doc/cppinternals.texi: Ditto.
8724 * doc/gcc.texi: Ditto.
8725 * doc/gccint.texi: Ditto.
8726 * doc/gcov.texi: Ditto.
8727 * doc/install.texi: Ditto.
8728 * doc/invoke.texi: Ditto.
8730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8732 * vector-builder.h (vector_builder::m_full_nelts): Change from
8733 unsigned int to poly_uint64.
8734 (vector_builder::full_nelts): Update prototype accordingly.
8735 (vector_builder::new_vector): Likewise.
8736 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
8737 (vector_builder::operator ==): Likewise.
8738 (vector_builder::finalize): Likewise.
8739 * int-vector-builder.h (int_vector_builder::int_vector_builder):
8740 Take the number of elements as a poly_uint64 rather than an
8742 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
8743 from unsigned int to poly_uint64.
8744 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
8745 (vec_perm_indices::new_vector): Likewise.
8746 (vec_perm_indices::length): Likewise.
8747 (vec_perm_indices::nelts_per_input): Likewise.
8748 (vec_perm_indices::input_nelts): Likewise.
8749 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
8750 number of elements per input as a poly_uint64 rather than an
8751 unsigned int. Use the original encoding for variable-length
8752 vectors, rather than clamping each individual element.
8753 For the second and subsequent elements in each pattern,
8754 clamp the step and base before clamping their sum.
8755 (vec_perm_indices::series_p): Handle polynomial element counts.
8756 (vec_perm_indices::all_in_range_p): Likewise.
8757 (vec_perm_indices_to_tree): Likewise.
8758 (vec_perm_indices_to_rtx): Likewise.
8759 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
8760 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
8761 (tree_vector_builder::new_binary_operation): Handle polynomial
8762 element counts. Return false if we need to know the number
8763 of elements at compile time.
8764 * fold-const.c (fold_vec_perm): Punt if the number of elements
8765 isn't known at compile time.
8767 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8769 * vec-perm-indices.h (vec_perm_builder): Change element type
8770 from HOST_WIDE_INT to poly_int64.
8771 (vec_perm_indices::element_type): Update accordingly.
8772 (vec_perm_indices::clamp): Handle polynomial element_types.
8773 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8774 (vec_perm_indices::all_in_range_p): Likewise.
8775 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
8777 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
8778 polynomial vec_perm_indices element types.
8779 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
8780 * fold-const.c (fold_vec_perm): Likewise.
8781 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
8782 * tree-vect-generic.c (lower_vec_perm): Likewise.
8783 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8784 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
8785 element type to HOST_WIDE_INT.
8787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8788 Alan Hayward <alan.hayward@arm.com>
8789 David Sherwood <david.sherwood@arm.com>
8791 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
8792 rather than an int. Use plus_constant.
8793 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
8794 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
8796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8797 Alan Hayward <alan.hayward@arm.com>
8798 David Sherwood <david.sherwood@arm.com>
8800 * calls.c (emit_call_1, expand_call): Change struct_value_size from
8801 a HOST_WIDE_INT to a poly_int64.
8803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8804 Alan Hayward <alan.hayward@arm.com>
8805 David Sherwood <david.sherwood@arm.com>
8807 * calls.c (load_register_parameters): Cope with polynomial
8808 mode sizes. Require a constant size for BLKmode parameters
8809 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
8810 forces a parameter to be padded at the lsb end in order to
8811 fill a complete number of words, require the parameter size
8812 to be ordered wrt UNITS_PER_WORD.
8814 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8815 Alan Hayward <alan.hayward@arm.com>
8816 David Sherwood <david.sherwood@arm.com>
8818 * reload1.c (spill_stack_slot_width): Change element type
8819 from unsigned int to poly_uint64_pod.
8820 (alter_reg): Treat mode sizes as polynomial.
8822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8823 Alan Hayward <alan.hayward@arm.com>
8824 David Sherwood <david.sherwood@arm.com>
8826 * reload.c (complex_word_subreg_p): New function.
8827 (reload_inner_reg_of_subreg, push_reload): Use it.
8829 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8830 Alan Hayward <alan.hayward@arm.com>
8831 David Sherwood <david.sherwood@arm.com>
8833 * lra-constraints.c (process_alt_operands): Reject matched
8834 operands whose sizes aren't ordered.
8835 (match_reload): Refer to this check here.
8837 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8838 Alan Hayward <alan.hayward@arm.com>
8839 David Sherwood <david.sherwood@arm.com>
8841 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
8842 that the mode size is in the set {1, 2, 4, 8, 16}.
8844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8845 Alan Hayward <alan.hayward@arm.com>
8846 David Sherwood <david.sherwood@arm.com>
8848 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
8849 Use plus_constant instead of gen_rtx_PLUS.
8851 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8852 Alan Hayward <alan.hayward@arm.com>
8853 David Sherwood <david.sherwood@arm.com>
8855 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
8856 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
8857 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
8858 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
8859 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
8860 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
8861 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
8862 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
8863 * config/i386/i386.c (ix86_push_rounding): ...this new function.
8864 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
8866 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
8867 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
8868 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
8869 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
8870 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
8871 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
8872 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
8873 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
8874 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
8875 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
8877 * expr.c (emit_move_resolve_push): Treat the input and result
8878 of PUSH_ROUNDING as a poly_int64.
8879 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
8880 (emit_push_insn): Likewise.
8881 * lra-eliminations.c (mark_not_eliminable): Likewise.
8882 * recog.c (push_operand): Likewise.
8883 * reload1.c (elimination_effects): Likewise.
8884 * rtlanal.c (nonzero_bits1): Likewise.
8885 * calls.c (store_one_arg): Likewise. Require the padding to be
8886 known at compile time.
8888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8889 Alan Hayward <alan.hayward@arm.com>
8890 David Sherwood <david.sherwood@arm.com>
8892 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
8893 Use plus_constant instead of gen_rtx_PLUS.
8895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8896 Alan Hayward <alan.hayward@arm.com>
8897 David Sherwood <david.sherwood@arm.com>
8899 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
8902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8903 Alan Hayward <alan.hayward@arm.com>
8904 David Sherwood <david.sherwood@arm.com>
8906 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
8907 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
8908 via stack temporaries. Treat the mode size as polynomial too.
8910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8911 Alan Hayward <alan.hayward@arm.com>
8912 David Sherwood <david.sherwood@arm.com>
8914 * expr.c (expand_expr_real_2): When handling conversions involving
8915 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
8916 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
8917 as a poly_uint64 too.
8919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8920 Alan Hayward <alan.hayward@arm.com>
8921 David Sherwood <david.sherwood@arm.com>
8923 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
8925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8926 Alan Hayward <alan.hayward@arm.com>
8927 David Sherwood <david.sherwood@arm.com>
8929 * combine.c (can_change_dest_mode): Handle polynomial
8930 REGMODE_NATURAL_SIZE.
8931 * expmed.c (store_bit_field_1): Likewise.
8932 * expr.c (store_constructor): Likewise.
8933 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
8934 and polynomial REGMODE_NATURAL_SIZE.
8935 (gen_lowpart_common): Likewise.
8936 * reginfo.c (record_subregs_of_mode): Likewise.
8937 * rtlanal.c (read_modify_subreg_p): Likewise.
8939 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8940 Alan Hayward <alan.hayward@arm.com>
8941 David Sherwood <david.sherwood@arm.com>
8943 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
8944 numbers of elements.
8946 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8947 Alan Hayward <alan.hayward@arm.com>
8948 David Sherwood <david.sherwood@arm.com>
8950 * match.pd: Cope with polynomial numbers of vector elements.
8952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8953 Alan Hayward <alan.hayward@arm.com>
8954 David Sherwood <david.sherwood@arm.com>
8956 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
8957 in a POINTER_PLUS_EXPR.
8959 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8960 Alan Hayward <alan.hayward@arm.com>
8961 David Sherwood <david.sherwood@arm.com>
8963 * omp-simd-clone.c (simd_clone_subparts): New function.
8964 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
8965 (ipa_simd_modify_function_body): Likewise.
8967 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8968 Alan Hayward <alan.hayward@arm.com>
8969 David Sherwood <david.sherwood@arm.com>
8971 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
8972 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
8973 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
8974 (expand_vector_condition, vector_element): Likewise.
8975 (subparts_gt): New function.
8976 (get_compute_type): Use subparts_gt.
8977 (count_type_subparts): Delete.
8978 (expand_vector_operations_1): Use subparts_gt instead of
8979 count_type_subparts.
8981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8982 Alan Hayward <alan.hayward@arm.com>
8983 David Sherwood <david.sherwood@arm.com>
8985 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
8986 (vect_compile_time_alias): ...this new function. Do the calculation
8987 on poly_ints rather than trees.
8988 (vect_prune_runtime_alias_test_list): Update call accordingly.
8990 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8991 Alan Hayward <alan.hayward@arm.com>
8992 David Sherwood <david.sherwood@arm.com>
8994 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
8996 (vect_schedule_slp_instance): Likewise.
8998 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8999 Alan Hayward <alan.hayward@arm.com>
9000 David Sherwood <david.sherwood@arm.com>
9002 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
9003 constant and extern definitions for variable-length vectors.
9004 (vect_get_constant_vectors): Note that the number of units
9005 is known to be constant.
9007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9008 Alan Hayward <alan.hayward@arm.com>
9009 David Sherwood <david.sherwood@arm.com>
9011 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
9012 of units as polynomial. Choose between WIDE and NARROW based
9015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9016 Alan Hayward <alan.hayward@arm.com>
9017 David Sherwood <david.sherwood@arm.com>
9019 * tree-vect-stmts.c (simd_clone_subparts): New function.
9020 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
9022 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9023 Alan Hayward <alan.hayward@arm.com>
9024 David Sherwood <david.sherwood@arm.com>
9026 * tree-vect-stmts.c (vectorizable_call): Treat the number of
9027 vectors as polynomial. Use build_index_vector for
9030 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9031 Alan Hayward <alan.hayward@arm.com>
9032 David Sherwood <david.sherwood@arm.com>
9034 * tree-vect-stmts.c (get_load_store_type): Treat the number of
9035 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
9036 for variable-length vectors.
9037 (vectorizable_mask_load_store): Treat the number of units as
9038 polynomial, asserting that it is constant if the condition has
9039 already been enforced.
9040 (vectorizable_store, vectorizable_load): Likewise.
9042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9043 Alan Hayward <alan.hayward@arm.com>
9044 David Sherwood <david.sherwood@arm.com>
9046 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
9047 of units as polynomial. Punt if we can't tell at compile time
9048 which vector contains the final result.
9050 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9051 Alan Hayward <alan.hayward@arm.com>
9052 David Sherwood <david.sherwood@arm.com>
9054 * tree-vect-loop.c (vectorizable_induction): Treat the number
9055 of units as polynomial. Punt on SLP inductions. Use an integer
9056 VEC_SERIES_EXPR for variable-length integer reductions. Use a
9057 cast of such a series for variable-length floating-point
9060 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9061 Alan Hayward <alan.hayward@arm.com>
9062 David Sherwood <david.sherwood@arm.com>
9064 * tree.h (build_index_vector): Declare.
9065 * tree.c (build_index_vector): New function.
9066 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
9067 of units as polynomial, forcibly converting it to a constant if
9068 vectorizable_reduction has already enforced the condition.
9069 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
9070 to create a {1,2,3,...} vector.
9071 (vectorizable_reduction): Treat the number of units as polynomial.
9072 Choose vectype_in based on the largest scalar element size rather
9073 than the smallest number of units. Enforce the restrictions
9076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9077 Alan Hayward <alan.hayward@arm.com>
9078 David Sherwood <david.sherwood@arm.com>
9080 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
9081 number of units as polynomial.
9083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9084 Alan Hayward <alan.hayward@arm.com>
9085 David Sherwood <david.sherwood@arm.com>
9087 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
9088 * target.def (autovectorize_vector_sizes): Return the vector sizes
9089 by pointer, using vector_sizes rather than a bitmask.
9090 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
9091 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
9092 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
9094 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
9095 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
9096 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
9097 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
9098 * omp-general.c (omp_max_vf): Likewise.
9099 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9100 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
9101 * tree-vect-loop.c (vect_analyze_loop): Likewise.
9102 * tree-vect-slp.c (vect_slp_bb): Likewise.
9103 * doc/tm.texi: Regenerate.
9104 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
9106 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
9107 the vector size as a poly_uint64 rather than an unsigned int.
9108 (current_vector_size): Change from an unsigned int to a poly_uint64.
9109 (get_vectype_for_scalar_type): Update accordingly.
9110 * tree.h (build_truth_vector_type): Take the size and number of
9111 units as a poly_uint64 rather than an unsigned int.
9112 (build_vector_type): Add a temporary overload that takes
9113 the number of units as a poly_uint64 rather than an unsigned int.
9114 * tree.c (make_vector_type): Likewise.
9115 (build_truth_vector_type): Take the number of units as a poly_uint64
9116 rather than an unsigned int.
9118 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9119 Alan Hayward <alan.hayward@arm.com>
9120 David Sherwood <david.sherwood@arm.com>
9122 * target.def (get_mask_mode): Take the number of units and length
9123 as poly_uint64s rather than unsigned ints.
9124 * targhooks.h (default_get_mask_mode): Update accordingly.
9125 * targhooks.c (default_get_mask_mode): Likewise.
9126 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
9127 * doc/tm.texi: Regenerate.
9129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9130 Alan Hayward <alan.hayward@arm.com>
9131 David Sherwood <david.sherwood@arm.com>
9133 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
9134 * omp-general.c (omp_max_vf): Likewise.
9135 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
9136 (expand_omp_simd): Handle polynomial safelen.
9137 * omp-low.c (omplow_simd_context): Add a default constructor.
9138 (omplow_simd_context::max_vf): Change from int to poly_uint64.
9139 (lower_rec_simd_input_clauses): Update accordingly.
9140 (lower_rec_input_clauses): Likewise.
9142 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9143 Alan Hayward <alan.hayward@arm.com>
9144 David Sherwood <david.sherwood@arm.com>
9146 * tree-vectorizer.h (vect_nunits_for_cost): New function.
9147 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
9148 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
9149 (vect_analyze_slp_cost): Likewise.
9150 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
9151 (vect_model_load_cost): Likewise.
9153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9154 Alan Hayward <alan.hayward@arm.com>
9155 David Sherwood <david.sherwood@arm.com>
9157 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
9158 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
9159 from an unsigned int * to a poly_uint64_pod *.
9160 (calculate_unrolling_factor): New function.
9161 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
9163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9164 Alan Hayward <alan.hayward@arm.com>
9165 David Sherwood <david.sherwood@arm.com>
9167 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
9168 from an unsigned int to a poly_uint64.
9169 (_loop_vec_info::slp_unrolling_factor): Likewise.
9170 (_loop_vec_info::vectorization_factor): Change from an int
9172 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
9173 (vect_get_num_vectors): New function.
9174 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
9175 (vect_get_num_copies): Use vect_get_num_vectors.
9176 (vect_analyze_data_ref_dependences): Change max_vf from an int *
9177 to an unsigned int *.
9178 (vect_analyze_data_refs): Change min_vf from an int * to a
9180 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9181 than an unsigned HOST_WIDE_INT.
9182 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9183 (vect_analyze_data_ref_dependence): Change max_vf from an int *
9184 to an unsigned int *.
9185 (vect_analyze_data_ref_dependences): Likewise.
9186 (vect_compute_data_ref_alignment): Handle polynomial vf.
9187 (vect_enhance_data_refs_alignment): Likewise.
9188 (vect_prune_runtime_alias_test_list): Likewise.
9189 (vect_shift_permute_load_chain): Likewise.
9190 (vect_supportable_dr_alignment): Likewise.
9191 (dependence_distance_ge_vf): Take the vectorization factor as a
9192 poly_uint64 rather than an unsigned HOST_WIDE_INT.
9193 (vect_analyze_data_refs): Change min_vf from an int * to a
9195 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
9196 vfm1 as a poly_uint64 rather than an int. Make the same change
9197 for the returned bound_scalar.
9198 (vect_gen_vector_loop_niters): Handle polynomial vf.
9199 (vect_do_peeling): Likewise. Update call to
9200 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
9201 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
9203 * tree-vect-loop.c (vect_determine_vectorization_factor)
9204 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
9205 (vect_get_known_peeling_cost): Likewise.
9206 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
9207 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
9208 (vect_transform_loop): Likewise. Use the lowest possible VF when
9209 updating the upper bounds of the loop.
9210 (vect_min_worthwhile_factor): Make static. Return an unsigned int
9212 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
9213 polynomial unroll factors.
9214 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
9215 (vect_make_slp_decision): Likewise.
9216 (vect_supported_load_permutation_p): Likewise, and polynomial
9218 (vect_analyze_slp_cost): Handle polynomial vf.
9219 (vect_slp_analyze_node_operations): Likewise.
9220 (vect_slp_analyze_bb_1): Likewise.
9221 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9222 than an unsigned HOST_WIDE_INT.
9223 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
9224 (vectorizable_load): Handle polynomial vf.
9225 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
9227 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
9229 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9230 Alan Hayward <alan.hayward@arm.com>
9231 David Sherwood <david.sherwood@arm.com>
9233 * match.pd: Handle bit operations involving three constants
9234 and try to fold one pair.
9236 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9238 * tree-vect-loop-manip.c: Include gimple-fold.h.
9239 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
9240 niters_maybe_zero parameters. Handle other cases besides a step of 1.
9241 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
9242 Add a path that uses a step of VF instead of 1, but disable it
9244 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
9245 and niters_no_overflow parameters. Update calls to
9246 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
9247 Create a new SSA name if the latter choses to use a ste other
9248 than zero, and return it via niters_vector_mult_vf_var.
9249 * tree-vect-loop.c (vect_transform_loop): Update calls to
9250 vect_do_peeling, vect_gen_vector_loop_niters and
9251 slpeel_make_loop_iterate_ntimes.
9252 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
9253 (vect_gen_vector_loop_niters): Update declarations after above changes.
9255 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
9257 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
9258 128-bit round to integer instructions.
9259 (ceil<mode>2): Likewise.
9260 (btrunc<mode>2): Likewise.
9261 (round<mode>2): Likewise.
9263 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9265 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
9266 unaligned VSX load/store on P8/P9.
9267 (expand_block_clear): Allow the use of unaligned VSX
9268 load/store on P8/P9.
9270 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9272 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
9274 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
9275 swap associated with both a load and a store.
9277 2018-01-02 Andrew Waterman <andrew@sifive.com>
9279 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
9280 * config/riscv/riscv.md (clear_cache): Use it.
9282 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
9284 * web.c: Remove out-of-date comment.
9286 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9288 * expr.c (fixup_args_size_notes): Check that any existing
9289 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
9290 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
9291 (emit_single_push_insn): ...here.
9293 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9295 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
9296 (const_vector_encoded_nelts): New function.
9297 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
9298 (const_vector_int_elt, const_vector_elt): Declare.
9299 * emit-rtl.c (const_vector_int_elt_1): New function.
9300 (const_vector_elt): Likewise.
9301 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
9302 of CONST_VECTOR_ELT.
9304 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9306 * expr.c: Include rtx-vector-builder.h.
9307 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
9308 directly on the tree encoding.
9309 (const_vector_from_tree): Likewise.
9310 * optabs.c: Include rtx-vector-builder.h.
9311 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
9312 sequence of "u" values.
9313 * vec-perm-indices.c: Include rtx-vector-builder.h.
9314 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
9315 directly on the vec_perm_indices encoding.
9317 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9319 * doc/rtl.texi (const_vector): Describe new encoding scheme.
9320 * Makefile.in (OBJS): Add rtx-vector-builder.o.
9321 * rtx-vector-builder.h: New file.
9322 * rtx-vector-builder.c: Likewise.
9323 * rtl.h (rtx_def::u2): Add a const_vector field.
9324 (CONST_VECTOR_NPATTERNS): New macro.
9325 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
9326 (CONST_VECTOR_DUPLICATE_P): Likewise.
9327 (CONST_VECTOR_STEPPED_P): Likewise.
9328 (CONST_VECTOR_ENCODED_ELT): Likewise.
9329 (const_vec_duplicate_p): Check for a duplicated vector encoding.
9330 (unwrap_const_vec_duplicate): Likewise.
9331 (const_vec_series_p): Check for a non-duplicated vector encoding.
9332 Say that the function only returns true for integer vectors.
9333 * emit-rtl.c: Include rtx-vector-builder.h.
9334 (gen_const_vec_duplicate_1): Delete.
9335 (gen_const_vector): Call gen_const_vec_duplicate instead of
9336 gen_const_vec_duplicate_1.
9337 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
9338 (gen_const_vec_duplicate): Use rtx_vector_builder.
9339 (gen_const_vec_series): Likewise.
9340 (gen_rtx_CONST_VECTOR): Likewise.
9341 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
9342 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9343 Build a new vector rather than modifying a CONST_VECTOR in-place.
9344 (handle_special_swappables): Update call accordingly.
9345 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
9346 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9347 Build a new vector rather than modifying a CONST_VECTOR in-place.
9348 (handle_special_swappables): Update call accordingly.
9350 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9352 * simplify-rtx.c (simplify_const_binary_operation): Use
9353 CONST_VECTOR_ELT instead of XVECEXP.
9355 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9357 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
9358 the selector elements to be different from the data elements
9359 if the selector is a VECTOR_CST.
9360 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
9361 ssizetype for the selector.
9363 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9365 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
9366 before testing each element individually.
9367 * tree-vect-generic.c (lower_vec_perm): Likewise.
9369 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9371 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
9372 * selftest-run-tests.c (selftest::run_tests): Call it.
9373 * vector-builder.h (vector_builder::operator ==): New function.
9374 (vector_builder::operator !=): Likewise.
9375 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
9376 (vec_perm_indices::all_from_input_p): New function.
9377 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9378 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
9379 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
9380 instead of reading the VECTOR_CST directly. Detect whether both
9381 vector inputs are the same before constructing the vec_perm_indices,
9382 and update the number of inputs argument accordingly. Use the
9383 utility functions added above. Only construct sel2 if we need to.
9385 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9387 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
9388 the broadcast of the low byte.
9389 (expand_mult_highpart): Use an explicit encoding for the permutes.
9390 * optabs-query.c (can_mult_highpart_p): Likewise.
9391 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
9392 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9393 (vectorizable_bswap): Likewise.
9394 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
9395 explicit encoding for the power-of-2 permutes.
9396 (vect_permute_store_chain): Likewise.
9397 (vect_grouped_load_supported): Likewise.
9398 (vect_permute_load_chain): Likewise.
9400 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9402 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
9403 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
9404 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
9405 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9406 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
9407 (vect_gen_perm_mask_any): Likewise.
9409 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9411 * int-vector-builder.h: New file.
9412 * vec-perm-indices.h: Include int-vector-builder.h.
9413 (vec_perm_indices): Redefine as an int_vector_builder.
9414 (auto_vec_perm_indices): Delete.
9415 (vec_perm_builder): Redefine as a stand-alone class.
9416 (vec_perm_indices::vec_perm_indices): New function.
9417 (vec_perm_indices::clamp): Likewise.
9418 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
9419 (vec_perm_indices::new_vector): New function.
9420 (vec_perm_indices::new_expanded_vector): Update for new
9421 vec_perm_indices class.
9422 (vec_perm_indices::rotate_inputs): New function.
9423 (vec_perm_indices::all_in_range_p): Operate directly on the
9424 encoded form, without computing elided elements.
9425 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9426 encoding. Update for new vec_perm_indices class.
9427 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
9428 the given vec_perm_builder.
9429 (expand_vec_perm_var): Update vec_perm_builder constructor.
9430 (expand_mult_highpart): Use vec_perm_builder instead of
9431 auto_vec_perm_indices.
9432 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
9433 vec_perm_indices instead of auto_vec_perm_indices. Use a single
9434 or double series encoding as appropriate.
9435 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
9436 vec_perm_indices instead of auto_vec_perm_indices.
9437 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9438 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9439 (vect_permute_store_chain): Likewise.
9440 (vect_grouped_load_supported): Likewise.
9441 (vect_permute_load_chain): Likewise.
9442 (vect_shift_permute_load_chain): Likewise.
9443 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9444 (vect_transform_slp_perm_load): Likewise.
9445 (vect_schedule_slp_instance): Likewise.
9446 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9447 (vectorizable_mask_load_store): Likewise.
9448 (vectorizable_bswap): Likewise.
9449 (vectorizable_store): Likewise.
9450 (vectorizable_load): Likewise.
9451 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9452 vec_perm_indices instead of auto_vec_perm_indices. Use
9453 tree_to_vec_perm_builder to read the vector from a tree.
9454 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9455 vec_perm_builder instead of a vec_perm_indices.
9456 (have_whole_vector_shift): Use vec_perm_builder and
9457 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9458 truncation to calc_vec_perm_mask_for_shift.
9459 (vect_create_epilog_for_reduction): Likewise.
9460 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9461 from auto_vec_perm_indices to vec_perm_indices.
9462 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9463 instead of changing individual elements.
9464 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9465 the vector in d.perm.
9466 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9467 from auto_vec_perm_indices to vec_perm_indices.
9468 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9469 instead of changing individual elements.
9470 (arm_vectorize_vec_perm_const): Use new_vector to install
9471 the vector in d.perm.
9472 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9473 Update vec_perm_builder constructor.
9474 (rs6000_expand_interleave): Likewise.
9475 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9476 (rs6000_expand_interleave): Likewise.
9478 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9480 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9481 to qimode could truncate the indices.
9482 * optabs.c (expand_vec_perm_var): Likewise.
9484 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9486 * Makefile.in (OBJS): Add vec-perm-indices.o.
9487 * vec-perm-indices.h: New file.
9488 * vec-perm-indices.c: Likewise.
9489 * target.h (vec_perm_indices): Replace with a forward class
9491 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9492 * optabs.h: Include vec-perm-indices.h.
9493 (expand_vec_perm): Delete.
9494 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9495 (expand_vec_perm_const): Declare.
9496 * target.def (vec_perm_const_ok): Replace with...
9497 (vec_perm_const): ...this new hook.
9498 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9499 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9500 * doc/tm.texi: Regenerate.
9501 * optabs.def (vec_perm_const): Delete.
9502 * doc/md.texi (vec_perm_const): Likewise.
9503 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9504 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9505 expand_vec_perm for constant permutation vectors. Assert that
9506 the mode of variable permutation vectors is the integer equivalent
9507 of the mode that is being permuted.
9508 * optabs-query.h (selector_fits_mode_p): Declare.
9509 * optabs-query.c: Include vec-perm-indices.h.
9510 (selector_fits_mode_p): New function.
9511 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
9512 is defined, instead of checking whether the vec_perm_const_optab
9513 exists. Use targetm.vectorize.vec_perm_const instead of
9514 targetm.vectorize.vec_perm_const_ok. Check whether the indices
9515 fit in the vector mode before using a variable permute.
9516 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
9517 vec_perm_indices instead of an rtx.
9518 (expand_vec_perm): Replace with...
9519 (expand_vec_perm_const): ...this new function. Take the selector
9520 as a vec_perm_indices rather than an rtx. Also take the mode of
9521 the selector. Update call to shift_amt_for_vec_perm_mask.
9522 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
9523 Use vec_perm_indices::new_expanded_vector to expand the original
9524 selector into bytes. Check whether the indices fit in the vector
9525 mode before using a variable permute.
9526 (expand_vec_perm_var): Make global.
9527 (expand_mult_highpart): Use expand_vec_perm_const.
9528 * fold-const.c: Includes vec-perm-indices.h.
9529 * tree-ssa-forwprop.c: Likewise.
9530 * tree-vect-data-refs.c: Likewise.
9531 * tree-vect-generic.c: Likewise.
9532 * tree-vect-loop.c: Likewise.
9533 * tree-vect-slp.c: Likewise.
9534 * tree-vect-stmts.c: Likewise.
9535 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
9537 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
9538 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
9539 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
9540 (aarch64_vectorize_vec_perm_const): ...this new function.
9541 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9542 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9543 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
9544 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
9545 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9546 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9547 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
9549 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
9550 check for NEON modes.
9551 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
9552 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
9553 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
9554 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
9556 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
9557 the old VEC_PERM_CONST conditions.
9558 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
9559 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
9560 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
9561 (ia64_vectorize_vec_perm_const_ok): Merge into...
9562 (ia64_vectorize_vec_perm_const): ...this new function.
9563 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
9564 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
9565 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
9566 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
9567 * config/mips/mips.c (mips_expand_vec_perm_const)
9568 (mips_vectorize_vec_perm_const_ok): Merge into...
9569 (mips_vectorize_vec_perm_const): ...this new function.
9570 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
9571 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
9572 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
9573 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
9574 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
9575 (rs6000_expand_vec_perm_const): Delete.
9576 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
9578 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9579 (altivec_expand_vec_perm_const_le): Take each operand individually.
9580 Operate on constant selectors rather than rtxes.
9581 (altivec_expand_vec_perm_const): Likewise. Update call to
9582 altivec_expand_vec_perm_const_le.
9583 (rs6000_expand_vec_perm_const): Delete.
9584 (rs6000_vectorize_vec_perm_const_ok): Delete.
9585 (rs6000_vectorize_vec_perm_const): New function.
9586 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9587 an element count and rtx array.
9588 (rs6000_expand_extract_even): Update call accordingly.
9589 (rs6000_expand_interleave): Likewise.
9590 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
9591 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
9592 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
9593 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
9594 (rs6000_expand_vec_perm_const): Delete.
9595 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
9596 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9597 (altivec_expand_vec_perm_const_le): Take each operand individually.
9598 Operate on constant selectors rather than rtxes.
9599 (altivec_expand_vec_perm_const): Likewise. Update call to
9600 altivec_expand_vec_perm_const_le.
9601 (rs6000_expand_vec_perm_const): Delete.
9602 (rs6000_vectorize_vec_perm_const_ok): Delete.
9603 (rs6000_vectorize_vec_perm_const): New function. Remove stray
9604 reference to the SPE evmerge intructions.
9605 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
9606 an element count and rtx array.
9607 (rs6000_expand_extract_even): Update call accordingly.
9608 (rs6000_expand_interleave): Likewise.
9609 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
9610 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
9612 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
9614 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9616 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
9617 vector mode and that that mode matches the mode of the data
9619 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
9620 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
9621 directly using expand_vec_perm_1 when forcing selectors into
9623 (expand_vec_perm_var): New function, split out from expand_vec_perm.
9625 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9627 * optabs-query.h (can_vec_perm_p): Delete.
9628 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
9629 * optabs-query.c (can_vec_perm_p): Split into...
9630 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
9631 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
9632 particular selector is valid.
9633 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9634 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9635 (vect_grouped_load_supported): Likewise.
9636 (vect_shift_permute_load_chain): Likewise.
9637 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9638 (vect_transform_slp_perm_load): Likewise.
9639 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9640 (vectorizable_bswap): Likewise.
9641 (vect_gen_perm_mask_checked): Likewise.
9642 * fold-const.c (fold_ternary_loc): Likewise. Don't take
9643 implementations of variable permutation vectors into account
9644 when deciding which selector to use.
9645 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
9646 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
9647 with a false third argument.
9648 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
9649 to test whether the constant selector is valid and can_vec_perm_var_p
9650 to test whether a variable selector is valid.
9652 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9654 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
9655 * optabs-query.c (can_vec_perm_p): Likewise.
9656 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
9657 instead of vec_perm_indices.
9658 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
9659 (vect_gen_perm_mask_checked): Likewise,
9660 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
9661 (vect_gen_perm_mask_checked): Likewise,
9663 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9665 * optabs-query.h (qimode_for_vec_perm): Declare.
9666 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
9667 (qimode_for_vec_perm): ...this new function.
9668 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
9670 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9672 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
9673 does not have a conditional at the top.
9675 2018-01-02 Richard Biener <rguenther@suse.de>
9677 * ipa-inline.c (big_speedup_p): Fix expression.
9679 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9682 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
9685 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
9689 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
9690 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
9691 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
9692 cond_taken_branch_cost 3->4.
9694 2018-01-01 Jakub Jelinek <jakub@redhat.com>
9696 PR tree-optimization/83581
9697 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
9698 TODO_cleanup_cfg if any changes have been made.
9701 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
9702 convert_modes if target mode has the right side, but different mode
9706 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
9707 last argument when extracting from CONCAT. If either from_real or
9708 from_imag is NULL, use expansion through memory. If result is not
9709 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
9710 the parts directly to inner mode, if even that fails, use expansion
9714 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
9715 check for bswap in mode rather than HImode and use that in expand_unop
9718 Copyright (C) 2018 Free Software Foundation, Inc.
9720 Copying and distribu